JP2009163753A - Method and device for routing, query and response based on reliable and efficient content in publish-subscribe network - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To achieve persistent caching of messages delivered via a publish-subscribe network. <P>SOLUTION: A publisher device 118 and a subscriber device 140 receive the messages containing data via the publish-subscribe network by an intelligent router, add a mark indicating time to the received data, cache the data to a cache memory, determine whether or not a time grain period has passed since caching of the data, transfer the cached data from the cache memory to a disk if determined that the time grain period having passed, determine whether or not a persistent time frame period has passed since caching of the data, and delete the cached data if determined that the persistent time frame period having passed. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  This application is filed on March 28, 2002, US Provisional Application No. 60 / 368,833, entitled "Reliable Publish / Subscribe System Architecture in Reliable Networks". System Architecture over Unreliable Networks), US Provisional Application No. 60 / 369,105, filed March 28, 2002, entitled “Methods for Propagating Content Filters for Issuing / Subscribing Networks” Content Filters for a Publish / Subscribe Network), filed March 28, 2002, US Provisional Application No. 60 / 368,831, entitled “Practical Content-Based Packet Routing in a Publish / Subscribe Network” Practical Method for Content-Based Packet Routing in a Publish / Subscribe Network ”, March 2, 2002 US Provisional Application No. 60 / 368,870, filed on the same day as the title of the invention "Mapping query data as a subscription and mapping a query as a notification, enabling query-response interaction in the publish-subscribe infrastructure (Implementing Query-Response Interactions On a Publish-Subscribe Infrastructure By Mapping Data Advertisements as Subscriptions and Queries as notifications), US Provisional Application No. 60 / 369,832, filed April 3, 2002, invention “Method and Apparatus for Implementing Persistent and Reliable Message Delivery”, US Provisional Application No. 60, filed on Feb. 19, 2003. / 447,782, title of invention "Propagation of content filter, compact file Publish-subscribe infrastructure by filter storage and offline pre-computation, reliable publishing and subscription, enabling sustained and reliable message delivery, mapping data advertisements as subscriptions, and mapping queries as notifications Propagating Content Filters, Content-Based Packet Routing Using Compact Filter Storage and Off-Line Pre-computation, Reliable Publishing and Subscribing, Implementing Persistent and Reliable Message Delivery, and Implementing Query- Response Interactions By "Mapping Data Advertisements as Subscriptions and Queries as Notifications in a Publish-Subscribe Network" ", and these patent documents are hereby incorporated by reference.

  Further, US patent application Ser. No. 10 / 199,356, title of invention “Packet Routing Via Payload Inspection”, US patent application Ser. No. 10 / 199,368, title of invention “Channel in Router”. Method and Apparatus for Content-Based Routing And Filtering At Routers Using Channels ", US Patent Application No. 10 / 199,439, entitled" Network and Networking " Method For Sending And Receiving A Boolean Function Over A Network ”, US Patent Application No. 10 / 199,369, entitled“ Evaluation, Modification, Reuse and Reuse Through Network ” Method For Storing Boolean Functions To Enable Evaluati on, Modification, Reuse, And Delivery Over A Network), US patent application Ser. No. 10 / 199,388, title of invention “Efficient implementation of wildcard matching in variable length fields in content-based routing (Efficient Implementation of Wildcard Matching On Variable-Sized Fields In Connect-Based Routing) is also incorporated herein by reference.

  Network bandwidth is growing exponentially. However, the network infrastructure (including routers, servers, daemons, protocols, etc.) still uses relatively old technology. As a result, Internet applications and network routers cannot cope with the speed of bandwidth expansion. At the same time, devices and applications corresponding to the network are increasing. The load on the network node by these devices and applications is significantly increased. Also, the complexity of implementing and maintaining network applications is increasing due to the increase in network load and applications. Increased network bandwidth and ubiquitous use of network equipment and applications has led to issues related to routing and transmission of data in older network infrastructures, particularly publishing content to subscribers. Has occurred.

  The model of pushing information from a server to a client in a network is called a publish-subscribe style. In this model, the server acts as a simplified publisher of that information regardless of which clients are interested in the information or where the client is located in the network. Clients become subscribers of information, potentially delivering that information as it becomes available, regardless of the details about where the information was made available on the network. The network is responsible for efficiently routing published information to subscribers, collating information for active subscriptions, and performing these processes in a transparent manner for publishers and subscribers.

  In the publish-subscribe model, the server configuration is greatly simplified, so there is no distinction between a heavyweight server and a lightweight client, which can be issued to issuers, subscribers, or both. Can be integrated into the designation of peers. Various types of applications have characteristics close to publish-subscribe interaction between peers. In the issues that exist at the base of many of these applications, published and subscribed information is provided in the form of events. For example, the stock price fluctuates as an investor purchases or sells stock. In addition, when a traffic accident occurs on a highway, traffic congestion occurs. When security holes in software systems are discovered, patches need to be developed for software users. When an internet game player uses a weapon, the other player's avatar dies. These exemplary phenomena are all events that are potentially of interest to a large number of subscribers and are propagated within the network to inform these subscribers that such events have occurred. ). That is, an event is a self-contained and succinct piece of information about something of potential interest that occurred at some point on the network at a certain time. .

  Another example is scheduled broadcast, which has a different nature than an application that includes only asynchronous events that are unpredictable at the time of event occurrence. In this case, the event is first scheduled to occur at a known time. Second, an event need not necessarily be a brief piece of information. Alternatively, the event may be a large amount of data. Substantial server processing is required to send such a large amount of data to the part of the network where the interested subscribers are detected.

  In the publish-subscribe model, a routing decision is usually made by the server or issuer to instruct the network where to send the published content. The issuer stores a subscription related to the content that the issuer issues. When the issuer receives or generates new content, the issuer compares the content with each subscription and checks for any matches. If the content (event) satisfies any subscription, the publisher pushes the content to the corresponding subscriber over the network. In this conventional publish-subscribe model, especially when more devices are compatible with the network and the number of subscriptions is increased, the issuer is heavily burdened. Alternatively, there is a way for subscribers to evaluate their subscriptions for all published events, but this is equally cumbersome.

  As the concentration of countless applications on the Internet increases, the possibilities for using event notifications are infinite. Now, because of these possibilities, it is necessary to implement a method to reduce the issuer's burden by efficiently determining the route and efficiently determining when the event satisfies the subscription. It is. Thus, the realization of a universal and persistent event notification service adds significant value to Internet applications and other applications and implementations.

JP 2001-285287 A JP 2001-291044 A Bill Segall, David Arnold, Julian Boot, Micheal Henderson, Ted Phelps “Content Based Routing with Elvin4” In Proc. AUUG2K (June 2000) Guruduth Banavar and five others "An Efficient Multicast Protocol for Content-Based Publish-Subscribe Systems" INTERNATIONAL CONFERENCE ON DISTRIBUTED COMPUTING SYSTEMS, 1999

  A communication method and apparatus for propagating a filter in an publish-subscribe network is provided. First, a plurality of filters related to content subscription in the network are received. Based on certain criteria, reduce the number of filters and propagate the reduced filters in the network to match content subscriptions.

  In addition, a communication method and a communication device for realizing query-response interaction in an issue-subscription network are provided. First, an advertisement associated with a data set and a query representing a logical expression are received. The advertisement is then mapped to the corresponding subscription reservation. The query is then mapped to the corresponding notification. These corresponding subscriptions and notifications are used to implement advertisements and queries in the network.

  Also provided are a communication method and a communication device that realize the sustainability of messages distributed via a network. First, a message is received via the network. The message is then saved for later retrieval. Provide persistent and reliable delivery of messages in response to network-related failures.

  Other objects of the present invention and specific advantages obtained by the present invention will become more apparent from the embodiments described below with reference to the drawings.

  The present invention is also a communication method for executing issue-subscribe processing on a low-reliability network, the step of receiving a content subscription via the network, and the content subscription A step of issuing the content via the network; a step of receiving a notification regarding the content at a node in the network; and a step of determining whether the notification should be transferred to an adjacent node. And selectively transferring a notification to the adjacent nodes using a reliable transmission protocol based on the determination.

  In addition, the present invention is a router that executes issue-subscribe processing on a low-reliability network, and includes a module that realizes the communication method.

  The present invention is also a wide area network for executing issue-subscribe processing, comprising one or more subscriber devices and a plurality of routers, each of the plurality of routers being connected via the wide area network. A subscription reservation receiving module for receiving a subscription reservation for content corresponding to one or more of the subscriber devices, a notification receiving module for receiving a notification regarding the content at a node in the network, and the notification adjacent to each other. A content-based routing module that determines whether or not to forward to a certain node, and a reliable transmission that selectively forwards notifications to the adjacent nodes using a reliable transmission protocol based on the determination And a issuing module having a forwarding module, wherein the plurality of routers are one or two corresponding to the subscription reservation. Until notification to a plurality of subscribers are delivered, characterized by selectively transferring reliably the notification.

  The present invention is also a communication method for executing issue-subscribe processing on an unreliable network, wherein a first node in the network receives a content subscription via the network. Using the content subscription reservation based on content-based routing (CBR) at the first node in the network, receiving a notification about the content, Determining whether or not to forward the notification to an adjacent second node, and selectively transmitting the notification to the adjacent node using a reliable transmission protocol based on the determination. And a transferring step.

  In addition, the present invention is a router that executes issue-subscribe processing on a low-reliability network, and includes a module that realizes the communication method.

  The present invention is also a wide area network (WAN) for executing issue-subscribe processing, comprising one or more subscriber devices and a plurality of routers, each of the plurality of routers being in the WAN. A subscription subscription receiving module for receiving subscription subscriptions of content corresponding to one or more of the subscriber devices via the WAN, and content subscription at the first node in the WAN. Using a notification receiving module for receiving notifications, a content-based routing module for determining whether the notification should be forwarded to a second node in the WAN, and using a reliable transmission protocol based on the determination And an issuance module having a reliable transfer module for selectively transferring a notification to the second node. Data, until notification to one or more subscribers correspond to the subscription is delivered, wherein the selectively transferring reliably the notification.

  The present invention is also a communication method for executing issue-subscribe processing on an unreliable network, and establishes a reliable tunnel between two adjacent nodes using a reliable transmission protocol. Receiving a subscription to the content via the network at a first of the adjacent nodes, and receiving a notification regarding the content at the first node in the network And determining whether or not to forward the notification to the adjacent second node using the content subscription based on content-based routing (CBR) Based on the step and the determination, the trusted tunnel is used to communicate with the adjacent second node. Characterized by a step of transferring.

  In addition, the present invention is a router that executes issue-subscribe processing on a low-reliability network, and includes a module that realizes the communication method.

  In addition, the present invention is a network that realizes issue-subscribe processing, and includes a plurality of the routers.

  The present invention is also a communication method for propagating a filter in an publish-subscribe network, the method comprising: receiving a plurality of filters related to subscription subscription of content in the network; and the filter based on a specific criterion. And a step of propagating the reduced number of filters in the network to match subscriptions of the content.

  Further, the present invention is a router that propagates a filter in an publish-subscribe network and includes a module that realizes the communication method.

  In addition, the present invention is an issue-subscribe network, and includes a plurality of routers including modules for realizing the communication method.

  According to another aspect of the present invention, there is provided an issue-subscription network including a user device including a module for realizing the communication method.

  The present invention also provides a communication device for propagating a filter in an publish-subscribe network, wherein the filter receiving module receives a plurality of filters related to content subscription in the network, based on a specific criterion, A filter reduction module for reducing the number of filters, and a filter propagation module for propagating the filter with the reduced number in order to check the subscription of the content in the network.

  The present invention is also a communication method for propagating a filter in an issue-subscribe network, comprising: (a) receiving a plurality of filters at a node of the network; and (b) one or more of the received ones Processing the filter and reducing the number of filters; (c) deciding whether to propagate the filter based on the rules of behavior of the receiver; and (d) to the next node in the network And propagating the processed filter.

  According to another aspect of the present invention, there is provided a router for propagating a filter in an publish-subscribe network, comprising a module for realizing the communication method.

  In addition, the present invention is an issue-subscribe network, and includes a plurality of routers including modules for realizing the communication method.

  In addition, the present invention is an issue-subscription network, and includes a plurality of routers including user devices for realizing the communication method.

  The present invention is also a communication device for propagating a filter in an issue-subscription network, a filter receiving module that receives a plurality of filters at a node in the network, and a filter reduction module that reduces the number of filters. A filter propagation module for deciding whether or not to propagate a filter based on the rules of behavior of the receiver and propagating the reduced number of filters to the next node in the network It is characterized by providing.

  The present invention is also a communication method for propagating a filter in an publish-subscribe network, the step of transmitting a notification requesting a filter from a router, the step of receiving a notification by one or more downstream routers, Propagating the filter request further downstream, waiting for a response from the downstream router to which the filter request was propagated, and collecting a filter from the downstream router to which the filter request was propagated Processing the propagated filter to reduce the number of filters, and transmitting the reduced filter to the upstream router that requested the filter. And

  In addition, the present invention is an issue-subscription network, and includes a plurality of routers and an issuer apparatus including a module that realizes the communication method.

  The present invention is also a communication method for performing content-based routing of packets in an publish-subscribe network, the step of receiving packets in the network, and the step of accessing a map that specifies a filter application range in an attribute space. Inspecting the contents of the packet for route determination of the packet, using the map to assist in route determination of the packet, and based on the content and map of the inspected packet, The method includes a step of determining a route and a step of routing a packet based on the route determination.

  In addition, the present invention is a router that performs content-based routing of packets in an issue-subscribe network, and includes a module that realizes the communication method.

  According to another aspect of the present invention, there is provided a network including the plurality of routers.

  In addition, the present invention is a computer-readable medium having an instruction for realizing the communication method.

  The present invention is also a communication method for performing content-based routing of packets, comprising: (a) spatial quantization of an attribute space; (b) a plurality of filters each having at least one described attribute The content-based routing (CBR) table is updated according to processing rules described by a plurality of filters covering the grid cells identified in step (b) and (c) step (b). And a step of performing.

  In addition, the present invention is a communication device that performs content-based routing of packets in an issue-subscription network, and includes a module that realizes the communication method.

  In addition, the present invention is a network, and includes a plurality of routers including modules that implement the communication method.

  In addition, the present invention is a computer-readable medium having an instruction for realizing the communication method.

  The present invention is also a communication method for realizing query-response interaction in an publish-subscribe network, the method comprising: receiving an advertisement related to a data set; receiving a query representing a logical expression; Mapping to a corresponding subscription, mapping the query to a corresponding notification, implementing an advertisement using the corresponding subscription, and querying in the network using the corresponding notification And a step of realizing.

  The present invention also provides a communication method for realizing query-response interaction in an publish-subscribe network, the step of creating a subscription for registering an advertisement, the step of issuing a notification and distributing the query, Performing content-based routing (CBR) using the subscription and pushing notifications to the advertiser and returning response data to the requester.

  In addition, the present invention is an issue-subscribe network that realizes query-response interaction, and includes a router including a module that realizes the communication method, an issuer device, and a subscriber device.

  The present invention is also a communication method for realizing query-response interaction in an issue-subscription network, the method comprising: receiving a subscription reservation for registering an advertisement; receiving a notification for distributing a query; And performing content-based routing to determine whether or not to forward notifications to adjacent nodes and selectively forwarding notifications based on the determination. And

  In addition, the present invention is a communication device that realizes query-response interaction in an issue-subscribe network, and includes a module that realizes the communication method.

  The present invention is also a communication method for realizing query-response interaction in an publish-subscribe network, issuing a notification including an initial subject, and issuing a notification including a past subject related to a previously issued advertisement. Submitting a subscription reservation for a past subject that is a query to a previously published ad, and performing content-based routing on the past subject using the subscription reservation to the previously published advertisement to the consumer And a step of pushing to

  In addition, the present invention is an issue-subscribe network that realizes query-response interaction, and includes a router including a module that realizes the communication method, an issuer device, and a subscriber device.

  The present invention is also a communication method for realizing query-response interaction in an issue-subscribe network, receiving a notification including an initial subject and a notification including a past subject related to a previously issued advertisement; Receiving a subscription reservation for an initial subject and a subscription reservation for a past subject that is a query to a previously published advertisement, and performing content-based routing using the subscription reservation for the past subject, Determining whether or not to forward an advertisement issued to a neighboring node, and selectively forwarding the previously issued advertisement based on the determination. .

  In addition, the present invention is a communication device that realizes query-response interaction in an issue-subscribe network, and includes a module that realizes the communication method.

  The present invention is also a communication method for realizing query-response interaction in an publish-subscribe network, the step of caching issued notifications, and the notification for advertising the contents in the cached issued notifications. , Submitting a subscription that is a query to the published notification, replaying the cached published notification from the cache, and based on the subscription Performing content-based routing (CBR) and pushing the replayed published notification to a consumer.

  In addition, the present invention is an issue-subscribe network that realizes query-response interaction, and includes a router including a module that realizes the communication method, an issuer device, and a subscriber device.

  The present invention also provides a communication method for realizing query-response interaction in an publish-subscribe network, the step of caching issued notifications, and the step of receiving a subscription reservation requesting the issued notifications; Replaying the cached issued notification from the cache, and performing content-based routing (CBR) using a subscription reservation that is a query to the issued notification, the cached, Determining whether to forward an issued notification to an adjacent node, and selectively transferring the notification to the adjacent node using a reliable transmission protocol based on the determination And a step.

  In addition, the present invention is a communication device that realizes query-response interaction in an issue-subscribe network, and includes a module that realizes the communication method.

  The present invention also provides a communication method for realizing query-response interaction in an publish-subscribe network, the step of receiving a data advertisement, the step of converting the data advertisement into a subscription reservation, and the step of receiving a query. , Converting the query into a notification, performing content-based routing (CBR) based on the subscription, pushing the notification to the advertiser, and sending response data to the requester based on the notification And a step of returning.

  In addition, the present invention is an issue-subscribe network that realizes query-response interaction, and includes a router including a module that realizes the communication method, an issuer device, and a subscriber device.

  The present invention is also a communication method for realizing query-response interaction in an publish-subscribe network, wherein the method receives a query, converts the query into a notification, and converts the query into a notification and subscription. Receiving response data based on content-based routing performed on the data advertisement.

  In addition, the present invention is a communication device that realizes query-response interaction in an issue-subscribe network, and includes a module that realizes the communication method.

  The present invention is also a communication method for realizing query-response interaction in an publish-subscribe network, which is performed for receiving a data advertisement, converting the data advertisement into a subscription reservation, and performing subscription reservation and notification. Receiving a query transformed into a notification based on the content-based routing, and transferring response data to the requester based on the notification.

  In addition, the present invention is a communication device that realizes query-response interaction in an issue-subscribe network, and includes a module that realizes the communication method.

  The present invention is also a communication method for realizing query-response interaction in an publish-subscribe network, receiving a notification changed from a query, content-based routing (CBR), and adjacent. The method includes a step of determining whether or not to transfer a notification to a node, and a step of selectively transferring the notification based on the determination.

  In addition, the present invention is a communication device that realizes query-response interaction in an issue-subscribe network, and includes a module that realizes the communication method.

  The present invention also provides a communication method for realizing persistent caching of a message provided via an publish-subscribe network, wherein (a) a message including data is transmitted via the network at a first node. Receiving, (b) attaching a mark indicating time to the data, (c) caching the data in the cache memory of the first node, and (d) using content-based routing. And routing the message to the second node.

  In addition, the present invention is a router that realizes continuous caching of a message distributed via an issue-subscription network, and includes a module that realizes the communication method.

  In addition, the present invention is an publish-subscribe network that realizes persistent caching of messages, and includes a node that includes a module that implements the communication method.

  In addition, the present invention is a computer-readable medium having an instruction for realizing the communication method.

  The present invention also provides a communication method for realizing continuous caching of a message delivered via an publish-subscribe network, the method comprising: receiving a message including data via the publish-subscribe network; Adding a mark indicating time to the data; caching the data in the cache memory; determining whether a time grain period G has elapsed from the step of caching the data; and the time grain If it is determined that the period G has elapsed, the sustained time frame period T has elapsed from the step of transferring the cached data from the cache memory to the disk and the step of caching the data for the last block of the cached data. Determine whether or not A step that, if it is determined that the duration time frame period T has elapsed, characterized in that a step of erasing the cached data.

  In addition, the present invention is a router that realizes continuous caching of a message distributed via an issue-subscription network, and includes a module that realizes the communication method.

  The present invention is also an publish-subscribe network that realizes persistent caching of messages, each comprising a first node comprising a module for executing the communication method and a downstream side of the first node. 2 nodes.

  In addition, the present invention is a computer-readable medium having an instruction for realizing the communication method.

  The present invention is also a communication method for realizing continuous caching of a message delivered via an publish-subscribe network, wherein (a) a plurality of upstream nodes have data via the network. (B) giving time information to the data; (c) caching the data in a cache at one upstream node of the plurality of upstream nodes; (D) routing the message to the edge node using content-based routing; and (e) steps (a) through (a) at the edge node, except that the data is cached in the edge node cache memory. c) repeating step And wherein the Rukoto.

  Furthermore, the present invention is a publish-subscribe network that realizes persistent caching of messages, and is characterized by comprising a plurality of upstream nodes and edge nodes that comprise modules that implement the communication method.

  According to the present invention, it is possible to provide a communication method and a communication apparatus for propagating a filter in an publish-subscribe network.

  In addition, according to the present invention, it is possible to provide a communication method and a communication apparatus that realize query-response interaction in an issue-subscription network.

  In addition, according to the present invention, it is possible to provide a communication method and a communication device that realize the sustainability of a message distributed via a network.

  Furthermore, the present invention can provide a persistent and reliable delivery of messages in response to network related failures.

It is a block diagram which shows notionally this intelligent routing in a network core. 1 is a network diagram illustrating an intelligent router for issuers and subscribers. FIG. FIG. 2 illustrates an example network infrastructure configuration for intelligent routers and backbone routers. It is a figure which shows the hardware constitutions of an intelligent router. It is a figure which shows the structure of an issuer apparatus and a subscriber apparatus. It is a figure which shows the structure of the channel manager of an intelligent router. It is a figure which shows the structure of the software component of the user apparatus for connecting to the network which has an intelligent router. It is a figure which shows the structure of the software component for intelligent routers. FIG. 3 shows a packet structure for a message. It is a flowchart of an issuing process. It is a flowchart of a subscription process. It is a figure which shows a channel and a subscriber screen. It is a flowchart of a content-based routing process. It is a flowchart of a caching process. It is a figure which shows a cache index. It is a flowchart of the agent process for a transmission message. It is a flowchart of the agent process for a received message. It is a figure explaining the specific example of encoding of a message. It is a figure which shows the database structure for preserve | saving a subscription reservation. It is a flowchart of a wild card process. It is a flowchart of the process which improves the reliability of the notification in an issue-subscription network. It is a flowchart of a filter propagation process. It is a flowchart of a filter reduction process. It is a flowchart of another filter propagation process. FIG. 2 is a diagram illustrating an issue-subscribe network topology. It is a flowchart of a content-based routing process using pre-computation. It is a flowchart of a content-based routing process using pre-computation. It is a flowchart of the query response process using issue-subscription network. FIG. 24B is a sequence diagram for further explaining the processing of FIG. 24A. It is a flowchart of the other query response processing using issue-subscription network. FIG. 24D is a sequence diagram for further explaining the processing in FIG. 24C. It is a flowchart of the other query response processing using issue-subscription network. It is a sequence diagram which further demonstrates the process of FIG. 24E. It is a flowchart of the other query response processing using issue-subscription network. It is a flowchart of the other query response processing using issue-subscription network. FIG. 25 is a sequence diagram for further explaining the processes in FIGS. 24G and 24H. 1 is a block diagram illustrating a portion of a publish-subscribe network that provides persistence by caching. FIG. It is a figure which shows the specific example of a cache manager routine. It is a flowchart of a continuous caching process. It is a flowchart of a continuous message search process.

  Hereinafter, the best mode for carrying out the present invention will be described in detail with reference to the drawings.

  Internet-wide or other distributed network-wide event notification systems provide powerful and flexible publish-subscribe networking in applications. In this system, an application program uses an event notification application program interface (hereinafter referred to as API) to issue and / or subscribe to or receive notifications regarding events occurring within the network.

  Notifications in this system are given a subject, which is represented as a string or other data structure for classifying the type of information that the notification encapsulates. The notification is completed by a set of attributes including information specific to the notification. For example, the application is subject quote. You may issue the notification regarding the transaction in the New York Stock Exchange using nyse (subject quotes.nyse), an attribute symbol, and a price. For example, the application issues individual notifications with specific attribute values including SNE (Sony Corporation quote receiver symbol) and $ 85.25 stock price. Most, if not all, notification attributes are predetermined in the sense that they are found in all notifications for the same subject family. Note that the issuer can add any attribute to the notification for each notification or in other units to add event specific information. Accordingly, some or all of the attributes may not be predefined.

  In this system, subscribers are not limited to subscribing to subjects or entire channels. Channels will be explained and defined later. A channel can include, for example, a hierarchical structure that specifies one or more levels of subject fields and associated sub-fields (sub-subjects). Therefore, the subscriber can specify the expression adjusted more precisely according to the interest by specifying the filter for each content with respect to the attribute of the notification. For example, a subscriber has a subject quote .. with a SNE symbol and a price of $ 90.00 or more. All notifications may be subscribed to nyse (eg, indicating an opportunity to sell stock owned by the subscriber). All notifications that match the subscription are delivered to the subscriber when the callback or subscriber subscribes to the subscription, or through other types of features identified at other times. A single subscription can be divided into multiple filters.

  Callbacks can perform a lot of processing, for example, from simple processing such as sending a message to a terminal device or sending an email, to selling a portion of a stock, etc. Start complex publishing and new publish-subscribe activities (eg, replace existing subscriptions with new subscriptions with a $ 75.00 share opportunity, or modify subscriber portfolio) A more complicated process such as issuing a new notification).

  Publishing and subscription activities in the application may be supported by an agent, for example. The agent may use a proxy or may be implemented as a proxy. The agent provides a connection for delivery of sent notifications and subscriptions, and matching notifications received by subscribers. When a notification is sent to the network, the router's network of this system propagates the notification to all subscribers whose subscription matches the notification. One technique for realizing this process is a technique in which notifications of all points in the network are broadcast and an application agent determines whether or not the notification is relevant to the corresponding subscriber. However, this is not necessarily a scalable technique. Typically, a network cannot function properly due to message traffic load, especially if there are a large number of active subscribers. Furthermore, even if the bandwidth is sufficiently secured, it is considered that the subscriber cannot process a huge amount of notifications.

  The network shown as an example of a system according to the present invention routes notifications very efficiently. First, the network can use multicast routing to propagate notifications, for example, at most once on all links in the network. The network can then employ a number of advanced optimization processes for filters, thereby reducing the number of notification propagations as much as possible.

  FIG. 1 is a diagram conceptually showing this intelligent routing in the network core. The issuer 14 transmits the content of the message via the edge router 16 to the network core 10 used in the issue-subscribe network. The publish-subscribe network may be any type of network for routing data or content from the publisher to the subscriber. Content is transmitted over one or more channels 18 that represent logical connections between routers or other devices. The intelligent router 12 in the network core 10 determines whether to route or forward the message. Specifically, the intelligent router 12 can determine whether the message includes content that is subscribed to by the subscriber 24.

  Each subscription reservation encapsulates a subject filter and an attribute filter. The router can extend the subject filter to a matching set of subjects, or merge attribute filters for each subject. The intelligent router evaluates the subject filter for the notification subject and the attribute filter for the attribute value in the notification. In the syntax for the subject filter, wildcards can be used, and in the syntax for the attribute filter, Boolean expressions can be used, which will be described in more detail later. The term “filter” means a set of events that a subscriber wants to receive from an publisher. The routing rule is generated from the filter, and the intelligent router performs route determination using this rule.

  Thus, if the message 26 does not satisfy the entire filter set, for example, the intelligent router 12 excludes (deletes) the message 26, which means that the message is not forwarded. If it is determined, based on the subject and attribute filter evaluation, that the message 20 satisfies part or all of the filter set, for example, the intelligent router 12 passes through the edge router 22 and possibly other devices. To route (forward) the message 20 to the subscriber or perform other functions on the message 20 within the intelligent router 12 based on all routing and / or operating rules defined for the matched filter. . The search continues until the processing on the entire set of filters is complete or until a decision on all rules is obtained, and ends when either of these two conditions is met.

  Such intelligent content-based routing at the network core provides real-time data delivery for alerts and updates, for example. Specific examples of real-time data distribution for warnings are not limited to the following, but include, for example, stock prices, traffic information, news, travel, weather, fraud detection, security, telematics, factory automation, and supply. There are chain management and network management. Specific examples of real-time data delivery for updates include, but are not limited to, software updates, antivirus updates, movie and music delivery, workflow, storage management and cache consistency. Etc. The system for distributing subscribed information can be applied to various other fields.

  Table 1 shows the subscription reservations saved for filtering, along with the subject and predicate. These may be stored in any part of the network using any kind of data structure if desired or required. As will be described later, the predicate is an element of a subscription reservation. The subscription reservation may be expressed in any format, and a specific example is shown below.

  Table 2 shows specific examples of issuance and subscription reservations for a quote server. These are merely illustrative examples for purposes of illustration, and subscriptions may include any number and type of parameters for any type of data or content.

  The predicate provides a logical expression for subscription reservation, and the subject indicates a channel related to subscription reservation. Subscription reservations can be expressed in various other ways. A logical expression is an example of such a technique, and by using this, a function for converting a subscription reservation into a subject filter and an attribute filter can be easily realized for content-based routing. Alternatively, a subscription can be expressed without referring to the subject, but by using the subject or channel (discussed in detail later), the attribute can be interpreted and the filter applied to the attribute. Context is realized.

  In the network core, route determination can be performed and distributed via the network, thereby reducing the processing burden on the issuer and subscriber devices and significantly increasing network efficiency. In FIG. 1, only a single issuer, a single subscriber, and a single intelligent router are shown for the purpose of explanation. Also good. The term intelligent router refers to a router or other entity that has the ability to determine a route by examining the payload of a packet or message at the network core or other location.

Network Infrastructure FIG. 2 is a network diagram showing an intelligent router for issuers and subscribers. For example, a routing entity 30 that provides services to a channel is effectively layered within the network infrastructure to route messages between intelligent routers, as described below. For example, publisher 32 conceptually encodes content for network transmission via channel service 30 and application 34 for receiving published content indication information, eg, a pointer to search for content. And an agent 36 for A group of logically interconnected intelligent routers 38, 40, 42, 44, 46, 48 uses the routing rules generated from the subject filters and attribute filters for subscription reservations to route content from the publisher. Route. The plurality of links 39, 41, 43, 45 logically connect the intelligent routers 38, 40, 42, 44, 46, 48. The other links 37 and 47 logically connect between the issuer 32 and the intelligent router 38 and between the subscriber 54 and the intelligent router 46, respectively. The subscriber 54 includes an agent 50 for detecting and receiving the subscribed content and an application 52 for expressing the content.

  A channel can include, for example, an associated set of logical multicast connections implemented in a distributed manner. A channel in this illustrative embodiment is a collection of logically related network resources shared by publishers and subscribers to exchange content. The content is classified according to the channel subject namespace, and the resources are managed, controlled, and supplied through channel services provided by the channel manager. Multiple channels may share the same resource. Channels can provide highly scalable directory services, including but not limited to publisher and subscriber information, authentication and authorization information message types, management information, billing and payment information, etc. Is included. Also, for example, a channel can provide persistent through caching, high speed data delivery mechanisms, security, user and network management, etc. that can be provided. The channel can also be used for any other purpose.

  Filtering by an intelligent router may be performed at the network core, and route determination may be distributed within the network. Furthermore, the intelligent router can function as an edge router that connects user equipment such as an issuer or a subscriber to the network core. Also, the same device connected to the network can function as both an issuer that pushes content to subscribers and a subscriber that receives pushed content based on routing decisions in the network. The intelligent routers and channels may be connected in any configuration as needed or desired in a particular embodiment, and the configuration of FIG. 2 is merely shown as an example.

  FIG. 3 shows an exemplary network infrastructure configuration for intelligent routers and conventional backbone routers, which also shows channel logical connections. The intelligent router in this embodiment uses an existing backbone router in a network such as the Internet or another distributed network, and the intelligent router is effectively layered on the backbone router. In this embodiment, Internet Service Provider (ISP) networks 58, 59, 60 each include several backbone routers for conventional routing of messages or packets. . In the ISP networks 58, 59 and 60, a plurality of intelligent routers 61 to 70 are connected to one or more backbone routers. In addition, the intelligent routers 61 to 70 are interconnected by a plurality of links 73 to 85 exemplarily showing actual links, and can also be connected to end user equipment via these links. The intelligent routers 61 to 70 are controlled by one or more administrator machines such as an entity 71 and one or more virtual private network (hereinafter referred to as VPN) controllers such as an entity 72, for example. Is done. The ISP networks 58, 59, 60 may be connected to issuer devices and subscriber devices (not shown in FIG. 3). The backbone routers within and between ISPs 58, 59, 60 may be interconnected in any conventional manner within the existing network infrastructure.

  As shown in FIG. 3, the existing network infrastructure can be used to implement intelligent routers 61-70 and links 73-85, which provide content-based routing in the network core. Links 73-85 represent logical connections between intelligent routers 61-70 and can be implemented using, for example, existing network infrastructure or other equipment. The link may be realized by using a logical connection called a tunnel, for example. The tunnel includes hardware, possibly software, and network infrastructure to implement the link, and the tunnel may be an element of multiple channels. Channels facilitate content-based routing in intelligent routers by providing a logical configuration for specific types of content, thereby providing context for attributes transmitted over the channel . Intelligent routers can also make routing decisions without using channels, but channels improve the efficiency of content-based routing by intelligent routers in the network core.

  This exemplary embodiment includes the use of channels and links. A link represents a connection between two routers, for example intelligent routers. A channel is a network entity that includes a group (usually a number) of routers that are statically or dynamically configured by interconnecting links to implement a one-to-many or many-to-many logical connection. In particular, a channel is the highest level logical entity that describes the essential characteristics of a channel. Many subjects may exist in one channel. Each subject forms a subnetwork (such as a multicast tree) that includes a group of interconnected routers. These subject-based sub-networks can be allocated, adapted and configured in various ways. A channel that is a collection of all sub-networks and under which a subject is formed can be compared, for example, to a mesh of networks.

  FIG. 4 shows a specific configuration example of the intelligent router 92, and the intelligent router 92 may correspond to any other intelligent router having a reference numeral. The network node 90 includes an intelligent router 92 connected to a conventional backbone router 95. The intelligent router 92 includes a processor 93 connected to a memory 94 and an auxiliary storage device 97 (for example, may be realized as an independent device). Both the memory 94 and the auxiliary storage device 97 are cache data. And the application program executed by the processor 93 is saved. The auxiliary storage device 97 is a nonvolatile data storage device. As will be described later, the processor 93 is controlled by software and issues a command to the backbone router 95. In response to this command, the backbone router 95 generates a routing rule generated from the subject filter and the attribute filter for the subscription reservation. The message or packet is routed (forwarded) or not routed (deleted). Here, although the intelligent router 92 is shown as an independent device controlled by the processor, the intelligent router 92 is realized as an application specific integrated circuit (ASIC) in the backbone router 95, and the hardware is realized. , An intelligent routing function may be provided, and software may be incorporated in this hardware. Alternatively, the intelligent routing function may be realized as a combination of software and hardware in one or a plurality of routing devices.

  FIG. 5 shows a specific configuration example of the issuer device and the subscriber device. The issuer device 100 or 118 is, for example, a memory 102 that stores one or more issuer applications 104 and an agent application 106, an auxiliary storage device 112 that is a non-volatile storage device, and for inputting information or commands. An input device 108, a processor 114 for executing an application stored in the memory 102 or received from another storage device, an output device 110 for outputting information, and a display device for visually displaying information 116.

  The subscriber device 122 or the subscriber device 140, for example, inputs information or commands to a memory 124 that stores one or more applications 126 and an agent application 128, an auxiliary storage device 130 that is a non-volatile storage device, and so on. An input device 132, a processor 134 for executing an application stored in the memory 124 or received from another storage device, an output device 136 for outputting information, and a visual display of the information The display device 138 is provided. The issuer device and the subscriber device may have any configuration, may include components different from those described above, and may include more or less components than those described above.

  The publisher device 100, 118 is connected to the subscriber device 122, 140 via a network, such as the network 120 described above, and the network 120 is a distributed routing of data or content via packets or messages at the network core. Includes an intelligent router to realize. Although only two issuer devices and subscriber devices are shown here, the network 120 can be scaled to include more issuer devices and subscriber devices. The issuer device and the subscriber device may be realized as any device controlled by a processor, and these devices include, but are not limited to, for example, a server, a personal computer, a notebook computer, A personal digital assistant, telephone, mobile phone, pager, or other device is included. The network 120 including the intelligent router may be any wired or wireless distributed network that connects wired devices, wireless devices, or both. The network 120 can also be a potentially existing or conventional network infrastructure.

  FIG. 6 shows the configuration of the channel manager 150 for the intelligent router. In this embodiment, the channel manager 150 is realized by a plurality of servers 152, 154 and 156. Each server includes local storage devices 158, 160, and 162, respectively. The intelligent routers 164, 166, 168 obtain information about a specific channel from the channel manager. The channel manager can also provide data persistence, fail over functions, and other functions. This allows the channel management program to establish a database or set of databases that specify, for example, channel related information, properties for data persistence, user information for publishers and subscribers, infrastructure information, etc., at any location in the network. The infrastructure information for providing the channel service including, for example, can include identification information of intelligent routers and tunnels connecting them, channel subjects, channel attributes (name and type of each attribute), and the like. Also, the packet or message can carry channel related information including identification information of fixed attributes and variable attributes.

  The user can download the channel information while online. For example, a user can register himself using a user name and password. Once the user's logon is authenticated, the user can open (call) the channel and retrieve information about the channel from the channel manager. Publishers can use this information when publishing content, and subscribers can use this information when entering and registering subscriptions.

  In this embodiment, each channel manager 152, 154, 156 functions as a primary for each intelligent router. Specifically, in this embodiment, each intelligent router has two Internet Protocol (IP) addresses, one for the primary channel manager and the other for the backup channel. For managers. The intelligent router communicates with the channel manager and uses these IP addresses to retrieve channel information. If the primary channel manager fails, the intelligent router can communicate with the backup channel manager. As a result, the channel managers 152, 154, and 156 share data relating to channel properties and other information, as indicated by the lines connecting them. Each channel manager has a designated backup channel manager, and when a failure occurs in the operation of the channel manager, other backup channel managers can take over the processing. Devices in the network can search for channel information, for example, using commands, and Table 3 shows a specific example of this information. Alternatively, the intelligent router may have only a primary channel manager or may have more than two channel managers.

  FIG. 7 illustrates a specific example of a stack 180 of software components in a user device or device for connecting to a network having an intelligent router. The user device can be used as an issuer, a subscriber, or both, and the user device can include various devices illustrated above as specific examples. The stack 180 can include one or more user applications 182 that are used to receive subscriptions from users, receive channel information from publishers, or receive content or data to be published. Can do. Also, the user application 182 may include any other type of application that is executed by a user device or device.

  The stack 180 may include an agent 184, an event library 186, a cache library 188, a channel library 190, a messaging library 192, and a dispatcher library 194. Agent 184 provides network connection establishment or other functions, and Table 3 shows specific examples of commands implemented by agent 184 that may be used as proxy commands or other types of commands. Can do. The event library 186 records a log of events relating to user equipment or other events or information. The cache library 188 provides a local data caching function. The channel library 190 stores channel identification information and information related to the channel. The dispatcher library 194 provides a connection to the control path 196, the channel manager 198, one or more intelligent routers 200, and can include the functions illustrated in Table 4. Messaging library 192 provides a connection to data path 204.

  Specific examples of the messaging API written in the C programming language are shown in Tables 5 to 9. Tables 5 and 6 show specific examples of APIs for message transmission and retrieval. Tables 7 and 8 show specific examples of APIs for notification transmission and retrieval. Table 9 shows specific examples of APIs for transmitting and searching for control messages. The APIs and other APIs, programs, and data structures that implement the particular functions or features shown here are merely exemplary, and these functions can be any type of API or programming language written in any programming language. You may implement | achieve as another software entity.

  FIG. 8 illustrates an exemplary software component 210 for an intelligent router, such as the intelligent router described above and the intelligent router 92 shown in FIG. The software component 210 is stored in, for example, the memory 94 and can be executed by the processor 93 of the intelligent router 92. The component 210 includes, for example, a filtering daemon 212, a dispatcher 214, a routing daemon 216, and a cache manager 218. The filtering daemon 212 provides a filtering function based on routing rules in content-based routing for processing content for subscription reservation, as will be described later. The dispatcher 214 is responsible for communicating control messages that are needed to propagate the filter over the path 220, and the dispatcher 214 has a single point of entry and channel manager for the user. Provide secure sockets and improve network security. In other words, in this example, the user does not communicate directly with the channel manager. However, as a modification, the user may communicate directly with the channel manager. The dispatcher 214 obtains attributes (name-value pairs) from the channel manager using control messages.

  The routing daemon 216 provides a communication function via a data path 222 performed via a conventional backbone router or other routing device as shown in FIG. The cache manager 218 provides local caching of data at the network node that includes the corresponding intelligent router. The operation of the cache manager 218 will be described in detail later. The cache manager 218 provides a distributed caching function of data via the network core.

  Content-based routing may be implemented at the kernel level rather than the application level. The memory accessible by the kernel is independent of the memory in the application layer. In order to execute content-based routing in an application, for example, it is necessary to copy message data from the kernel storage area to the application area and switch the application context from the kernel context to the routing application context. Both of these processes cause considerable overhead. Here, if the kernel is changed to support content-based routing, the overhead described above is eliminated, and routing can be performed at higher speed.

  Such a content-based routing function in the kernel allows the routing daemon 216 to send or receive data directly or indirectly via the data path 222, depending on the specific example. The daemon is a process that is executed in the application layer and calculates in advance a content-based routing table to be inserted into the kernel. However, the kernel can determine a route using the routing table after the routing table is inserted. Similarly, the filtering daemon calculates a filtering table in advance and inserts the calculated filtering table into the kernel. In the kernel in this example, neither the routing daemon nor the filtering daemon interacts directly with the data path.

  FIG. 9 shows an example of a packet structure 230 for a message that can include subscription content. A packet or message used for content-based routing includes, for example, a header section and a payload section. The header section specifies routing or other information. The payload section identifies data or content or indicates data or content. The packet structure 230 includes an IP header 232, a user datagram protocol (hereinafter referred to as UDP) transmission control protocol (hereinafter referred to as TCP) header 234, a length value 238, one or more subjects. It includes a field 240 and one or more attributes 242. The packet structure 230 shows the basic structure for length values, subjects and attributes. Further, the packet used in the content-based routing can include other or different elements as shown in the embodiment of FIG. 18 to be described later, and the packet for the content-based routing can be used in any manner. It may be configured. The attribute can include, for example, an arbitrary attribute added to the end of the message. These arbitrary attributes are temporary information added by, for example, the issuer (or router), and are information that does not necessarily need to be transmitted using the message format defined for the channel.

Publisher and Subscriber Methodologies
FIG. 10 is a flowchart of an exemplary publishing process 250 for setting up channels and publishing content by publishers. The process 250 can be realized by a software module including the agent 106 executed by the processor 114 in the issuer apparatus 100, for example. In process 250, the issuer device agent 106 receives a proxy for the channel created by the issuer (step 252). The proxy is used for communication with the network. Agent 106 determines the message format to use on the channel via the interface (step 253). Format information can be obtained, for example, from a channel manager or other entity in the network. The agent 106 sets up a proxy for the channel using the received channel information (step 254), which includes receiving a channel attribute (step 256) and creating a notification for the channel (step 258). Including. This notification provides content to devices that are “listening” for content on the channel. Attributes define notification parameters and characteristics.

  The agent 106 sends the channel and content information identifier (ID) to the intelligent router to process the subscription at the network core or elsewhere (step 260). The issuer sets an appropriate value for the notification attribute (step 261), so that the issuer can issue content related to the notification based on the channel attribute (step 262). In step 260 to 262 in this embodiment, the notification issuance is completed. This process may include other or further steps depending on the actual individual embodiment. Here, the information related to the notification in this embodiment is divided into a series of attributes based on a predetermined order. Each attribute has a name, a position (starting from 1) in the notification, a type, and a value. Have. Alternatively, the attributes may have different characteristics depending on the particular embodiment. For example, the attributes can include predetermined attributes, optional attributes, or both.

  The intelligent router can know the attribute of the corresponding channel based on the channel ID in the packet, and this attribute determines the structure or format of the packet transmitted through the channel. Specifically, each packet can include, for example, a tag associated with other header information such as a channel ID, issuer ID, and subject. This tag can be used to map a subject to a message format number, for example, as shown in FIG. This number can be a small integer value, for example a 16-bit value. Alternatively, the subject may be mapped using any other type of number or information. By mapping the subject to the number, for example, the following advantages can be obtained. That is, this can save space in the message format and provide a uniform or standardized way to specify subjects in the message, which allows for quick identification and identification of subjects. The intelligent router may store the mapping locally, or alternatively may obtain the corresponding subject remotely by command using a number.

  Table 10 shows a structure for mapping subjects to numbers. In this specific example, integer values are used for the numbers. The subject tree parameter in Table 10 indicates that a subject can include one or more subject fields in a hierarchical relationship, for example, a subject tree can include a series of subject fields that are partitioned by a particular symbol. . Specific examples of the subject tree are shown in Table 2. For example, subject tree quotes. The nyse includes a subject “quote” and a subfield “nyse”, and these two terms are separated by a “.” (period), like a URL or other network address. Instead of using a period and URL type character string, the subject tree may be specified by any method using any character and a symbol for delimiter.

  Knowing the packet format or structure of a particular channel allows the intelligent router to quickly locate the subject and attributes in the packet or other information for content-based routing. For example, a channel can easily identify the byte positions of subjects and attributes transmitted over the channel by counting bytes in the packet. Alternatively, the intelligent router can analyze the packet to detect the location of subjects and attributes or other information. Table 11 shows a specific example of an issuer program written in the C ++ programming language. Table 12 shows a specific example of an API for creating a channel. Table 13 shows a specific example of the channel setting file maintained by the channel manager (see FIG. 6) and channel related information. Alternatively, the system may include a comprehensive channel manager that provides the IP addresses of geographically distributed servers that act as local channel managers to distribute the processing load.

  FIG. 11 is a flowchart of a subscription process 264 for receiving and processing subscription reservations. Process 264 may be implemented, for example, by a software module that includes agent 128 executed by processor 134 at subscriber device 122. In the process 264, for example, a graphical user interface (hereinafter referred to as GUI) displays information indicating available channels to the user (step 266). This process is realized by the application 126. Information identifying a channel can be received, for example, from a channel manager that provides channel-related information. Any type of application 126 may be used to display information indicating the channel based on any particular technique or format. The application receives the channel selected by the user (step 268) and calls an API or other program for the selected channel (step 270). The API displays the channel subscription options for the selected option to the user (step 272). The API receives a value for the subscription from the user (step 274) and provides the subscription to the agent 128 for processing described below (step 276).

  The parameters for subscription reservation may include predicates as shown in Table 1, for example. Each channel can use its own API to process subscriptions according to specific requirements or parameters for the corresponding channel, for example. These APIs may include, for example, web-based or Java-based APIs for receiving subscriptions, receive information about subscriptions using any type of user interface and process, and Information may be transferred to the agent application.

  FIG. 12 conceptually illustrates channel and subscriber screens or GUIs 278, 284, which are used in connection with process 264 for receiving a subscription. Screen 278 includes a plurality of sections 282 showing available channels selected by the user. When a particular channel is selected, a section 286 for receiving user values for subscription is displayed, as shown on screen 284. If the user selects section 288, the subscription is confirmed, or if the user selects section 290, the subscription is canceled. For example, the screens 278, 284 may be formatted as a hypertext markup language (HTML) web page, or any other format may be used. The screen may also contain any configuration of sections and content, including text, graphics, pictures, various colors or multimedia information, if necessary, for subscribers, if necessary. Provide a user-friendly and visually attractive interface. Further, the screen may include, for example, a toolbar 280 that provides a browser function similar to the conventional one.

  Table 14 shows a specific example of a subscriber program written in the C ++ programming language.

Payload Inspection and Content-Based Routing Through Channels FIG. 13 is a flowchart of a payload inspection process 300 for content-based routing. Process 300 is implemented in a software module that is executed by processor 93 in intelligent router 92, for example, represented as filtering daemon 212. Alternatively, this processing may be realized by an ASIC, or may be realized by a combination of hardware and software. The content-based routing shown as process 300 can be performed in an intelligent router in any part of the network, such as a network core or edge router, for example.

  In general, content-based routing involves examining the payload section of a packet to determine how to process the packet. This content-based routing process, for example, processes a list of subscriptions in any order (eg, using a filter), compares messages by subject and attribute based on routing rules, and routes messages. And executing processing in the network core. This rule may include any rule relating to a rule or filter that governs processing at the router. As a result, these route decisions can be conveyed within the network core. By representing the subject as a channel, the message format is determined so that the intelligent router can quickly detect the position of the attribute in the message, for example, by knowing the byte position of the attribute in a message or packet on a particular channel. become able to.

  In process 300, intelligent router 92 receives a packet for a message (step 302). The intelligent router 92 determines the channel ID of the corresponding message from the packet (step 304), and reads the channel attribute using the channel ID (step 306). In this embodiment, the type of channel (determined from the channel ID) determines the attribute position and data type in the packet. Channel attributes may be stored locally, for example, remotely read via a channel manager. The intelligent router 92 searches for a filter corresponding to the subscription reservation (step 308). The filter includes one or more attribute checks, often a group of attribute checks for subscription reservations. Intelligent router 92 applies the corresponding attribute check in the filter description to the attribute in the packet (step 310).

  If the result of all attribute checks in the filter description is affirmative (step 312), that is, if the attribute meets the conditions of all attribute checks, the intelligent router 92 is predetermined by the rules associated with the filter. A set of processes is executed (step 314). These processes include, for example, a process of routing a packet to the next link and / or a process of executing some operation or operation on the contents of the packet in the local router based on a rule. This operation or next link is specified, for example, in a data structure that specifies the corresponding subscription. If the rule is a link, the rule typically specifies the next network node to receive the packet, which includes intelligent routers, backbone routers, devices connected to the network, or other entities . Alternatively, the next link may be identified or associated with a subscription reservation in other ways.

  If the results of all attribute checks in the filter description are not positive (step 312), that is, if the attributes do not satisfy all attribute check conditions, a filter mismatch is declared (step 315). The intelligent router repeats the above processing until all attribute checks in the filter description are completed or until the first negative result is obtained, and ends when either of these two conditions is satisfied.

  The intelligent router performs all attribute checks on this filter and then determines whether there are more filters (step 316). If there are more filters, processing returns to step 308; The attribute check for the next filter is read and the check for that filter is performed. The matching process (steps 308, 310, 312, 314, 315, 316) continues until the process on the entire set of filters is completed or until all rules are determined, and either of these two conditions is met. When satisfied, it ends. If the packet does not meet the filter criteria, the filter is excluded (deleted) and not forwarded.

  The intelligent router 92 may perform a series of processes using filters in any specific order. For example, as shown in Table 15, the intelligent router 92 can store filters for subscription reservations in a file or routing table, and can sequentially read these and apply filters to attributes (perform attribute checking). . Alternatively, the routing table may include a link or pointer to the filter.

  In content-based routing, two or more techniques may optionally be used simultaneously based on application and heuristics such as switching algorithms based on traffic conditions, for example. For use in examining payload sections for content-based routing, the filters used for processing may optionally be encrypted, decrypted, transformed, and merged at routers in the network. For example, if the issue information in the application does not include currency information that is less than two digits after the decimal point, the fraction of subscription information of> 3.54122 may be rounded to> 3.54 dollars. Alternatively, foreign currency may be converted into US currency when overseas issue information reaches the first router existing in the United States.

  Instead of a linear approach, the intelligent router 92 may select filters for processing in other orders that can improve the speed and efficiency of processing or based on various algorithms. Table 16 shows a specific example of the link corresponding to the subscription reservation. In these examples, the subject is associated with a particular channel, and the subscription for the subject can be represented by a routing rule for the filter. The subject can include a network address for identifying the source of the content, such as a uniform resource locator (URL).

  FIG. 14 is a flowchart of the caching process 320. The process 320 is implemented in a software module executed by the processor 93 in the intelligent router 92, for example, represented as a cache manager 218. Alternatively, this processing may be realized by an ASIC, or may be realized by a combination of hardware and software. These may be the same physical device as the corresponding intelligent router or a different physical device. In process 320, intelligent router 92 receives a message containing data or content, a channel ID, and a subject (step 322). The intelligent router 92 adds a mark indicating the time to the data (step 324) and caches the data in a local memory such as the memory 94 or the auxiliary storage device 97 (step 326). The intelligent router 92 indexes the cached data using the channel ID, subject, and time stamp (step 328).

  When intelligent router 92 receives a request for data (step 330), intelligent router 92 retrieves the cached data using the index based on the request (step 332). Intelligent router 92 forwards the cached data to backbone router 95 or other routing entity, which is ultimately transmitted to the requester or other entity. The process 320 may be repeated so that data is continuously cached and the cached data can be retrieved on demand.

  FIG. 15 shows a specific example of the cache index (336) used in the process 320. The cache index (336) receives the data (338) and stores it along with the time stamp (340). When collecting data, each data is marked for each period of Δt. Here, Δt represents a time between marks, for example, a period between t2 and t1. Any other type of index may be used for the time mark.

  Table 17 conceptually shows the indexes used for the cached data. Table 18 conceptually shows data for storing a connection history for caching. Table 19 shows a specific example of a data structure used for locally cached data in a network node having an intelligent router.

  The mark indicating the time may be given at any fixed interval or at a variable time interval. For example, data may be cached every 5 minutes and indexed. When the channel manager 218 receives a command to read cached data specifying a time and subject (eg, # .getCache), it can read the cached data corresponding to the request in step 332 using a cache index. Determine if you can.

  Each subject or channel can include, for example, its own IP address in the multicast tree and a set of intelligent routers. Table 18 shows the connection history between such routers that can be stored locally on the user equipment. If an edge router fails, the device accesses the connection history to see how the edge router can reconnect to the upstream router of the channel when it comes back online. In addition, the device can execute a command (get cache command) for acquiring cache data during a period in which the device is not connected to obtain pending content related to the subscription.

  These exemplary data structures include the following information: A subject node includes a subject identifier, a subject level, a pointer to a parent channel or subject node, a file descriptor for its own directory, a pointer to a hash table containing subject nodes of the next level, and a data node And a pointer to The data node has a pointer to the subject parent node, a file descriptor for the data directory, a circular buffer containing data structures for the data stored in each storage device, a head and tail of the buffer, And a lock to lock the data node during reading and storage. A saved time grain node is a node that represents an actual data file, and the last grain node represents a buffer that is not yet saved in the storage device but is held in memory. The caching and data storage thread in this embodiment uses the mutex of the last time grain node to prevent the last time grain node from being accessed simultaneously.

Processing at Agent FIG. 16 is a flowchart of agent processing 350 for a transmitted subscription message. The process 350 can be implemented, for example, by a software module that includes an agent 128 that is executed by the processor 134 in the user (subscriber) device 122. In the process 350, the agent 128 receives the subscription reservation generated by the process (step 352) described above with reference to FIGS. 11 and 12, for example. Agent 128 creates a string that identifies the logical expression for the subscription reservation (step 354) and parses the string to detect errors in the subscription reservation (step 356). If an error exists, agent 128 displays an error message to the user (step 360) and prompts the user to correct the error and re-enter the subscription. If the subscription does not contain any errors (step 358), agent 128 saves the representation in the data structure (step 362), as described below. The agent 128 converts the inequality operator included in the data structure into a positive form (step 364), and further converts the data structure into a corresponding disjunctive normal form (hereinafter referred to as DNF) ( Step 366). Agent 128 also simplifies the AND expression of the DNF structure so that only range filters and membership checks are included (step 368).

  DNF is a well-known standard form where a logical expression is represented as an OR (called disjunction) of one or more subexpressions, and each subexpression is represented by an AND of one or more attribute checks. For example, the equivalent DNF expression to the logical expression (price> = 10 AND (symbol = "LU" OR symbol = "T")) is ((price> = 10 AND symbol = "LU") OR (price> = 10 AND SYMBOL = "T")).

  The transformation in step 364 identifies all allowed values, replacing an expression with an “inequality” operator (represented as! = In the exemplary syntax) with one value that is not allowed. To the equivalent “positive” format. Since the router used in this example requires an expression in the positive form, this conversion is done before generating the DNF. For example, the expression (price! = 80) is converted to an equivalent positive expression (price <= 79 OR price> = 81).

  The transformation in step 368 is performed after the DNF is created and includes a process that further simplifies the AND expression generated here, thereby simplifying the processing of the router in this embodiment. Specifically, in particular, the AND of multiple attribute checks for the same attribute is a canonical “includes one lower limit, one upper limit, both lower and upper limits, and, in the case of equivalence checks, any single value. It can be simplified to “range filter”. A particular type of range filter is encoded based on, for example, Table 22.

  For example, the expression (price> = 10 AND price <= 80 AND price> = 20 AND price <= 100) can be reduced to the expression (price> = 20 AND price <= 80), which has both lower and upper limits. It is a specific example of a range filter having Specific examples of other types of values after reduction include (price> = 20) (lower limit only), (price <= 80) (upper limit only), (price = 50) (single value), and the like. In creating these range filters, any subexpression can be reduced to true or false, in which case the subexpression can be eliminated based on the laws of Boolean algebra, so that the encoding of the expression in the message is reduced. Further optimization. For example, the expression (price> = 50 AND price <= 20) can be simplified to false because there is no “price” value that satisfies the expression. In the special case where the entire filter representation is simplified to false, the agent does not need to create a message, thereby freeing the router from unnecessary work.

  If the subject filter contains wildcards, agent 128 can optionally convert them (step 370), as described below. Alternatively, any wildcard may be converted in the network rather than the user device or other device. In this illustrative example, the syntax for the subject filter is the only syntax that uses wildcards, and the syntax for the attribute filter is the only syntax that uses logical expressions. Alternatively, different or variable types of syntax may be used for the subject filter and the attribute filter.

  Agent 128 encodes the generated DNF representation into a message (step 372) and forwards the message to the intelligent router (step 374). This encoding includes a process of converting a subscription reservation into a flat message format, which means a format composed of a string of data. The forwarding process also includes the process of propagating the routing rules generated by the subject filter and attribute filter for the subscription reservation to one or more intelligent routers or other routing entities in the network. For this propagation, for example, a subscription representation can be mapped to a conventional packet structure.

  The encoding in step 372 includes converting the subscription corresponding to the channel to the messaging API messaging format for propagation through the channel. For example, the subscription reservation is subject #. Messaged internally as a notification including SUBSCRIPTION. Since there are both an arbitrary number of subject filter fields and an arbitrary number of attribute checks, in this embodiment, one set of bytes is used to store the number of subject filter fields, and another set of bytes is used to attribute. Save the number of examinations. The individual fields of the subject filter are rearranged in the order specified in the original subscription, for example, and stored in the 2-byte part of the message. The arrangement of the wild card field will be described later.

  In the attribute check array, the check operand is placed at the end of the message, similar to the placement of attribute values in the notification. Prior to placing attribute checks and operands, they are sorted in attribute order within each DNF choice by a positional order check for a given attribute followed by a name order check for any attribute. In addition, the set of relational checks for scalarally evaluated attributes within each choice can be a range filter with one limit (range open or right or left or equivalent test) or two limits (separate The closed range between the limits) is simplified to the standard form. The remaining information about the check is encoded in the same order as the operands, for example, into two byte pairs. This sequence of two byte pairs is placed in the message immediately after the sequence of 2-byte encoded data in the subject filter field. Two byte pairs can constitute one form of attribute check bit string encoded data sequence, which, apart from the two byte pairs, is also used to represent other types of encoded data. be able to. A specific example of attribute inspection will be described later.

  Table 20 shows a method for encoding the attribute check. Table 21 shows the encoding by two byte pairs and Table 22 shows the encoding of the operator ID in two byte pairs.

  Since the two byte pairs for the test already indicate both types of operands for the test, regardless of whether the test is applied to a given attribute or to any attribute, any There is no need to organize multiple tests performed on attributes or their types individually. This method assumes that the number of predetermined attributes in the notification is 127 or less. Alternatively, the attribute check may be encoded with more bits in this design.

  This arrangement convention determines the order of attribute checking and groups them based on the attribute filter's DNF, but infrastructure elements (eg, routers) can make a comprehensive evaluation of attribute filters more efficient. The tests may be selected and evaluated in another order (eg, in an order dynamically derived from local data regarding the probability of success or failure of different tests). The subscription ID field of the message is a value generated by the agent to uniquely identify the subscription to the agent's edge router in subsequent requests and to change or cancel the subscription. Specifically, the dynamic change to the subscription reservation attribute filter is subject to #. The message format shown in the embodiment of FIG. 18 is propagated except that it is RESUBSCRIPTION and the subscription reservation ID is registered in advance and is the ID of the subscription reservation that is currently being changed. The unsubscription is, for example, that the subject is #. UNSUBSCRIPTION is used, and the subscription reservation ID is registered first, and the subscription reservation ID for which subscription stop processing is performed is propagated through the subscription reservation ID field using the message format shown in FIG.

  Hereinafter, a specific example of the conversion and encoding by the agent described above will be described. First, a specific example of the following attribute filter expression will be considered. That is, price> = 10 and (symbol == “LU” or (volume> = 1000 and volume <= 10000)). FIG. 19 shows a diagram 390 representing, in step 362, the object used by the agent to store the representation in unified modeling language (UML). The diagram 390 illustrates a hierarchical relationship for identifying subscriptions that can include variables, constants, or both. Objects in diagram 390 may be instances of filter classes, depending on the particular implementation. Each SimpleFilter object represents an attribute value used to store information about the corresponding attribute check of the filter expression. In the expression shown in FIG. 19, the OR filter 396 combines two AND filters 392 and 400. The AND filter 392 includes a simple filter 394 having an attribute for subscription reservation. Similarly, the OR filter 396 includes a simple filter 398, and the AND filter 400 includes simple filters 402 and 404.

  In this embodiment, attribute price (price), symbol (symbol), and volume (volume) are defined in advance as attributes of related channels, and positions 0, 1, and 2 are defined for these attributes, respectively. It is assumed that Furthermore, the type of each attribute is an unsigned integer (type code 6), a character array (type code 12), and an unsigned integer (type code 6).

  Next, a subscription reservation including the attribute filter expression described above as an attribute filter will be considered. FIG. 18 shows a specific example in which subscription reservations are arranged in a message. The data structure 386 on the left side of FIG. 18 shows the actual message content, and the data structure 388 on the right side of FIG. 18 shows a description of each part of the message. The width of each data structure in this drawing corresponds to 4 bytes. Before doing the array, the filter is converted to an equivalent DNF like (price> = 10 and symbol == "LU") or (price> = 10 and volume> = 1000 and volume <= 10000) .

  The 16-bit attribute check encoded data is shown as a bit sequence having a gap separating different parts. Note that in this example, the two tests for the price are in separate discourses and cannot be combined, so they are in a range that does not have a right boundary (the right side is As an open range). On the other hand, the two exams for the volume can be combined because they are in the same discourse, so they are organized as a single “closed range” exam.

  Also, certain fields are characterized as “assumed”, which means that in this example, the values of these fields are arbitrarily chosen and are usually independent of the arranged subscriptions. Means. In addition, the subject filter for the subscription is arbitrarily selected as “>”, which matches any subject defined by the associated channel. The specific examples described above and shown in FIGS. 18 and 19 are exemplary, and this arrangement may be used with any other type of subscription. Further, the process 350 is merely an example of the arrangement of subscription reservations, and the subscription reservations may be arranged by any other method.

  FIG. 17 is a flowchart of an agent process 376 for a transmitted message. The process 376 can be realized by the agent 128 and the application 126 in the user device 122, for example. In process 376, agent 128 receives a message corresponding to the subscription from the intelligent router (step 378). For example, the 128 agent determines the channel corresponding to the subscription reservation by the channel ID of the message (step 380), and calls the API of the channel (step 382). The API displays the data for subscription reservation in the GUI or other format at the user device (step 384). As processing for the input message, data decoding processing that is the reverse of the encoding processing described above can be used, and this decoding (decoding) can be performed in the router or other network entity.

Wildcard Processing FIG. 20 is a flowchart of wildcard processing 410. The process 410 is an example of a process for converting a wild card into an expression for subscription reservation using a set of routing rules of the filter. Process 410 may be implemented, for example, as a software module represented by agent 128 executed by processor 134 in user device 122. Alternatively, the wild card may be processed by the intelligent router 92 or the processor 93 in the ASIC 91 having a corresponding function under software control in the network. Wildcards contain open fields or variable length fields, and Table 21 shows examples of such fields.

  In process 410, agent 128 or other entity receives a subscription that includes a wildcard (step 412). The subject length of the subscription can be specified when the issuer publishes the content, and the subject may be pre-processed in the issuer device. For example, the subject field is counted to calculate the field count (length). May be. The 128 agent counts the number of fields in the filter operand (step 414) and initializes a new rule (filter) with field length = N (step 416). The agent 128 reads the subscription reservation subfield (step 418) and determines whether the filter operand subfield 0 [i] is a wildcard (step 420). If the filter operand subfield is not a wildcard, the agent 128 adds a conjunctive clause to the rule, setting field [i] = 0 [i] (step 422). If the filter operand has additional subfields (step 424), agent 128 returns to step 418 to process this additional subfield. The parameter “i” represents a field, and i is an integer representing a field number in this embodiment.

  After processing the subfield, agent 128 determines whether the last filter operand subfield is ">" (step 426) and if the last filter operand subfield is ">", the length of The constraint condition is changed to field length> N−1 (step 428). Wildcard processing may use any kind of symbol, and “>” is just one example. In this specific example, “a.>” Causes a. b, a. c, a. d, etc. and all these sub-subjects at all levels (eg abx, acx, abxy, etc.) can be represented. Other symbols may be used as wild cards.

  If necessary, agent 128 propagates the modified rule to the intelligent router or other entity in the network (step 430). The process is repeated for all subfields, which converts wildcards into non-wildcard rules. Non-wildcard rules mean rules that do not contain wildcards. Wildcard conversion may occur anywhere in the network, for example, at a subscriber device or an intelligent router. Thus, this transformation may be performed on one entity by a transformation rule and propagated to other entities, or this transformation may be performed dynamically.

  Table 23 provides an overview of these exemplary routing rules for wildcard processing, along with specific examples. These routing rules may be generated by an intelligent router, for example, or may be generated by other network entities and propagated to the intelligent router. Furthermore, the routing rules in Table 23 are shown for illustrative purposes only, and wildcards may be converted using other routing rules.

Reliable publishing-subscription system in unreliable networks The main purpose of the publishing-subscription network, such as the publishing-subscription network described herein (e.g., shown in FIGS. 1-3), is for subscribers Deliver notifications that match subscriptions efficiently and in a timely manner. In order to implement an issue-subscribe network, it is necessary to transmit a notification from the issuer to the subscriber in the form of a network packet. Unreliable networks include, for example, networks in which some of these network packets can be lost during transmission. The main technical challenge to be overcome in this embodiment is to determine if the notification (ie network packet) was lost due to a transmission error or was excluded by the filtering mechanism in the upstream router. .

FIG. 21 is a flowchart of a process 520 for increasing the reliability of the issue-subscribe network. In this process 520, a highly reliable transmission protocol is used. Specifically, reliability is achieved by using a reliable transmission mechanism (reliable tunnel) between a pair of adjacent routers (or between a router and user equipment). If the router determines that the sent notification needs to be forwarded to an adjacent router, it forwards the message using a reliable transmission protocol, preferably TCP. Each router only distributes notifications to entities (hop distance = 1) adjacent to the downstream side. Thus, reliable delivery of notifications from issuers to subscribers takes place hop-by-hop. The process 520 is executed by the intelligent router 92 or the processor 93 in the ASIC 91 having a corresponding function under software control. Further, the process 520 may be realized by a software module represented by the agent 105 executed by the processor 114 of the user device 100, for example.
It has been found that using TCP allows a reliable transmission protocol to be incorporated into the content-based routing of the publish-subscribe system described herein. Another technique that can be used in a publish-subscribe system is forward error correction. In forward error correction, a plurality of notification packets are transmitted to the receiver, and one of them is expected to be received well. This increases the probability that a packet will be received at the expense of the bandwidth used, but does not guarantee delivery. A third technique is reliable multicast, where a set of multicast receivers is promised to receive the “same” set of packets from the multicast sender. This approach cannot be used in CBR issue-subscribe systems where different receivers need to receive different notifications based on different subscription reservations.

  A node in the publish-subscribe network receives a subscription for content over the network (step 522). The content received from the issuer is published over the network using a CBR based on subscription by receiving a notification about the content at a node in the network (step 524), and the notification is sent to the adjacent node. It is determined whether to send (step 526), and based on this determination, the notification is selectively forwarded to neighboring nodes using a reliable transmission protocol (step 528). These steps are preferably performed according to any of the CBR processes described above.

  The reliable transmission protocol used in step 528 is preferably TCP. TCP is a communication protocol for the transport layer of the OSI model. TCP provides a transfer function that promises that the total amount of bytes sent is received correctly at the other terminal. Accordingly, the transfer step 528 is executed from a single node to a single node, and when a transmission error occurs, the transfer step 528 receives the instruction information indicating the failure in transmission of the notification (step 5281). And processing for retransmitting the notification in accordance with the instruction information (step 5282). The request for information indicating such a transmission failure is preferably encapsulated in a notification packet.

  The notification packet is preferably buffered in a buffer in case it becomes necessary to retransmit. When the buffer for a particular adjacent node is full, the supply of data to that node may be stopped, the packet may be deleted on a first-in first-out basis, or the packet is stored in auxiliary storage May be. Alternatively, if a certain quality of service (QoS) is guaranteed for a given issuer or subscriber, packets may be selectively deleted from the buffer based on QoS (eg, by QoS Delete packets that are not related to constraints first).

  As shown in FIG. 21, if further routing is required for an adjacent node, the process for the adjacent node is repeated (step 530). In this manner, the process 520 can realize delivery of reliable notification from the issuer to the subscriber in a hop-by-hop manner.

Filter propagation Content-based routing (CBR) is a type of constraint-driven routing. In the publish-subscribe network (for example, see FIGS. 1 to 3) whose performance has been improved by the function of the CBR such as the publish-subscribe network described above, it is desirable that the subscribers in the issuance of specific contents (ie, subscription reservation) are desirable. Interest is encapsulated in an entity called a filter. Routers in the network preferably use a filter to determine the path. Unlike conventional filtering methods that can only be performed at the edge or user terminal, this CBR method that can operate at the network core involves the propagation of these filters. If the subscriber is given the ability to impose constraints on the content of the packet received by the subscriber, and if these content constraints are used at the core of the network to route the packet, both efficiency and network functionality are Benefit from CBR.

  This embodiment provides several highly scalable filter propagation methods that allow (1) filtering along the data path between the issuer and subscriber to the core router, from the core router, and / or between core routers. And (2) if necessary, CBR can be performed on the upstream side within the network core, not just at the edge of the network. By performing CBR on the upstream side (eg, within the network core or up to the issuer device), the number of data packets that need to be frequently routed to the downstream side can be significantly reduced. As a result, the traffic load on the network is reduced.

  The filter propagation process described herein includes aspects of propagating any name and any form content filter to, and from, the core of the network. Filter propagation has many aspects and involves many problems. An important issue to be solved by the filter propagation method described here is that the filter is “up” the propagation stream, ie, from a large number of filter sources (eg, subscribers) to a small number of filter destinations (eg, When migrating to a core router, edge router or issuer, it is to avoid building a large number of filters. This problem is also referred to as an “accumulation” problem, and occurs, for example, in a network composed of nodes connected in a tree topology.

  Solutions designed to solve this and other problems include a set of one or more of rules, procedures, policies, processes, mechanisms, protocols, message formats, data structures and / or algorithms. These solutions allow for quick (eg, simplification, merging, efficient organization, etc.), scalable (eg, merging, estimation, etc.), dynamic (eg, new subscriptions, Unsubscribing, remediation, etc.), secure (eg, by authentication), personal (eg, by encryption), and reliable (eg, TCP connection, reliable tunnel, encoding, recovery protocol) Etc.), flexible (eg, passive propagation (discussed below), various multicast topologies or protocols), and adapted to user requirements (eg, by subscriber / issuer behavior encoding). realizable. The propagation process described herein utilizes these solutions to achieve these and these benefits.

A. Network Topology Filter Propagation Strategy The publish-subscribe system can be implemented by multicast, broadcast or mobile publish-subscribe strategies. Filter propagation may be embedded in the network topology or may be performed “above” or in addition to the network topology that is enforced by the strategy selected to link the publisher and subscriber. Alternatively, the filter propagation strategy can be used to link publishers and subscribers, such as publish-subscribe schemes, that include published advertisements that can deliver subscription subscriptions with embedded filters directly to the advertising destination. It may be a strategy itself.

  In the multicast category, filter propagation is a protocol that matches the protocol used for multicast (eg, protocol independent multicast-sparse mode (PIMSM), protocol independent multicast-dense mode). mode: PIMDM), core-based tree (CBT), distance vector multicast routing protocol (DVMRP), multicast open shortest path first (MOSPF), etc.) Can be used. Depending on the protocol used, the filter propagation may be bidirectional or unidirectional, and many filter propagation strategies are possible. The strategies are: 1) When a filter is received, all filters are propagated to all possible upstream links (aggressive propagation scheme), 2) Filter propagation depending on the received issuer packet (Passive propagation scheme) strategy, 3) strategy to propagate according to the broadcast advertisement by the publisher, 4) strategy to propagate after calculation of network topology (eg MOSPF), 5) periodic Strategy to filter propagation (e.g., satisfying a refresh request in a system that supports filter timeout), 6) Strategy requiring filter re-propagation during fault repair in the network, 7) Filter during network restructuring Strategies that require repropagation and / or 8) There are these combinations. The strategy preferably follows the filter propagation methods and rules described herein.

B. Link-level connection mechanism for filter propagation There are various methods for connecting a source and a destination related to filter propagation processing. For example, the sources and destinations involved in filter propagation processing are 1) hop-by-hop (eg, from router to router), 2) end-to-end (eg, from user device to cache for offline data retrieval), 3) using a tunnel, 4) using a transmission control protocol (TCP), 5) using a user datagram protocol (UDP) (eg, instead of a reliable data recovery mechanism), 6) using a security protocol 7) using other protocols, 8) only in the kernel of the router and / or user equipment, 9) at the router and / or user equipment, only for user level processing, 10) at the router and / or user equipment For all possible combinations of both kernel and user level processing To (e.g., to the application from the kernel, the application to through kernel application, etc.), and / or 11) can be connected by any combination of the above.
C. Filter Propagation Destination As shown below, filters may be inserted anywhere in the publish-subscribe network: 1) All routers and user equipment that form an end-to-end connection between the issuer and the subscriber 2) Any router in the network including the edge router, 3) Core router only, 4) Edge router only, 5) Subscriber device, 6) Issuer device, 7) Administrator device, cache, 9) Backup router, 10) It may be inserted into any router that handles packets, 11) new routers involved in network repair or disaster recovery, and / or 12) any combination thereof. The filter is preferably changed in the approximation, merging process and other processes. The form of the propagated filter may be different for each category of the destination described above.
D. Source of filter propagation The source of filter propagation can be any part in the publish-subscribe network.
Specific examples of filter sources include the parts listed in Section C. Similarly, filters are preferably modified during propagation with approximations, merging and other processes, and therefore filters from different sources often have different forms.

E. Rules for Potential Filter Recipient Behavior Also, filter propagation can vary depending on potential recipient behavior and needs. For example, 1) an issuer located in a permanent or semi-permanent location wants the filter to be propagated to that location, and 2) a moving issuer will immediately leave the current location, so that Do not want the filter to be propagated to the location, 3) small publishers that will only publish a small amount of content before logging off, do not need filters, and 4) large publishers are You may want the filter to be propagated before publishing the first set of content to save breadth 5) Publishers that restrict access permissions will have subscriber privacy and confidentiality in their applications 6) The core router that implements the filter time-out strategy may start from the downstream router periodically. Require updating, and / or 7) edge router which is not always connected to the subscriber device may subscribe or interested filters desires to be updated regularly. All these behavioral properties and other properties include, for example, using issuer advertisements, encoding issuer behavior in the first issue, sending filter updates periodically, and receiving filters. Can be discriminated by other techniques that convey the needs of the elderly.

F. Rules for Filter Sender Behavior Similarly, filter propagation can also vary depending on filter sender behavior and needs, for example, as follows. 1) If a subscriber wishes to sample published content, he may want to use only temporary filters with a time limit (defined by the system or defined by the user). 2) Subscriber can move and expire filter after subscriber moves. 3) Subscribers who are not always connected to the edge router may wish to send periodic filter updates that support automatic filter timeout in the CBR system. 4) In some policies, a subscriber can be defined as needing to be authenticated before propagating the filter. 5) CBR system routers that support automatic filter timeout need to send filter updates periodically. 6) A specific query for only a predetermined set of data can include or be implemented by a filter (eg, data read from a cache in response to a query or subscription, persistent backups) Data, data restored during disaster recovery, etc.). 7) A router that has changed the filter for propagation can indicate to the upstream router how the propagated filter has been changed based on the needs of the sender. All these behavioral properties and other properties can be supported, for example, by encoding the behavior of the filter propagation message or by using a special protocol (eg, a protocol for authentication).

G. Propagated Filter Processing Filter processing may include transformation processing that changes the content or representation of the filter and / or management functions that simply organize the filter. The original filter representation in the subscription can be taken over and accepted by different sets of processing procedures in different parts of the data delivery path, including the following path sections: 1) From subscriber device to edge router. 2) From edge router to core router. 3) Between two core routers. 4) From the core router to the issuer edge router. 5) From the edge router to the issuer device. Some specific examples of these conversion processes are as described above (see, for example, agent process and wildcard process). These processes may be performed anywhere in the publish-subscribe network, such as a subscriber device, issuer device, edge router or core router. These processes may be performed by components such as agents (agents on subscriber and issuer devices) and daemons (daemons on routers).

  Specific examples of processing include the following processing.

  1) Conversion of attribute content (for example, conversion from subject name to subject ID, conversion from strings to their lexicographic values, conversion from one currency to another, etc.).

  2) Expression expansion (eg, expansion from wildcards to their components).

  3) Expression conversion (for example, conversion from> operator to> = operator, conversion from! = Operator to> and <operator, etc.).

  4) Display conversion (for example, conversion from string representations to symbolic representations such as numeric or Polish notation or parenthesis notation, or specific data used to store or match filters such as tree, radix trie, etc.) Conversion to a display suitable for the structure).

  5) Encoding / decoding (eg checksum).

  6) Encryption / Plain culture.

  7) Device level sorting and reverse sorting.

  8) Reordering protocol level to and from a given filter propagation message format.

  9) Decomposition (for example, disassembly into disjunctive standard form (DNF)).

  10) Simplification of redundant predicate expressions (eg, $ 1> 5 && $ 2> 3 && $ 1> 3).

  11) Inclusion of two or more filters.

  12) Approximate filter before and after the merge process.

  13) Merge processing of two or more different filters (eg based on approximation).

  14) Grouping of two or more separate filters (eg, for propagation to different uplinks).

  15) Unsubscribe or remove the filter.

  16) The filter expires.

  17) Update filter differences (eg, bookkeeping an old set of filters with some representation and subtracting a new set of filters from it to get a change in the filter configuration, ie, update the difference).

  18) Configuration of filter pattern (for example, a pattern in which when filter 1 matches, filter 2 also matches). Any processing related to filter propagation described above or not described above can be used in the filter propagation method described herein.

H. Filter Propagation Message Content and Format Messages can be used for transmission or propagation of filters. The content and format of the message containing the propagated filter may be affected by the transformation process applied to the filter and the conditions that trigger the filter propagation during propagation. A specific example of the format is shown below. 1) Any message layout at the device or user level. 2) The filter may include attributes at fixed or random locations (eg, strings that match any part of the published information). 3) The filter may have a fixed size (eg, integer, Boolean, etc.) or may have a variable size attribute (eg, array, string, c structure, etc.). 4) The data type of the filter may be simple (eg, integer, Boolean, etc.) or complex (eg, array, string, range variable, etc.). 5) Filters can be simple checks (eg>, ==, <= etc.) or complex checks, or operators (eg membership checks, regular expressions, string matching, intersection queries, enclosure queries, containment queries, etc. 6) The decomposed DNF format. 7) A simplified or original representation of the filter. 8) A compact or expanded format (eg, a one-way filter can be expressed using two limits, or a bidirectional filter can be rewritten as two one-way filters). 9) Partial representation of the filter (eg truncating string predicates, reducing data type resolution, reducing the number of attributes). 10) Conversion of filter expression (Polish notation, parentheses, tree, trie, etc.). 11) Inclusion / exclusion of subject fields (eg, subject filtering may already be implemented by the multicast backbone through which the filter is propagated). 12) Extension of wildcard subject or attribute predicate. 13) Conversion of content or expression. 14) Encoding information (for example, error correction checksum). 15) Encryption. 16) A complete set of filters (e.g., refresh information is updated with filter time limits in the system during failure recovery). 17) A partial set of filters (eg, only new filters are inserted into the included filter set or old filters are excluded). 18) Modified set of filters (merged filter, estimated filter). 19) Changes in differences in filter representations that are not similar to filters (eg, changes in the radix tree during filter insertion and deletion). 20) Filter identification label instead of coordinates (eg, during unsubscribing). 21) Filter expiration parameter. 22) Filter patterning information. 23) Information for encoding the behavior and request of the filter sender as shown in section F. These are merely illustrative of the possible content and format of the propagated filter that can be used in the processes described herein.

I. Procedures and policies for filter propagation At the system level, policies and procedures that govern some or all aspects of filter propagation may be established. For example, the following aspects are included in these aspects. 1) Management (eg, global shutdown and restart restart). 2) Maintenance (eg, failure recovery). 3) Filter persistence (eg, one-time filtering on expired, cached data). 4) Performance improvement policies (eg, averaging, edge selection, positive propagation versus passive propagation, etc.). 5) Monitoring (eg failure reporting). 6) Accounting (eg, statistical bookkeeping). 7) Billing. 8) Regulations (eg authentication for propagating filters, types of filters allowed in the network, etc.). 9) User authentication. 10) User assistance (eg, directory service for subject filtering). 10) Subscriber privacy (eg, hide filters from other subscribers). 11) Synchronization (e.g., approval to the subscriber when receiving the filter at the edge router, approval to the subscriber when filter propagation ends, approval to repair the router during recovery, etc.). 12) Prioritization. 13) Any policy that determines and regulates automatic filter insertion (e.g. based on the interest shown in the set of filters inserted by the consumer himself) Management related to the consumer to insert).

  Many aspects and possible features of filter propagation are as described above in sections A through I. However, to make filter propagation realistic, it is necessary to find a scalable solution that solves the “accumulation” problem described above. Accordingly, a method and apparatus for realizing scalable filter propagation will be described below.

  As noted above, an important issue in filter propagation is that the filter is “up” the propagation stream, ie, from a large number of filter sources (eg, subscribers) to a small number of filter destinations (eg, core routers, edges). When migrating to a router or issuer, it is to avoid building a large number of filters. This problem is also referred to as an “accumulation” problem, and occurs, for example, in a network composed of nodes connected in a tree topology. FIG. 22A shows a filter propagation process 450 that solves such a filter accumulation problem. Process 450 illustrates a specific example of a process for propagating a filter in a publish-subscribe network using the set of solutions shown in Sections A-I above. Process 450 may be implemented, for example, by a software module that includes agent 128 executed by processor 134 at subscriber device 122. Alternatively, the process 450 may be executed by the intelligent router 92 or the processor 93 in the ASIC 91 having a corresponding function under software control. Further, the process 450 may be executed by a combination of these.

  As shown in FIG. 22A, first, a plurality of filters are received (step 452). As described in Section C, the filter may be inserted anywhere in the publish-subscribe network. The received filters are processed to reduce the total number of filters (step 454). This reduction step reduces the number of propagated filters to avoid or mitigate filter accumulation problems. Too many filters accumulate on a single node in this publish-subscribe network, such as when there are a large number of downlink nodes adjacent to the current node, or when the current node is quite late in the filter propagation path. Since the calculation load for processing and collating all the filters is enormous, normal operation of the issuer and the router becomes difficult, and the efficiency of the network decreases.

  Hereinafter, the effect of the filter reduction step 454 will be further described. At each stage of the propagation path, there are many downstream nodes, but there are only one or a few upstream nodes. When each downstream node in a particular stage uses a factor close to or not significantly less than the number of downstream nodes in that stage, reducing the number of filters that are propagated to the upstream node will increase the The total number of filters accumulated in the stream nodes is the same as the number of filters in each downstream node. Based on induction, this reduction factor does not increase the number of filters at each destination of the filter propagation more rapidly than the number of stages in the propagation path, thus solving the accumulation problem.

  FIG. 22B shows preferred substeps in the reduction step. These sub-steps preferably include simplifying the filter and removing redundant predicate expressions (step 4542), grouping filters from only the downlink associated with the subject (step 4544), Estimating filters within a group of filters (before or after merging) (step 4546), clustering the filters based on the relative distance of the filters (step 4548), and including the filters (Step 4550) and merging two or more filters (step 4552). These steps may be performed in an order different from the order described here, and some of these steps may be omitted. The reduction process is as described above in relation to the agent process with reference to FIG.

  The filter grouping step 4544 is important to send the correct filter information upstream. In the following, some examples of this importance will be described. For example, a router can have many downlinks (eg, links with downstream routers). Not all downlinks have subscribers who send subscriptions or filters to them. As another example, in a bidirectional multicast scheme, each link to the router can function as both an uplink and a downlink because the issuer and subscriber can join the network from any arbitrary direction. When a filter request is received from a link, this link is considered the current uplink and only filters belonging to the same subject inserted from these downlinks will be later estimated, clustered, included, regardless of this uplink. And need to be grouped for merge operations.

  Grouping 4544 may also be required for other purposes such as billing and network traffic control. This process may not be performed entirely in some situations, for example when a comprehensive map of filters from all links is needed for some reason.

  Estimate step 4546 modifies the filter so that it captures more subscriptions. For example, the approximation includes processing such as reducing the number of predicates in the propagated filter representation, reducing the resolution of the filter coordinates, or reducing a long string to the first 12 characters. Furthermore, the approximation also has the effect of aligning the format of the filter, which makes the filter more likely to be a candidate for inclusion and merging (steps 4550, 4552). It is also possible to reduce the size of the message propagating through the filter by the estimation step.

  The inclusion step 4550 is a process of eliminating filters that are a subset of other filters. For example, the first filter is a filter for all condominiums whose land is 1000 square feet or more and 1300 square feet or less and whose price is 200,000 dollars or more and 300,000 dollars or less, and the second filter is 1100 square feet. The second filter is a subset of the first filter if it is a filter for all condominiums that are greater than or equal to square feet and less than or equal to 1200 square feet and whose price is greater than or equal to $ 225,000 and less than or equal to $ 375,000 Therefore, it is eliminated by inclusion.

  Steps 4542, 4544, 4546, 4548, 4550, 4552 may be performed in any order, depending on the particular filter reduction scheme and / or algorithm selected. For example, the positions of step 4546 and step 4550 are interchangeable. An estimate may be made after inclusion. Further, step 4550 and step 4552 may be integrated and executed as a single step. For example, the merge process can include an inclusion process as shown in the following graph.

  Still referring to FIG. 22B, merge step 4552 is similar to inclusion in that it combines two or more filters. However, in the merge process, an approximation can be used to combine two or more partially or completely different filters.

  Various processes are shown geometrically to illustrate this process. For example, graphs 1 to 4 are shown below.

  Graph 1 shows two filters sl and s2 geometrically, and filter s2 is a subset of filter s1, as described above, and can be eliminated by inclusion processing. In graph 2, neither filter s3 nor filter s4 is a subset of each other. However, the filter s3 and the filter s4 partially overlap each other. In a merge processing step 4552, s3 and s4 are merged by an approximate process with a minimum filter sA including both s3 and s4. Here, since s3 and s4 are both subsets of sA, s3 and s4 are included in sA. Graph 3 shows another specific example of the merge process. In graph 3, filters s5 and s6 are completely separate filters. In the merge processing step 4552, the filter sA including these filters s5 and s6 and the region between them is estimated. Thereby, s5 and s6 of the filter are included in sA. Graph 4 shows still another specific example of the merging process. Here, the filters s7 and s8 are partially different only for one attribute. Since s7 and s8 overlap with each other and have the same value, in the merge processing step 4552, the two filters are merged into sA to include the overlapping portion.

  Thus, inclusion (step 4550) and merging (4552) are performed when the filters are close to each other. In order to improve the efficiency of the linear search over the entire list of filters, an optional step may be to use a clustering algorithm that integrates filters with boundaries that are close to or overlapping each other (step 4548). The clustering algorithm can be executed incrementally in real time as filters are added or removed. In addition, in the attribute space of the filter attribute, filters may be grouped based on grid coordinates using a fixed-size or variable-size grid having an arbitrary accuracy.

  The portion of sA that is not included in either s3 or s4 or s5 and s6 allows the issue information that is not required by s3 or s4 or s5 or s6 to pass through, so the leakage region (or mitigation region, redundancy Also called area. Therefore, the merge processing step 4552 generates such a leak as a trade-off for filter reduction. Here, if a large rough estimate is made, the leakage area becomes too large, and the benefit of filter reduction is offset. Thus, the publish-subscribe network in which the process 450 is performed defines a threshold for the leaked area and determines whether the leaked area for each approximation used in the merge process step 4552 is too large. Merge processing step 4552 performs a statistical analysis to determine the percentage of extra packets (ie, packets included in the leaked area) that are received for approximation with respect to the leaked area. This statistical analysis is constructed based on the packet distribution profile and using the history of received or sent packets (eg, at the router performing process 450). If the expected rate exceeds the leakage area threshold, no approximation is performed.

  Usually, the threshold value of the leak region is a numerical value indicating a ratio in many cases, and two or more threshold values may be used as an evaluation criterion for the merge process. As a specific example of the threshold value related to the leakage region, for example, in graph 3, when the threshold value # 1 of the leakage region is designated as 10% of the number of packets accepted by the actual filters s5 and s6, the filter s5 related to the merge processing step 4552 , S6 is estimated to include 20% of the total packets, the merging operation can be avoided. As another specific example of the leakage threshold, in the graph, the leakage area threshold # 2 is 2% of the total number of packets passing through the router, and the gap between the filters s5 and s6 is the total number of packets passing through the router. If it is estimated that it may constitute 1%, the percentage of packets belonging to the leaky area that consumes downstream bandwidth when routing without filtering is small, so the first leak Even if the threshold (# 1) of the region is exceeded, the merge operation can be performed. Thresholds can be designed and specified for a large number of differently intersecting leak regions.

  The leak region threshold can be determined based on the number of filters being merged, the region of filter coverage, or any characteristic of the filter distribution. For example, if the total number of filters is small, the accumulation problem is not a tight problem, so the constraints on merging filters may be relaxed.

  As described above using graph 4, not all merge operations are approximate. For example, a merged filter sA composed of two or more separate filters has no leakage area in some situations and is therefore an accurate filter. For example, for a two-dimensional filter (ie, s7, s8 in graph 4), if two rectangles representing the filter are aligned with each other without a gap between them and are adjacent, the merging step 4552 resulting in filter sA is an exact It is processing. In general, in multi-dimensions, in two or more filters, this situation occurs when the asserted range for one attribute and the same asserted range for another attribute have overlapping or intersecting portions. It can happen. Similarly, if an asserted range for a filter (eg, for an integer-only attribute) starts with the next possible value after the asserted range for another filter ends, By assuming the same asserted range for any attribute, these filters can be approximated by a merged filter sA with no leakage region.

  Returning to FIG. 22A and continuing the description, process 450 may include additional filtering, such as the filtering described in Section G above (in addition to step 454). As shown here, process 450 determines whether to propagate the filter based on the rules of behavior of the recipient (step 456). A specific example is as described in section E above. In decision step 456, it is determined which filters (and / or which filters) are to be propagated to potential recipients. For example, a potential recipient may be a mobile issuer that has sent a request not to propagate the filter to the node (eg, edge router) to which it is connected. Instead, a potential recipient is a new issuer who has sent a request to propagate all filters to the node to which he is connected before starting to publish a huge amount of content. Also good. Other potential recipient behavior rules shown in Section E may be used. The order of step 458 and step 456 can be interchanged. Further, for example, if the number of subscriptions (ie, filters) is limited or if the network contains a small number of propagation stages, the accumulation of filters is not a serious problem, so step 454 (more details , Some steps such as merging sub-step 4552) and step 456 may not be performed.

  As exemplarily shown in section F, a determination may be made regarding whether to propagate the filter (step 458) based on the rules of the sender's behavior. For example, the transmission side of the filter may provide an expiration date for the filter. In this case, a decision step 458 tests whether the filter has expired. Also, authentication may be required for the filter sender to propagate the filter to the next node. In this case, the decision step 458 can reject the filter if the source of the filter is not authenticated. Rules relating to the behavior of the filter transmission side, such as other rules shown in section F, may be used.

  As described in Section D, this filter propagation procedure 450 can be initiated from anywhere in a publish-subscribe network that includes a network administrator. The next node is propagated with all filters or the reduced set of filters (step 460). Also, as described in Section C, the next node (filter recipient) may be any part of the publish-subscribe network. The number of filters can be reduced by each of steps 454-458. In practice, the number of filters may be reduced to zero by step 456 or step 458 (eg, based on potential recipient rules indicating that no filter propagates). Process 450 may also perform filter propagation 460 based on one of the filter propagation strategies in the publish-subscribe network, as described in Section A. Thus, the propagation step can include determining a filter propagation strategy that can be included and propagating a filter that satisfies the requirements of the determined filter propagation strategy. For example, the propagation step 460 may always be performed in response to receiving the filter, or only when there is a broadcast advertisement by the issuer. Similarly, process 450 may be performed according to any of the policies and procedures shown in Section I. Process 450 can also format the filter propagation message based on section H. As described in section B, various methods can be used for connecting the source and the destination regarding the propagation step 460.

  After propagation step 460, process 450 may determine whether there is another potential recipient (another node) (step 462). If there is another potential recipient, process 450 determines whether the procedure needs to be repeated for this potential recipient. In some publish-subscribe network strategies (eg, CBT), the filter has no deadline and once it is propagated to one node, any changes to this filter (eg, adding a filter, removing a filter, etc.) This filter may remain persistent on the link until is set. In such cases, filter propagation to that link is unnecessary. Such an example can also be found in the passive propagation method described below.

  If a filter needs to be propagated, process 450 can determine whether to propagate the filter based on behavioral rules (steps 456, 458). Process 450 may be repeated until there are no potential additional recipients of the reduced filter.

  FIG. 22C shows another filter propagation method 470. This process is another embodiment using a set of solutions for propagating filters in the publish-subscribe network described in Sections A-I. The process 470 may be executed by the intelligent router 92 or the processor 93 in the ASIC 91 having a corresponding function under software control. Alternatively, the process 470 can be implemented by a software module including the agent 128 or the agent 105 executed by the processor 134 or the processor 114 in the subscriber device 122 or the issuer device 100, for example.

  The network topology in this example is a bi-directional topology in which a series of issuers P are connected via any arbitrary path from any arbitrary direction to the network of intelligent router C, as shown in FIG. 22D. is there. Such a network is realized by, for example, a core-based tree multicast scheme. In this embodiment, the new issuer PN and the new router CN are connected to the network via the router C. When a new issuer PN connects to a particular issue-subscribe network, it may need to connect to the core through a series of new routers that form an extension of that network and / or already In some cases, it is possible to connect to a network via a router that is a part. For illustration, only one new router CN is shown in FIG. 22D.

  The filter propagation process is performed in two stages. In the bidirectional network shown in FIG. 22D, any link can be uplink and downlink at the same time. Here, for purposes of illustration, the uplink (only one) is defined as the link that is currently receiving the filter request, and the downlink (many) is different from the uplink and is immediately downstream of the recipient of the filter request. Define a link to Downstream and upstream directions are defined similar to uplink and downlink.

  Referring to FIG. 22C, in the first stage, the new issuer PN broadcasts or multicasts a notification requesting a filter to the downstream router C (or all downstream nodes). (Step 472). After Router C receives the filter request (step 474), Router C preferably updates the bookkeeping with a reference counter to ensure that the same set of filters is not sent again on the same uplink ( Step 476). Router C then determines whether the filter request from this uplink has been previously processed, preferably by examining a reference counter for this uplink (step 478). . If this is a new uplink that has never been processed before, for example, this is the first issuer joining the network from this uplink, router C relays the request (step 480) and router Wait for a response from the other downlink (ie downstream router) (step 482). Subsequently, these downstream routers send requests to further downstream routers and wait for responses from these routers. This process is repeated until a response is determined by the receiving node. This completes the first stage.

  The second stage is when either node that receives the filter request gets a response from all of its downstream routers, or that node is the last node in the network and It begins when no response is expected (step 484). If there are no filters to send, a negative response (or an arbitrarily expressed empty set of filters) is sent upstream. When all new filters have been received from downstream routers of all nodes, after the optional filter reduction (step 454), all or part of the filter set (eg, change differences) It is sent upstream (step 486). For the current uplink, only the filters collected from the downlink are grouped, merged and reduced according to process 450. This response collection-filter propagation process is repeated on the upstream side until the issuer or the first router that sent the filter request receives a filter or negative response from all routers downstream of that issuer or initial router. . The actual propagation of the filter described above or the propagation of the empty set (negative response) constitutes the second stage. These two stages differ in the content being propagated and in their propagation direction, since filter requests may be sent farther than other filter requests to the downstream side of some branches of the network. These two stages may be performed overlapping in time.

  This filter propagation scheme starts the filter propagation process only when it receives a filter request from the uplink, i.e., when there is a filter request, rather than actively sending the filter to all uplinks. This is referred to as a passive propagation scheme because the filter is supplied only to the upstream router that requested it. Note that the filter request need not be a specially formatted notification. The initial notification from any publisher can be considered as an advertisement of its presence, and the filter is requested based on section E, using the rules of the filter receiver behavior encoded and embedded in the data notification. It can be determined whether or not.

  The filter of each issuer P may be stored in the issuer user device 100 or the router C to which the issuer P is connected. Similar to process 450 described above, process 470 can use additional solutions, such as those described in Sections A-I.

  CBR with compact filter storage and off-line pre-computation The CBR embodiment described in this section includes the process of routing data packets based on deep packet inspection and filter matching at the core of the publish / subscribe network. In this CBR process, the filter is converted into a more compact form, and the filter application range is pre-calculated in the attribute space (pre-computing: hereinafter also referred to as precompute / precomputation). Create a map for routing instructions from a finite set offline and use this map in real time to make routing decisions. This CBR process is very fast and achieves a data transmission rate comparable to that of conventional Internet protocol (IP) routing. This CBR process is therefore a good candidate for actual commercial content-based routing at the core of the network.

  As described above, a subscriber selects a packet sent from one or more issuers in an issue-subscribe network by specifying a predicate for a range of attribute values. Each predicate is an attribute test that associates one or two operands (operands may be a complex object such as a regular expression) with an attribute at a position in the packet in a relationship check or membership check (also The attribute check may include a subject check.) For example, 10 <$ 1 <20 is a test that checks whether the value of the first attribute position indicated by $ 1 is greater than 10 and less than 20 (the definition of the predicate is 1 It can be simplified by limiting it to relate only one operand, for example, a test with a two-way range can be expressed as the conjunction of two one-way tests). The subscriber presents a collection of these tests, and in one or more subscription reservations, each subscription reservation may include a subject field that represents the subscriber's interest.

  The embodiments described herein relate to content-based routing using filters, which are objects derived from subscription subscriptions, the representation of which differs only from subscription subscriptions. One of the filters is a conjunction filter described later. A subscription can also be said to be the original filter that was specified, used to filter the packets, and / or the subscriber sent directly to the network. To facilitate propagation and packet matching at routers in the network, subscriptions are converted into other formats of objects called filters. For example, in some embodiments, each subject is implemented using a multicast network. A filter that has the same subject and is inserted into each node of the multicast network for a particular subject filters packets that have only the same subject field, effectively performing subject field filtering. Thus, filters can be stored under their subject heading without providing their subject field at each node in the network. A filter without a subject field is also called an attribute filter. In the following description, subscriptions, filters and attribute filters are used interchangeably, and their meanings are only apparent by context.

  To facilitate matching of received data packets with these subscriptions, as described above, each subscription is sent to the equivalent disjunction standard form (DNF) by the proxy before being supplied to the edge router. Processed and split. DNF is the logical sum of one or more sub-filters called conjunction filters. Each conjunction filter is the logical product of one or more attribute checks where each attribute appears at most once in an expression with the same check operator. For example, the conjunction filter can restrict matching of attributes in the following expressions (in the example shown below, attributes other than $ 1 are also included).

$ 1> 10and $ 1 <20and $ 2 <4and $ 3 == 5and ...
As far as routers are concerned, all attribute filters are conjunctive filters for proxy operations. Conjunctive filters have the advantage of facilitating the propagation of these filters at the core of the network, as well as the advantage of having a uniform format, and each occurrence of each attribute in the filter representation. By limiting to, these filters can be interpreted as multidimensional geometric objects, which is advantageous in the CBR described here. Therefore, in the following embodiment, only the conjunction filter is used.

  When complicated processing such as collation for each filter and for each predicate is performed on a huge set including a large number of filters, the required amount of calculation becomes enormous, and it is obtained by using CBR regardless of the use of DNF. The benefit of the downstream bandwidth gain that is made is easily offset. This enormous amount of operations in critical data transfer paths can be a serious obstacle to publish-subscribe network efficiency. In the CBR embodiment described herein, this enormous amount of computation is reduced and most of the computation processing is excluded from the critical data transfer path.

  Specifically, in the embodiment described herein, very efficient processing is achieved in CBR by approximating the filter representation to a more compact format and pre-calculating filter coverage information offline. . In this CBR process, the filter coverage above every point in the attribute space is predetermined offline when a filter is inserted or deleted. The correct processing for a packet whose attribute value is contained at that point is the integration of the processing indicated by all the filters covering that point. These processes include processing on packets, for example, packet forwarding to physical or encapsulated interfaces, packet conversion to user level processing, parallel devices (eg, cache, backup router, etc.) Includes packet forwarding. Some processes do not include packet-related processes, such as a process that triggers a general purpose application running on the router in response to the arrival of a specially designed packet, active issue information Are still present in the area associated with the filter, so there are processes to extend the life of some filters, automatically insert new filters based on published content, etc. In practice, the processing indicated by each filter is, for example, a processing bit mask (hereinafter referred to as ABM) in which each bit position represents an individual (transfer) processing, or the number of processing is a bit mask. If there are too many (for example, when the edge router is connected to a huge number of subscribers), an individual processing list (Distinct Action List: hereinafter referred to as DAL). ) Or the like. In the following description, even when ABM or DAL is explicitly specified, the context includes all forms of processing for realizing the data structure.

  Thus, precomputation generates a CBR map from a predetermined number of coordinates in the attribute space to the same number of integrated processes or lists of processes. Since this simple map is the same as the multicast routing table used in subject group multicast, this is understood as a CBR table. In this way, data is moved very quickly through the critical data transfer path by moving the heavy computational load from the critical data transfer path to the non-critical filter pre-computation process that runs in the background. Transmitted and is hardly affected by the processing required to match the original filter representation.

  The CBR processing also preferably includes processing that approximates the filter representation to a more compact form, for example, by spatial quantization and predicate reduction. Since there are a huge number of coordinates in each attribute dimension, it is necessary to quantize the attribute space and generate a grid. Only the quantized regions instead of the individual coordinates are used to build the CBR table. Spatial quantization has the effect of grouping individual points in the actual attribute space into a single point in the quantized attribute space so that every point in the original actual attribute space is CBR. Shares the list of treatments shown for these quantized coordinates with adjacent points that are considered indirectly in the table and quantized to the same coordinates.

  Various spatial quantization schemes can be used in this process. Furthermore, if the filter expression contains many predicates, this expression may be approximated only to the subset represented in the CBR table.

  The attribute space for attributes is potentially infinite. Therefore, the attribute space may be too large to handle even if a grid is provided. Therefore, in the CBR processing described here, the region of the attribute space to be quantized is preferably reduced. Specifically, it is preferable to quantize only those regions of the attribute space where matching data packets are expected to be received. As will be discussed below, this area where matching data may be received is obtained from statistical analysis of information received from the issuer and / or past (ie, previously received) packets and / or filter distributions. It is determined based on the obtained information. For example, if the issuer provides information that the attribute never exceeds data packets above a certain value, it is desirable not to quantize the attribute space beyond that certain value. Similarly, if statistical analysis indicates that no attribute values below a certain value will be received, it is desirable not to quantize attribute spaces below that value.

  FIG. 23A illustrates a CBR process 480 based on the embodiment described herein that approximates the filter representation to a more compact format and pre-calculates filter coverage information offline. This processing of the CBR minimizes collation for each predicate in real time. The process 480 may be executed by the intelligent router 92 or the processor 93 in the ASIC 91 having a corresponding function under software control. In many cases, pre-computation steps including filter management functions are computationally expensive and it is desirable to do this in software. However, for example, in some filter expressions such as a radix expression, the filter can be realized by hardware. The process 480 may be realized by a software module expressed as the agent 106 executed by the processor 114 of the user device 100, for example.

  As shown in FIG. 23A, in process 480, it is first determined whether or not spatial (re) quantization should be performed (step 482). Spatial quantization may be triggered by periodic or irregular deadlines, or if, for example, it is observed that a packet in the current quantized attribute space is overly uneven May be. Spatial quantization is performed off-line when necessary, when the processing of the router is relatively free (step 484). Spatial quantization preferably involves the calculation of a new grid that has the property of rendering either a uniform packet distribution or a filter distribution (or both) over the quantized attribute space that is formed. Other processing and evaluation criteria may be used for spatial quantization.

  The grid area can be constructed to cover only the reduced area of the attribute space where matching packets are expected to be received. For example, if the first attribute $ 1 represents the number of shares in an industrial sector in the United States, spatial quantization need only be performed in the range from 0 to the total number of shares on the US Stock Exchange. After obtaining information about the statistical distribution of this attribute from the collected packet statistics data, using adaptive quantization as described below, the initial maximum number of shares for all sectors is very high. This range can be adjusted dynamically to smaller numbers and in the future to larger numbers as the market grows.

  The router determines whether any filters have been inserted or removed from the network location (eg, core or edge intelligent router 92, or issuer device 100) where the process 480 is being performed (steps). 486). The reference count process monitors whether or not the newly inserted filter has the same information (for example, attribute, matching attribute value and filter process) as the already inserted filter (step 488, step 494). Only different filters need to be processed. For the sake of explanation, it is assumed that the position in the network is the intelligent router 92. For example, a new filter may be propagated to the intelligent router 92 on the upstream side, or the expiration date set for the old filter may expire.

  If a request to insert or delete a filter is received (step 486), (1) update the reference count for the correct filter in the actual attribute space (step 488), and (2) insert / delete To determine whether the power filter is a new filter or a filter that has already been received (step 490), first, using the actual coordinates, the filter and the router have already received, Compare with other saved filters. If the filter is not a new filter, the execution flow is returned to the packet matching routine. In other cases, to obtain the quantized coordinates of the attribute value, the attribute value of the packet is mapped to the quantized space (step 492). Then, using the quantized coordinates, a similar filter identification procedure is performed to (1) update the reference count for the correct filter in the quantized attribute space (step 494), (2) this It is determined whether the filter is a new filter at the quantized coordinates of the filter (step 496). For step 488 and step 494, two maps from filter coordinates to filter reference values are preferably used.

  If the filter is new in both the actual attribute space and the quantized space, the filter can be approximated and reduced (step 498). For example, only a subset of filter attributes that have too many predicates may be selected for quantization and mapping. Similarly, attributes with more complex nature predicates (eg, predicates that require string matching, regular expressions, non-strings, in-arrays, range operators, etc.) or higher Attributes that require resolution may be handled individually.

  One of the main advantages of the embodiment of the present invention is that if the filter can be made compact, packet matching and transfer can be efficiently performed in a scalable manner. Filter compaction includes both the process of mapping the filter to the quantized attribute space (step 492) and the process of estimating and reducing the filter representation (eg, step 498). As shown in FIG. 23A, the compression step includes a process of spatially quantizing the attribute space (step 484). In this step, a grid is generated in the attribute space as described above and below.

  As a result, the CBR table can be updated (step 500). Each filter is associated with a rule that indicates one or more (routing) processes for the issuance information that matches the filter. The (routing) process indicated by the filter is inserted into the grid cell in the CBR table by the following method. In step 500, the grid generated in the attribute space of the filter coverage information is used to first identify all grid cells (ie, quantized regions) that cover or overlap the quantized coordinates of the filter. . Next, step 500 inserts the (routing) process indicated by this filter into these grid cells. For example, if there are five possible destination nodes, the two filters that need to be routed to the user equipment connected to node 2 and node 3, respectively, cover the grid cell or quantized region X. (Refer to the above description). As a result, step 500 inserts or declares rules that require routing to node 2 and node 3, respectively, for the grid cell or quantized region X. This process is repeated for each grid cell that covers or overlaps the quantized coordinates of the filter. Preferably, step 500 represents a mapping by generating or updating an ABM for each grid cell or quantized region. In this embodiment, the ABM for the grid cell or the quantized region X is expressed as follows, with node numbers 0-4.

  In order to remove the filter, bookkeeping is preferably performed. For example, a process in which two filters cover a grid cell and one of the filters is excluded will be described. The corresponding bit in the ABM should not be reset to zero. For each grid cell, confusion can be avoided by bookkeeping the reference count of the filter covering the cell area. Subsequently, an FBM is generated from the updated reference counter. These reference counters are also called process reference counters. Thus, the mapping step 500 may include a bookkeeping step.

  The CBR table is constructed / updated based on the ABM generated from each grid cell or quantization region (step 500). This process preferably includes, directly or indirectly, spatial requantization and CBR table update and all additional steps 482-500. Steps 482-500 are preferably performed offline (ie, when intelligent router 92 is not receiving issued packets) or as the background of real-time routing processing, filters are inserted, suspended, Stopped or deleted. Thus, if no issued packet has been received, or if steps 482-500 are being performed in the background, process 480 again first determines whether the attribute space should be requantized. Then, it is determined whether or not any filter has been inserted or deleted (step 486). Then, as shown in FIG. 23B, real-time routing processing is executed in the data transfer path using the CBR table.

  The size of the CBR table increases when the number of attributes is large or when the number of grid divisions related to each attribute is large. In principle, in the CBR table, only the grid cells that contain the processing rules (ie covered by the filter) need be stored. The CBR table may be compressed using any technique.

  FIG. 23A shows an aspect of CBR management of the process 480, and FIG. 23B shows an aspect of data transfer by the CBR of the process 480. If it is determined in process 480 that the issued packet has been received (step 502), the intelligent router 92 receives the issued packet (step 504) as shown in FIG. 23B. Received packets can be inserted into the network via a point-to-point connection (eg, a wireless connection to an edge router). The received packet is mapped to the grid (step 506). In step 506, the quantized coordinates of the packet are determined, thereby referring to the correct grid cell in the CBR table, extracting the processing rules contained in the grid cell, and processing the packet (step 508). . These processes may include determining which node or other location in the network will route the packet or delete the packet.

Leakage Reduction As a result of the filter approximation, the CBR processing 480 described so far may cause a packet leakage error (see step 492). Since the packet is matched using the grid cell boundary instead of the original filter boundary, the actual region used for matching (eg, the area in two dimensions and the volume in three dimensions) is expanded and actually called a relaxation region. A packet having an attribute value existing in a region between the filter boundary and the grid cell boundary may pass through the filter by mistake. As described above, the reduction in filter resolution is considered to be an acceptable sacrifice when the high-speed performance and scalability of the process is realized or improved and the leakage error is small compared to the total number of received packets.

  However, there are cases where the leakage error is not small. For example, the system has at least one attribute predicate corresponding to a CBR that is an equivalence check or inequality check (eg, $ 1 = 4 & $ 2 = 100 & $ 3 <30, or $ 1 = 4 & $ 2> 3). May have a filter. In these equivalence-check or inequality-checked attribute dimensions, the filter resembles a point. If the grid boundary is imposed to encompass these points (step 490 above), the relaxation region (eg, area, volume, supervolume) will be expanded much larger than the region of the actual filter itself. There is a case.

  This problem is illustrated by the following figure. This figure represents a two-dimensional attribute space and shows two point filters F1 and F2 respectively defined by an equivalence check in each dimension. Numbers represent grid cell coordinates, and alphabetic characters represent actual coordinates. Packets whose attributes are contained anywhere within the grid cell (4, 1) are accepted as packets that match both FI and F2.

  Various processes that can be performed in the leakage reduction step 516 can suppress such undesirable properties of the CBR process 480, and a specific example of such a process is shown below.

  As one of the techniques for the leakage reduction step 516, there is a technique of performing accurate matching in one dimension (1D). For all attributes, the process of matching packets against a filter at the same time is a multidimensional problem that does not have a scalable solution. On the other hand, accurate matching in one dimension is known to have many efficient and scalable solutions. These solutions include, but are not limited to, for example: (1) a binary search for an array of filter boundaries, (2) a binary tree search through all filter boundaries, and (3) a filter boundary. There are a radix search below a tree in which information is embedded, and (4) a search having a more complicated data structure such as an interval binary search tree.

  In the leakage reduction step 516, it is desirable to perform one-dimensional accurate matching using some one-dimensional search method for all attributes. Some specific examples of the leakage reduction step 516 by one-dimensional accurate matching are shown below.

  a. As described above, the CBR verification step 508 without leakage reduction is performed (if there is no match, the packet is deleted).

  b. For matching packets (see step 508 above), select one attribute dimension and use some one-dimensional exact matching method and the corresponding data structure to change the attribute value of the packet to its attribute axis Match the projections of all the filter boundaries above. As a result, a list of all forwarding nodes where a matched filter is detected is obtained.

  c. If there is no one-dimensional match, all the attribute checks guarantee that there is no filter that matches the packet, so the packet can be excluded.

  d. If there is a match, another dimension is selected and step b is repeated until one of the following conditions is met. (1) When a mismatch is detected in any dimension. (2) When the list of forwarding nodes generated by all previously executed one-dimensional searches does not include a common interface (that is, when a one-dimensional match is generated from filters from different interfaces). Or (3) If correct match checking is performed for all attribute dimensions, including attribute dimensions disabled for CBR.

e. If none of the exclusion conditions described in step d is met and no mismatch is detected in the inspection of all attributes, a decision can be made not to exclude the packet.
f. If more than one interface is included, it is necessary to compare the list of interfaces from which packets are not excluded with the list of interfaces obtained from step a. Packets are routed only to interfaces that overlap in these two lists.

  The order of the sub-steps a to f may be changed. For example, sub-step a may be executed after executing the one-dimensional leakage reduction step (sub-steps b to e). Several substeps may be combined into one substep.

  Note that the one-dimensional leakage reduction step 516 described above is not accurate when all dimensions are considered. Using the last example, a packet with attribute values $ 1 = a and $ 2 = d passes an exact match check in any dimension, but the check in each dimension is actually a different filter. Is matched. With correct multidimensional analysis, this packet should be excluded. Therefore, this leak removal process is effective for packet exclusion, but is not accurate for packet acceptance. For this reason, this process is approximate, and a leakage error may remain. However, although this leakage reduction step 516 is inaccurate, leakage errors caused by using the entire area of the grid cell (4, 1) for filtering can be greatly reduced.

  It is determined whether the data transfer path traffic through the intelligent router 92 on which the process 480 is performed exceeds a predetermined threshold (step 510) (eg, the intelligent router 92 is critical for forwarding packets). If it is occupied by a busy task, it will be “busy”). If intelligent router 92 is busy, the packet mapped to the covered grid cell is routed (step 508) and the corresponding packet statistics are updated based on the routing process determined by step 508 (step 512). (Step 514). Thus, when the intelligent router 92 is busy, a “leak” caused by routing of packets that are mapped to a partially covered grid cell but outside the actual filter coverage of that grid cell is Accepted as a sacrifice to increase speed.

  On the other hand, if it is determined in step 510 that the intelligent router 92 is not busy (that is, if the traffic does not exceed the threshold), preferably leakage reduction processing is performed (step 503). Leak reduction step 516 reduces the leakage described in the previous paragraph. The leakage reduction process can be realized by a method described below, for example.

  Here, inaccurate filter matching for each predicate is executed in real time. However, in process 480, the lowest level of accurate filter matching is allowed only if the traffic does not exceed the threshold. This CBR process essentially groups the filters based on their position relative to the grid. Grid cells are tools that group filters based on their geometric location. Within any group of filters (also referred to as a filter bucket), in some situations, more precise and detailed filtering may be possible. Such additional verification requests can be stored in a filter bucket.

  Some specific examples are shown below.

  1. An attribute check that requires a higher resolution than the CBR map can be performed after the filter bucket is derived from the first grid matching step of the CBR (step 508).

2. Attributes with more complex nature of predicates (eg string matching, regular expressions, non-strings, non-arrays, range operators, etc.) filter bucketing method
Can be processed by.

  3. For filters that include many attribute checks, the attributes may be divided into two groups, one group included in the CBR map and the other group included in the filter bucket.

  The filter predicates for the attributes described in 1-3 can be reduced by step 498 described above. Furthermore, if the intelligent router 92 is not busy (see step 520), the filter determined by the partially covered grid cell based on the CBR table using per-predicate or other exact filter matching. Match such packets to the bucket. Since the filter bucket for that grid cell or quantized region has already been reduced (from the total number of filters) by precomputation (in steps 482-500), this process performs a simple per-predicate CBR. It's just a more efficient process.

  Prior to routing the packet based on the CBR table (determined in step 508) (step 512), the process was mapped to the partially covered grid cell for the filter bucket determined by mapping to the grid cell. It is determined whether there is a reduced filter predicate or packet (step 522). If there are such reduced filter predicates (eg, Examples 1-3 above) or if there are such mapped packets, a per-predicate or other exact filter match is performed. (Step 524). The matched packet is then routed based on the routing process determined by checking the CBR table with an exact match, for example, by matching the interface list obtained from the two processes (step 512). The statistical information is updated (step 514).

  In a variation of the process 480, the traffic threshold check step 510 is performed prior to only some steps with a large computational load, such as step 518 and step 524, and the balance between processing waste and computational load is balanced. Be considered. Further, process 480 may have separate paths for packets mapped to fully covered grid cells and packets mapped to partially covered grid cells. In this case, the traffic threshold step 510 may be performed only on the path of the partially covered grid cell.

  The leak reduction step 516 in process 480 can be replaced with a more sophisticated leak reduction procedure described herein. Advanced leak reduction procedures are processes that can further reduce leak errors. For example, by obtaining the outer product after searching the grid table, accurate multidimensional matching can be performed in the grid cell. The cross product operation includes the following processing.

  a. As described above, one-dimensional accurate matching is performed using any one-dimensional accurate matching method. The main feature of this alternative step is that a list of filter labels is created rather than a list of forwarding nodes.

  b. If the list is empty (ie there is no match), the packet is excluded.

  c. If there is at least one match, the search process proceeds to another attribute. Steps a through c are repeated until any of the following short-circuiting conditions are met. (1) A mismatch is detected in any dimension. (2) There is no common filter label in all the filter label lists generated by the previous one-dimensional inspection. (3) All dimensions were examined.

  d. If none of the exclusion conditions shown in step c is met and there is no mismatch in all attribute checks, a decision is made to accept the packet.

  For example, in the above diagram, a packet having attribute values (a, d) generates F2 for the $ 1 check and F1 for the $ 2 check. Since there is no overlap between the two filter label lists, the packet is excluded. On the other hand, packets with attribute values (a, c) generate F2 in both $ 1 and $ 2 checks, so the packet is accepted. Since this procedure is substantially accurate in the final multidimensional analysis, it can be said that this is not just a test to exclude packets, but a test to accept packets. An approximation is only made for truncation in the filter attribute check. This problem is not an unsolvable problem. For example, this problem can be avoided by the following procedure.

  1. Create multiple grid tables containing subsets of attributes corresponding to a complete set of subset permutations (eg, ($ 1, $ 2, $ 3) as one set, ($ 2, $ 4, $ 5) is selected as the other set). A filter with a long list of predicates is covered by at least one such grid table.

  2. For all attributes, filters with many predicates that perform cross product checking are inherently selective, and any of the packet exclusion conditions outlined in step c (eg, filter label list AND) The probability of short-circuiting is very high because it can occur before the attribute is examined.

  Filters containing many predicates are impractical and rare, so truncation of the filter expression is necessary but not frequently done in practice.

  As described above, the hybrid method and the multi-table search method combining spatial quantization and outer product constitute further main elements of the present invention, and as described later, these processes are all performed by the outer product method. Solve the problem of lack of scalability in the traditional method used.

  For point filters, the outer product method is accurate. However, in the case of a range filter, there are two problems in the process of performing an outer product as a list of filter labels to be stored between any pair of adjacent filter boundaries. As a first problem, information stored when a range filter having a high degree of overlap is inserted into the system is on the order of N2, where N is the number of filters. Second, the list AND step is on the order of N, and this process is very slow when N is large. In a hybrid method that combines grid search and cross product processing, only the filters contained in the same grid cell need be stored, and the packets contained in the boundary of the grid cell need only be matched, reducing this problem to a reasonable size. it can. In a sense, the grid cell takes over the role of storing the filter label list from between adjacent filter boundaries, and the number of grid cells is reduced, so the filter label list needs to be stored.

  If the grid is constructed based on the filter boundary distribution, an even number of filters will result in each grid cell spacing in all attribute dimensions in any attribute dimension. If N is small, the storage problem N2 is not a big problem. For example, in a grid with 32 intervals, the storage problem is mitigated by (N / 32) 2 * = N / 32. Further, the problem of the speed of the list logical product processing is similarly reduced.

  If the number of remaining grid cells is actually small, the cross product process described above works well. By dividing the attributes into various permutation pairs and assigning one CBR table for each pair, the number of grid cells in each table can be limited to a reasonable number. For example, if the number of departments on each attribute axis is L, the total number of grid cells for k attributes is LK. By using the multiple table (size = 4) method, the size is reduced to ML4. Where M is the permutation K! / (K4!) 4! The number is smaller than the total number. Note that some permutations overlap greatly, and one of them may be selected as a representative (for example, {$ 1, $ 2, $ 3}, {$ 1, $ 2, $ 4}, {$ 1, $ 2, $ 5}).

  The CBR processing (for example, processing 480) according to this embodiment (excluding the accurate filter matching steps 516 and 524 for each predicate) has at least the following characteristics. 1) The packet transfer rate is very fast, comparable to IP routing and requires only O [k log (L)] steps. Here, L is the number of departments on some attribute axis, and k is the number of attributes that are quantized and mapped. If there are 4 attributes and each attribute is divided into 32 regions, the required steps are about 22 steps. 2) The packet transfer rate is constant in real time. That is, it does not depend on the number of filters (N). This property makes it possible to estimate the maximum bandwidth that can be maintained by this CBR process. 3) Also, dynamic updates are fast as well. For example, filter insertion is realized in O (Lk) steps, depending on the embodiment. 4) Storage requirements are small and limited, approximately k (L-1) SA + NrLK / 8. Here, SA is the size of the attribute data type expressed in bytes, and Nr is the number of interfaces transferred in the router. The first term is the total size of k index components in the data structure, and the second term represents the actual transfer component (ie, CBR table) to which the index is attached. For example, when a router having 8 interfaces uses 4 integer attributes and 32-divided CBR, the amount of storage is only 1 MB. 5) The indexed data structure is separated from much larger transfer components and is very easy to cache. For example, in the above-described embodiment, the size of the index structure is only 4 × 31 × 4 = 496 bytes. 6) The grid is also a tool for grouping and organizing filters. This is particularly useful for merging filters before propagating the filters, which makes the process very scalable in systems with a large number of subscribers. Thus, the CBR process 480 is fast, predictable, dynamic, scalable, does not require a large amount of memory, is flexible, and thus can be used practically in the core of a publish-subscribe network. is there.

Adaptive Grid As described above, the grid is formed by spatially quantizing the attribute space. The attribute space is usually quantized uniformly. Here, by performing adaptive spatial quantization on the attribute space, for example, based on the past packet distribution, the resolution of the grid and, therefore, the efficiency of the CBR processing 480 can be improved. With adaptive quantization, the region of the attribute space where the most matching packets are expected to be received is determined. For example, in adaptive quantization, it is possible to determine an attribute space region where past packet distributions are most concentrated. The determined region or regions are quantized with higher accuracy than other quantization regions in the attribute space. In other words, in the determined area, the number of grid cells is larger and the size of the grid cells is smaller than in other quantization areas of the attribute space. This is shown in the two diagrams of the adaptive grid below.

  In FIG. 1, a grid with higher accuracy is overlaid in an area where packet traffic is crowded. In the second diagram showing an alternative approach, for example, in a region surrounded by quantized coordinates (3, 2) (4, 2) (3, 3) (4, 3), grid cells The grid spacing on the attribute axis has been changed to be more accurate and focused.

  The advantages of such an adaptive grid are numerous. As can be seen from this figure, the region Z is adaptively quantized. In the region Z, 64 (64) grid cells are provided in the region of 4 (4) grid cells unless adaptation is performed. As shown in this figure, the filter coverage information sS includes three grid cells partially covered before processing, which are 1 (1) partially covered grids by the adaptive grid. Only cell Y is present. Thus, the potential leakage potential is significantly reduced. Furthermore, if one or more of the 64 grid cells in region Z are partially covered, the potential leakage potential in each of these grid cells is even greater than the partially covered grid cell Y. Becomes smaller. This is because, in the region Z, the relaxation region in each partially covered grid cell is significantly smaller than the relaxation region in the partially covered grid cell Y. Therefore, the efficiency of the CBR process 480 can be increased by using an adaptive grid.

  In the second method, since the same number of grid cells are generated, there is an advantage that the storage requirement does not increase as compared with the first method. However, the second method is inferior in accuracy compared to the first method.

Implementing Query Response with Issue-Subscribe Architecture Here, a series of approaches using an issue-subscribe network that supports query-response type interaction is described. This process includes a query response style, ie, advertisement (also called offer), query and response mapping, that is different from the basic abstraction of publish-subscribe style, ie publish (also called notification) and subscription. As a further change, here we distinguish between a single set of advertisements and queries (characterizing a single content item) and a set of advertisements and queries (characterizing multiple content items).

  As described above, in a publish-subscribe network, publishers create and distribute content items called publishing or notifications. In many systems, this information is structured as a tuple of categorized attributes, and in some systems, notifications are initially classified using a distinct attribute called a subject. Subscribers create and send subscriptions that characterize the notifications they are interested in. Notifications are often a single set of content items, and subscription subscriptions use a complex representation to characterize a collection of interesting notifications. The same entity can function as both an issuer and a subscriber. The interaction is initiated by the publisher publishing the content item to the subscriber. This push distribution respects some sort of temporal order between subscription reservations and notifications. Specifically, notifications are only verified against subscription subscriptions that are valid at the time of notification and not against future subscriptions.

  Query response style, a different form of interaction, is the basis for many common distributed applications, including Internet search engines and peer-to-peer content content sharing systems. In a query response software application, an entity called an advertiser producer provides content to other entities, and the ad creator can announce or provide the presence of content via advertisements. Provides information characterizing an item or collection of content items. The other entity, called the requester or consumer, broadcasts queries or requests that characterize the single set of content items or sets of content items that are of interest. In response to the consumer, the application identifies a matching advertisement and delivers it to the consumer. Thus, the interaction is initiated by the consumer asking (pulling) the advertisement and the content item itself from the creator. This pull-type transmission respects some sort of temporal order in advertisements and queries. Specifically, the query gets a response only for advertisements that are valid at the time the query is issued, but does not get a response for future advertisements.

  Here, a series of processes for supporting query-response pull-type interaction using issue-subscribe push-type interaction will be described. In the following, five exemplary processes for realizing this will be described.

  1. FIG. 24A is a flowchart illustrating a first query response process 600 using an issue-subscribe system. Process 600 may be performed with various publish-subscribe systems. For example, the process 600 may be performed by the processor 93 in the intelligent router 92 (or ASIC 91 with corresponding functionality) under software control, eg, performed by the processors 134, 114 in the user devices 122, 100. May be implemented in software modules represented as agents 128, 105.

  In process 600, the publish-subscribe network is used to match the query with the advertisement, the advertiser registers a single set or set of advertisements using the subscription reservation (step 602), and the requester issues the issue / notify. Use to issue a single set query (step 604). In other words, a single set or set of advertisements is mapped to a subscription reservation, and a single set query is a filter for notification of subscription reservations (advertisements) by a publish-subscribe system that performs CBR mapped to notifications ( Query), all matching notifications are pushed to the advertiser (step 606). The CBR may be executed according to any of the processes described above using the infrastructure described above. The advertiser receives the request delivered via some non-issue-subscription means (eg unicast connection to the requestor) or via an agreement-based arrangement using the issue-subscribe network (eg notification) Therefore, the requester is requested to register an individual subscription reservation (corresponding to the requester's communication).), A notification is received and a response is returned to the requester (step 608).

  FIG. 24B is a sequence diagram further illustrating the process 600. FIG. 24B shows the relationship between an advertiser (eg, using publisher device 100), a CBR infrastructure (eg, network of intelligent routers 92), and a requester (eg, using subscriber device 122). Represents an interaction. This sequence diagram preferably illustrates the operation of multiple processors (eg, processor 93, processor 114, and processor 132) that perform process 600.

  2. FIG. 24C is a flowchart illustrating a second query response process 610 using the issue-subscribe system. Process 610 can be performed by various publish-subscribe systems. For example, the process 610 may be performed by the processor 93 in the intelligent router 92 (or ASIC 91 having a corresponding function) under software control, for example, by the processors 134, 114 in the user devices 122, 100. May be implemented in software modules represented as agents 128, 105.

  In process 610, the author delivers the content item twice using a pair of related subjects. The author issues a notification with one subject of this pair, called the initial subject, for the initial advertisement of the single set content item (step 616). In addition, the creator periodically reissues the advertisement previously issued by the initial subject using the other subject of the pair, called the past subject issuing the notification (step 618). The consumer submits one subscription reservation for the initial subject and receives a new advertisement (step 612), and submits another subscription reservation for the past subject to receive a previously published advertisement (step 614). ). Alternatively, the consumer may subscribe only to previously published advertisements. A previously published advertisement that matches this (and a new advertisement if subscribed) is pushed to the consumer using the CBR (step 620). At some point, the consumer stops subscribing for the past subject (step 622).

  FIG. 24D is a sequence diagram that further describes process 600. FIG. 24D shows the relationship between an advertiser (eg, using publisher device 100), a CBR infrastructure (eg, network of intelligent routers 92), and a requester (eg, using subscriber device 122). Represents an interaction. This sequence diagram preferably illustrates the operation of multiple processors (eg, processor 93, processor 114, and processor 132) that perform process 610.

  3. FIG. 24E is a flowchart illustrating a third query response process 630 using the publish-subscribe system. Process 630 can be performed by various publish-subscribe systems. For example, the process 630 may be performed by the processor 93 in the intelligent router 92 (or ASIC 91 having a corresponding function) under software control, for example, performed by the processors 134, 114 in the user devices 122, 100. May be implemented in software modules represented as agents 128, 105.

  The publish-subscribe network is added with a caching function as described above, for example, so that previously issued notifications are cached (step 632) and can be used by future subscribers. The creator issues a notification to advertise the single set content item (step 634). Consumers both to request previously advertised content items of interest (provided by playing relevant notifications from the cache) and to obtain future advertisements of interested content items A single subscription is submitted for purposes (step 636). Similar to traditional publish-subscribe networks, subscription subscriptions characterize a collection of content items of interest. A previously issued notification that matches this is replayed from the cache (step 638) and pushed to the consumer using the CBR (step 640).

  FIG. 24F is a sequence diagram for further explaining the process 630. FIG. 24F shows between an advertiser (eg, using publisher device 100), a CBR infrastructure (eg, network of intelligent routers 92), and a requester (eg, using subscriber device 122). Represents an interaction. This sequence diagram preferably illustrates the operation of multiple processors (eg, processor 93, processor 114, and processor 132) that perform process 630.

  4). 24G and 24H are flowcharts illustrating a fourth query response process 650 using the issue-subscribe system. Process 650 can be performed by various publish-subscribe systems. For example, the process 650 may be performed by the processor 93 in the intelligent router 92 (or ASIC 91 having a corresponding function) under software control, for example, by the processors 134, 114 in the user equipment 122, 100. May be implemented in software modules represented as agents 128, 105.

  FIG. 24G shows the data advertisement phase of process 650. First, a data advertisement is received (step 652). Necessary for consumers to use subscriptions to identify query sets, authors to use notifications to advertise single-set content items, and for such uses for subscriptions and notifications In order to achieve a temporal order reversal, an ontology transformation is performed that transforms the data advertisement into an equivalent subscription (step 654). Subscription subscriptions (data advertisements to be converted) are registered in the CBR network (eg, subscription subscriptions are stored in the intelligent router 92) (step 656).

  FIG. 24H shows the query response phase of process 650. First, a query is received (step 658). Similar to the data advertisement described above, the query is converted into an equivalent notification by ontological conversion (step 660). The conversion from the old space (eg, data advertisement) to the new space (eg, equivalent subscription) in step 654 and step 660 begins with the combination of attributes and expression operators from the old space in the new space. Includes processing to define attributes to represent. This transformation further includes creating notifications and subscriptions in the new space that are equivalent to queries and advertisements in the old space, respectively. The conversion must meet the requirement that the notification matches the subscription if and only if the unconverted data advertisement matches the unconverted query.

  The notification (converted query) is pushed to the advertiser by the publish-subscribe system that executes the CBR, using the subscription (converted data advertisement) as a filter for the notification (step 662). The CBR may be executed according to any of the processes described above using the infrastructure described above. The advertiser receives the transformed notification via some non-issue-subscription means (eg unicast connection to the requester) or via an agreement-based arrangement using the issue-subscribe network and responds to the requester Data is returned (step 664).

  FIG. 24I is a sequence diagram that further describes process 600. FIG. 24I illustrates between an advertiser (eg, using publisher device 100), a CBR infrastructure (eg, network of intelligent routers 92), and a requester (eg, using subscriber device 122). Represents an interaction. This sequence diagram preferably illustrates the operation of multiple processors (eg, processor 93, processor 114, and processor 132) that perform process 650.

5. In the fifth process, additional interface points and / or message formats extend the functionality of the publish-subscribe network so that authors can distinguish between notifications, advertisements and responses, and consumers can Can distinguish queries. By separating the various primitives in the network, both styles of interaction can be achieved while using a uniform set of data structures, algorithms and protocols.
Persistent and reliable message delivery In the following, in an publish-subscribe network, an apparatus for the persistent storage and playback of messages related to network, link and / or node failures, the resumption and / or continuation of message flows and A method will be described. Message persistence refers to the property that a message can be stored and the stored message can be retrieved later. Many specific applications, such as e-mail, typically require long message persistence for messages communicated over a network. In ideal conditions where no network failure occurs, always-connected subscribers do not require any persistence beyond what is required in these particular applications. In reality, however, messages may be “lost” for various reasons while being transmitted over the network. The reason for this is, for example, (1) a failure or buffer overflow on the network or user side, or (2) a case where the user disconnects the network and then reconnects after a predetermined period.

  The event notification platform persistence model in this example is divided into two levels: short-term persistence and long-term persistence. Short duration restores packets lost due to network congestion or short term link failures. Long-lasting recovers lost packets due to other failures, including, for example, user connection failures, user device failures, application failures and / or failures in the network for longer periods of time.

The channel (eg, as described above) may be persistent or real time. Real-time channels typically transmit useful data only in real time and do not have any application-specific persistence requirements. A persistent channel stores data transmitted over the network for a sustained time frame T. In other words, in a persistent channel, persistence is guaranteed for the time frame T. This persistence of data is realized, for example, as follows. Data is cached at each edge node for the duration of the channel. Data is read from a cache that is transparent to the user under conditions of failure. Allow users to retrieve data from the cache. Prevents router failures and establishes a reliable tunnel between routers to make the flow of data through the network persistent. Duplication protects the channel components from failure.
A. Persistence by caching Timed persistence (by time frame T) is the ability to retrieve data in the previous time frame T from the publish-subscribe network. For example, if a subscriber disconnects from the network, any data about the persistent channel that is received while the subscriber is absent is retained on the network until the time frame T elapses (from receipt of the data). The If the user returns within time frame T, no data is lost.

  FIG. 25A is a block diagram illustrating a portion of a publish-subscribe network that provides persistence through caching. As shown in FIG. 25A, the network includes a core routing node and an edge routing node. Each routing node preferably comprises an intelligent router 92 (shown with an edge routing node) and a conventional backbone router (not shown), as described above with reference to FIG. Each intelligent router 92 that needs to perform caching for persistent channels comprises a cache manager 218, preferably in the same location, as shown in FIG. 25A. The cache manager 218 is as described above with reference to FIG. Intelligent router 92 preferably provides short-term persistence for retrieving lost data or recovering from a router failure. The cache manager 218 caches data to achieve long-term persistence in the channel. Cache manager 218 preferably caches this data in cache 540. Cache 540 preferably includes memory and disk (not shown). The agent 128 shown in FIG. 25A is preferably provided in the subscriber device 122 (not shown in FIG. 25A), as described above with reference to FIG. Agent 128 communicates with cache manager 218 to retrieve data from cache 540, receive the retrieved data, and organize the retrieved data.

  Although not shown in FIG. 25A, each core routing node at the first level upstream from the edge routing node preferably comprises a cache manager 218. The first level upstream core routing node means a routing node adjacent to the upstream side of the edge routing node. Although the publish-subscribe network often includes multiple first level core routing nodes, FIG. 25A shows only one first level core routing node, ie, core routing node 548. Yes. As described above, the cache manager 218 realizes local caching of data in the network node where the cache manager 218 exists. Thus, for example, the operation of the cache manager 218 present at various core routing nodes, including the core routing node 548, provides distributed caching of data within the network core. This distributed caching provides a backup for edge routing node caching.

  Cache manager 218 preferably includes routines or subroutines that manage cache 546. These routines preferably include a data caching routine 545, a cache data search routine 547, and a cache data purge routine 549, as shown in FIG. 25B. These routines can each operate separately, or can communicate with each other and / or pass control, as shown in FIG. 25B.

  FIG. 25C is a flowchart of a persistent caching process 550 for data contained in a single message. As shown in FIG. 25C, the process 550 shows a specific example of steps executed by each routine shown in FIG. 25B. In actual operation, these routines are executed independently of each other. Process 550 may be performed at intelligent router 92, edge routing node, or core routing node that provides caching.

  The process 550 can be realized by, for example, a software module (for example, the cache manager 218) executed by the processor 93 in the intelligent router 92. Alternatively, this processing may be realized by an ASIC, or may be realized by a combination of hardware and software. These may be the same physical device as the corresponding intelligent router 92 or a different physical device. Intelligent router 92 forwards messages that need to be cached to cache manager 218. The cache manager 218 receives the message containing the data (step 552), puts a mark indicating the time on the data in the message (step 554), and caches the data in the cache 546 (step 556) (step 322 in FIG. 14). To 326 (see above)). As described above, the cache manager 218 indexes data cached by various methods including, for example, a channel identifier, a subject, an issuer identifier, a time stamp, and the like (step 558). The index assigning step may be performed before the caching step 556, may be performed at the same time, or may be performed after this. The cached data is indexed and stored in a hierarchical directory structure (see, for example, FIG. 15). Tables 17-19 also relate to this process.

  Data is cached in memory (cache 546) and periodically transferred to disk (step 560). The sustained time frame “T” for a particular channel is divided into N time grains, each having a size G. Caching in the memory (step 556) is performed only for the period G. If the cache manager determines that period G has elapsed (step 559), the data is moved to disk (step 560). The cache manager 218 stores data on the disk for a sustained timeout period T.

  When the time exceeds the duration timeout (T) + the upper limit of the period (step 562), the data corresponding to the period G is deleted from the disk (step 564). For the sake of explanation, assume that T is 2 hours for a certain channel. Furthermore, it is assumed that the cache manager 218 uses a time grain G of 15 minutes. To delete data from the disk, preferably if the last data cached during period G (15 minutes) is stored for T (2 hours), the entire cached data for that 15 minute period Use a policy that discards the data. Therefore, the data cached at the beginning of the 15 minute period is stored for more than 2 hours before being discarded. In this embodiment, the data cached every 15 minutes is a block of data. If the duration time frame T is divided into N periods, there will be N + 1 data blocks (N on disk and 1 in memory) in the cache 540 for each subject at any time.

  When cache manager 218 receives a request for data (step 566), cache manager 218 uses the index to retrieve the requested cached data (time period T has not elapsed) (step 568); The retrieved data is forwarded to the backbone router for routing to the requested subscriber device 122 (step 570). As will be described later with reference to FIG. 25D, in the edge routing node, if the data is not in the cache 540, the cache manager 218 may clear the cache in the upstream first level core routing node (eg, core routing node 548). Call and retrieve data. If cache manager 218 does not receive the request, process 550 returns to step 552.

  In addition to storing data in cache 546, in the event of a failure in cache manager 218 of intelligent router 92, the data is cached in a first level upstream core routing node (eg, core routing node 548). Is also saved. Accordingly, steps 552-564 are preferably performed at the first level upstream core routing node before routing the message and before performing the process 550 at the edge routing node.

  The location of the cache in cache 546 (eg, within the edge routing node) and the first level upstream core routing node (eg, first upstream core router 548) is stored locally at agent 128. Thus, if the subscriber device 122 has moved (eg, if the subscriber is a mobile phone subscriber reconnecting to the network via a different edge routing node), the agent 128 can The subscriber device 122 informs the cache location of the edge routing node and the first level upstream core routing node that were connected before moving. Thereby, even if the subscriber device 122 moves, the data persistence can be maintained.

  As shown in FIG. 25D, which is a flowchart of the persistent message retrieval process 580, the user application 182 of the subscriber device 122 preferably calls the agent 128 to request data from the cache (step 582). Agent 128 uses the saved cache location to invoke cache 546 at the appropriate edge routing node (eg, using a getcache command) to obtain data (step 584). The agent 128 may specify which time frame of data to read. Alternatively, the cache manager 218 may examine the connection history for the subscriber device 122 (see Table 18 above) to determine which cached data to read. If the connection history reveals that the subscriber device 122 has been offline for period X, the cache manager 218 may retrieve all data that has been cached for period X. Further, the cache manager 218 may simply retrieve all the data cached during the sustained time frame T. If there is data in the cache 546 of the edge routing node (step 585), the agent 128 reads the data from the cache 546 (step 586).

If the cache 546 has no data, the cache manager 218 uses these lists of caches at the first level upstream routing node (eg, core routing node 548) provided by the agent 128 to The upstream cache is called to retrieve data (step 588). When cache manager 218 detects the upstream cache or the location of the data in the cache, cache manager 218 reads the data from the upstream cache (step 590). The cache manager 218 may transfer the data to the agent 128 (step 594) after executing the duplication suppression processing (that is, processing for eliminating duplicate data) (step 592).
B. Persistence with Fault Protection The infrastructure according to the present invention includes a channel manager, event agents, routing nodes and caches, or equivalent components. To prevent messages from being lost during a failing condition. Each of these components needs to be protected.

1. Channel Manager Failure Protection The channel manager has been described above with reference to FIG. The channel manager is protected from failure by duplication. The channel manager operates in a primary backup group that includes one primary module and another backup module. The primary module is responsible for performing all updates related to data and responding to queries. The backup module only responds to the query and does not have the ability to perform updates, i.e. the backup module sends all updates to the primary.

  The component of the publish-subscribe network that needs to retrieve channel information from the channel manager is called the channel manager client, and the channel manager client is assigned a channel manager. These clients include intelligent router 92 and its cache manager 218, subscriber device 122 and its agent 128, and the like. When a channel manager assigned to a client fails, the client accesses and connects to another channel manager.

2. Router and cache fault protection There are several levels of recovery for routing infrastructure components (level 0, level 1 to level 3). Level 0 controls the impact of the failure locally (ie, within the failed router) and recovery is limited to the failed routing node. Level 0 recovery relies on hardware redundancy to recover from a hardware failure. That is, the router hardware resources are duplicated. The main benefits of level 0 recovery are short recovery time and transparency. Level 1 controls the impact of failures within the LAN. Level 1 recovery is realized by IP failure bypass using cluster technology. However, the tunnel from the failed router to the neighboring router needs to be rebuilt after recovery. Therefore, level 1 recovery is less transparent than level 0 recovery and takes time to execute.

  The purpose of Level 2 recovery is to recover the router failure by the remote backup router on the WAN without impacting the network topology causing complex reconfiguration and data loss. Level 2 recovery implements a set of recovery by grouping routers (eg, primary and backup routers). One router can function as a primary router in one group and can also function as a backup router in another group. The primary router exchanges information about routing and network configuration with the backup router, and when the backup router detects a failure in the primary router, the backup router takes over the function of the primary router. Make sure that the failure is the primary router, not the link to the primary router, by making sure that the backup can connect to the majority of the primary routers adjacent to the primary router that you suspect has failed To do. Level 3 recovery is triggered when it is not possible to connect to the majority of the primary routers adjacent to the primary router suspected of having failed.

  Level recovery 3 performs coordinated recovery across the entire system. Level 3 recovery is performed when everything else has failed. A router that detects such a failure sends a restart notice to all downstream routers. The router that received the notification suspends all other recovery, stops all network services, and deletes all multicasting information in the resume message. When all network services are finished, the router resumes all services including the tree building process for multicasting. Any failure in the network is handled by a lower level recovery procedure. If a lower level recovery is not applicable, the recovery may perform a higher next level recovery sequentially. If none of the low level recovery works, level 3 recovery is used.

3. Reliable tunnels Use reliable tunnels that carry traffic between routing nodes to prevent message loss during network failures. The tunnel is as described above with reference to FIG. Any message loss during the transmission period is detected by the receiving terminal, and the receiving terminal asks the transmitting side to retransmit the lost message.

  Although the present invention has been described using exemplary embodiments, it will be apparent to those skilled in the art that these embodiments can be variously modified and the present invention includes any adaptations or variations thereof. For example, various types of issuer devices, user or subscriber devices, their channels and configurations, content-based routing and other functions can be implemented in hardware and software without departing from the scope of the present invention. is there. The invention is limited only by the claims and the equivalents thereof.

  10 Network core, 14, 32 Issuer, 16, 22 Edge router, 18 channels, 12, 38, 40, 42, 44, 46, 48, 61-70, 92, 164, 166, 168, 200 Intelligent router, 24 , 54 subscribers, 20, 26 messages, 30 routing entities, 34, 52, 126 applications, 30 channel services, 36, 50, 184, agents, 39, 41, 43, 45, 37, 47, 73-85 links, 58, 59, 60, 120 network, 71, 72 entity, 90 network node, 93, 114, 134, processor, 94, 102, 124 memory, 95 backbone router, 97 auxiliary storage device, 112, 130 auxiliary storage device, 100, 18 issuer device, 104 issuer application, 106, 128 agent application, 108, 132 input device, 110, 136 output device, 116, 138 display device, 122, 140 subscriber device, 150, 198 channel manager, 152, 154 156 server, 158, 160, 162 storage device, 180 stack, 182 user application, 186 event library, 188 cache library, 190 channel library, 192 messaging library, 194 dispatcher library, 196 control path, 210 software component, 212 filtering daemon 214 dispatcher, 216 routing daemon, 218 cash money Turbocharger, 204, 222 data path

Claims (193)

A communication method for performing publish-subscribe processing on an unreliable network, comprising:
Receiving a subscription subscription for content via the network;
Publishing the content via the network based on a subscription to the content;
Receiving a notification regarding the content at a node in the network;
Determining whether the notification should be forwarded to an adjacent node;
Selectively forwarding notifications to the adjacent nodes using a reliable transmission protocol based on the determination.   The step of selectively transferring the notification includes the step of receiving instruction information indicating information indicating failure of transfer of the notification, and retransmitting the notification according to the instruction information. The communication method described.   2. The communication method according to claim 1, wherein the transferring step includes a step of transferring a packet encapsulating the notification.   The communication method according to claim 1, wherein the transferring step uses an Internet protocol for transferring the notification.   The communication method according to claim 1, wherein the transferring step includes a step of transferring a notification for each single node.   The communication method according to claim 1, wherein the determining step includes a step of determining whether or not the content satisfies one or more conditions in a subscription for the content.   The communication method according to claim 1, wherein the transferring step includes a step of transferring a notification to the adjacent node using a router.   The communication method according to claim 1, further comprising the step of repeating the issuing step for each node in the network until a notification is distributed to one or a plurality of subscribers corresponding to the subscription.   The communication method according to claim 1, wherein the step of selectively transferring the notification uses a transmission control protocol.   The router for executing the issue-subscribe process on a low-reliability network, comprising a module that implements the communication method according to claim 1. A wide area network to perform publish-subscribe processing,
One or more subscriber devices;
With multiple routers,
Each of the plurality of routers
A subscription reservation receiving module for receiving a subscription for content corresponding to one or more of the subscriber devices via the wide area network; and a notification receiving module for receiving a notification regarding the content at a node in the network; A content-based routing module that determines whether the notification should be forwarded to an adjacent node, and a selective transmission of the notification to the adjacent node using a reliable transmission protocol based on the determination And an issuance module having a reliable transfer module for transferring to
The wide-area network, wherein the plurality of routers selectively transfer the notification with high reliability until the notification is distributed to one or a plurality of subscribers corresponding to the subscription reservation. A communication method for performing publish-subscribe processing on an unreliable network, comprising:
Receiving a subscription for content at a first node in the network via the network;
Receiving a notification regarding the content at the first node in the network;
Determining whether to forward the notification to an adjacent second node using content subscription based on content-based routing (CBR);
Selectively forwarding notifications to the adjacent nodes using a reliable transmission protocol based on the determination. Receiving instruction information indicating a transfer error when transfer between the first node and the second node fails;
13. The communication method according to claim 12, further comprising a step of retransmitting the notification to the second node in accordance with the instruction information.   The communication method according to claim 12, further comprising a step of buffering the notification in a buffer before performing the selective transfer.   The communication method according to claim 14, further comprising the step of transferring the buffered notification to an auxiliary storage device after a predetermined period of time has elapsed.   The communication method according to claim 14, further comprising a step of deleting the buffered notification in a first-in first-out manner.   The communication method according to claim 14, further comprising the step of deleting the buffered notification based on quality of service.   The communication method according to claim 12, further comprising a step of caching the notification in the cache of the first node.   13. The communication method according to claim 12, wherein the step of transferring with high reliability uses a transmission control protocol.   13. A router that executes issue-subscribe processing on a low-reliability network, comprising a module that implements the communication method according to claim 12. A wide area network (WAN) for performing publish-subscribe processing,
One or more subscriber devices;
With multiple routers,
Each of the plurality of routers
A subscription receiving module for receiving subscriptions for content corresponding to one or more of the subscriber devices via the WAN at a first node in the WAN; and a first node in the WAN. A notification receiving module for receiving a notification about the content, a content-based routing module for determining whether the notification should be transferred to the second node in the WAN, and a reliable transmission protocol based on the determination Using an issuance module having a reliable transfer module for selectively transferring notifications to the second node,
The wide-area network, wherein the plurality of routers selectively transfer the notification with high reliability until the notification is distributed to one or a plurality of subscribers corresponding to the subscription reservation.   The wide area network according to claim 21, wherein the reliable transfer module uses a transmission control protocol. A communication method for performing publish-subscribe processing on an unreliable network, comprising:
Establishing a reliable tunnel between two adjacent nodes using a reliable transmission protocol;
Receiving a subscription for content at the first of the adjacent nodes via the network;
Receiving a notification regarding the content at the first node in the network;
Determining whether to forward the notification to an adjacent second node using content subscription based on content-based routing (CBR);
And forwarding the notification to the adjacent second node using the reliable tunnel based on the determination.   24. A router that executes issue-subscribe processing on a low-reliability network, comprising a module that implements the communication method according to claim 23.   25. A network comprising a plurality of routers according to claim 24 and realizing issue-subscribe processing. A communication method for propagating a filter in a publish-subscribe network, comprising:
Receiving a plurality of filters related to subscription subscriptions of content in the network;
Reducing the number of filters based on certain criteria;
Propagating the reduced-number filter to match the content subscription in the network. The above reduction step is
Performing an approximate process for one or more of the plurality of filters;
27. The communication method according to claim 26, further comprising at least one of a step of executing a merging process on one or more of the plurality of filters.   28. The communication method according to claim 27, wherein the step of executing the merging process includes a step of determining whether or not the merged filter exceeds a packet leakage threshold.   27. The communication method according to claim 26, wherein the receiving step comprises receiving a plurality of filters via a content-based routing network.   The propagation steps include: active propagation schemes, passive propagation schemes, broadcast advertisement response schemes, after-network topology computation schemes, periodic propagation schemes, schemes for repropagating filters upon failure recovery, network reconfiguration schemes 27. The communication method of claim 26, comprising the step of propagating the filter using one or more of a plurality of specific procedures including a scheme for repropagating the filter in between.   The source node and the destination node for propagating the filter may have at least one of a plurality of specific mechanisms including a hop-by-hop scheme, an end-to-end scheme, a tunnel scheme, and a transmission control protocol scheme. 27. The communication method according to claim 26, wherein the communication method is connected by using.   27. The communication method according to claim 26, further comprising the step of inserting a filter in the network at one or more of a plurality of specific locations in the network.   Propagating the filter comprises propagating the filter based on at least one of the specific behaviors of the intended recipient selected from moving, persistent and temporary behavior. 27. A communication method according to claim 26, wherein:   Propagating the filter comprises propagating the filter based on at least one of the one or more filter sender specific behaviors selected from moving, persistent and temporary behaviors The communication method according to claim 26, further comprising:   27. Performing one or more of a plurality of specific processing steps including attribute content conversion, expression expansion, expression conversion, display transformation, encoding / decoding, encryption / plain culture. The communication method described.   27. The communication method of claim 26, comprising formatting a message including the one or more filters based on one or more of a plurality of specific formats.   Propagating the filter comprises propagating the filter according to one or more of a plurality of specific policies or procedures including management, maintenance, filter persistence, policy highlighting performance, monitoring, and billing. The communication method according to claim 26. The step of propagating the filter is:
Determining a network topology underlying the filter propagation strategy, including one or more of a plurality of requests for filter propagation that must be satisfied to propagate the filter;
27. The communication method according to claim 26, further comprising the step of reducing the number of filters based on satisfaction of the determined filter propagation strategy requirement.   The determined filter propagation strategies are: active propagation scheme, passive propagation scheme, broadcast advertisement response scheme, after-network topology calculation scheme, periodic propagation scheme, scheme for re-propagating filters upon failure recovery, network 39. The communication method of claim 38, wherein the communication method is one of a scheme for repropagating a filter during reconfiguration.   27. The communication method according to claim 26, wherein the receiving step includes a step of receiving a plurality of filters at a core router in the network.   27. The communication method according to claim 26, wherein the receiving step includes a step of receiving a plurality of filters at an edge router in the network.   27. The communication method according to claim 26, wherein the receiving step includes a step of receiving a plurality of filters at an issuer device in the network.   27. A router that propagates a filter in an publish-subscribe network, comprising a module that implements the communication method according to claim 26.   An issue-subscribe network comprising a plurality of routers comprising modules for implementing the communication method of claim 26.   An issue-subscribe network comprising a user device comprising a module for implementing the communication method according to claim 26. A communication device for propagating a filter in a publish-subscribe network,
A filter receiving module for receiving a plurality of filters related to content subscriptions in the network;
A filter reduction module that reduces the number of the filters based on specific criteria;
A communication device comprising: a filter propagation module for propagating a filter with a reduced number in order to match a subscription to the content in the network. A filter estimation module that performs an estimation process on one or more of the plurality of filters;
The communication apparatus according to claim 46, further comprising: a filter merging module that performs a merging process on one or more of the plurality of filters. A filter processing module that transforms the one or more filters;
The communication apparatus according to claim 46, further comprising: a filter grouping module that groups filters based on mutual similarity of the filters. A communication method for propagating a filter in a publish-subscribe network, comprising:
(A) receiving a plurality of filters at a node of the network;
(B) processing the received one or more filters to reduce the number of filters;
(C) determining whether to propagate the filter based on the rules of behavior of the recipient;
(D) Propagating the processed filter to the next node in the network.   50. The communication method according to claim 49, further comprising the step of determining whether or not to propagate the filter based on rules of behavior of the sender.   50. The communication method according to claim 49, further comprising the step of determining whether or not there is an additional node after the next node in the network.   50. The communication method according to claim 49, further comprising the step of repeating the steps (b) to (d) when it is determined that there is a further node after the next node.   50. The communication method of claim 49, further comprising the step of determining whether a next node in the network has additional filters to receive.   54. The communication method according to claim 53, further comprising the step of repeating the steps (a) to (d) at a new node when it is determined that there is a further filter to be received at a next node in the network.   The step of processing the filter includes the steps of simplifying the filter by removing redundant predicate expressions, estimating the filter, clustering the filter, including the filter, and merging two or more filters. 50. The communication method according to claim 49, comprising one or more of them. Processing the filter includes grouping the filters;
The step of approximating the filter estimates the filters in the group, and the step of clustering the filters identifies and collects the filters in the group and includes the filters based on the similarity of the filters. 56. The communication method according to claim 55, wherein the step includes a filter in the group, and the step of merging the filter merges the filter in the group.   57. The communication method according to claim 56, wherein the step of grouping the filters comprises the step of grouping only the filters from the downlink associated with the subject.   56. The communication method according to claim 55, wherein the step of merging the filters includes the step of determining whether or not the merged filters exceed a packet leakage threshold.   50. The communication method of claim 49, wherein the step of processing the filter comprises the steps of estimating a filter and including and merging the estimated filter.   The nodes in the network have stored filters, and the step of processing the filters processes the received filters and the stored filters to reduce the number of these filters. The communication method according to claim 49.   61. The communication method of claim 60, wherein the saved filter is a previously received and processed filter.   50. The communication method according to claim 49, wherein said receiving step (a) receives a plurality of filters at a node in the core of the network.   50. A router for propagating a filter in an publish-subscribe network, comprising a module for implementing the communication method.   50. An issue-subscribe network comprising a plurality of routers comprising modules for implementing the communication method of claim 49.   50. An issuance-subscription network comprising a plurality of routers comprising user equipment for implementing the communication method according to claim 49. A communication device for propagating a filter in a publish-subscribe network,
A filter receiving module for receiving a plurality of filters at a node in the network;
A filter reduction module that reduces the number of filters;
Based on the rules of the behavior of the receiver, decide whether to propagate the filter, and if it decides to propagate the filter, a filter propagation module that propagates the reduced number of filters to the next node in the network A communication device provided.   68. The communication device of claim 66, further comprising a reduction module that reduces the filter to remove redundant predicate expressions.   The communication apparatus according to claim 66, wherein the filter processing module includes a grouping module that groups filters based on selection of a downlink.   67. The communication device according to claim 66, wherein the filter processing module includes a rough estimation module for roughening a filter.   The communication apparatus according to claim 66, wherein the filter processing module includes an inclusion module that includes a filter.   The communication device according to claim 66, wherein the filter processing module includes a clustering module that clusters the filters based on the similarity of the filters.   The communication device according to claim 66, wherein the filter processing module includes a merge processing module for merging filters. A communication method for propagating a filter in a publish-subscribe network, comprising:
Sending a notification requesting a filter from the router;
Receiving a notification by one or more downstream routers;
Further propagating the filter request further downstream;
Waiting for a response from a downstream router to which the filter request has been propagated;
Collecting a filter from a downstream router through which the filter request has been propagated;
Processing the propagated filters to reduce the number of filters;
Transmitting the reduced filter to the upstream router that requested the filter. Updating the bookkeeping database at each downstream router to reflect that the filter request has been received from the router;
74. The communication method according to claim 73, further comprising a step of determining whether or not the router that has received the filter request is operating.   An issue-subscribe network comprising a plurality of routers and an issuer device comprising a module for realizing the communication method according to claim 73. A communication method for content-based routing of packets in an issue-subscribe network,
Receiving a packet in the network;
Accessing a map specifying filter coverage in attribute space;
Inspecting the contents of the packet for route determination of the packet;
Using the map to assist in routing the packet;
Determining a route based on the content and map of the inspected packet;
Routing a packet based on the route determination.   77. The communication method according to claim 76, wherein the step of determining a route further comprises a step of collating content with a corresponding subscription reservation.   78. The communication method according to claim 77, wherein the collating step includes a step of identifying a match using a plurality of conjunction filters.   78. The communication method according to claim 77, wherein the collating step includes a step of performing an attribute check.   77. The communication method according to claim 76, wherein the step of determining the route includes a step of using the map online.   77. The communication method according to claim 76, further comprising a step of calculating the filter application range in advance by offline or background processing.   77. The communication method according to claim 76, further comprising a step of calculating a filter application range in advance, and the step of calculating in advance includes a step of approximating and reducing the filter expression.   77. The communication method according to claim 76, further comprising a step of calculating a filter application range in advance, and the step of calculating in advance includes a step of spatially quantizing the attribute space into a grid including grid cells. The step of calculating the filter application range in advance includes
Combining the filter coverage information and the grid;
84. The communication method according to claim 83, further comprising the step of: generating a map based on integration of processing described by the filter application range information in each grid cell.   77. A router that performs content-based routing of packets in an publish-subscribe network, comprising a module that implements the communication method of claim 76.   86. A network comprising a plurality of routers according to claim 85.   77. A computer readable medium having instructions for implementing the communication method of claim 76. A communication method for content-based routing of packets,
(A) spatially quantizing the attribute space;
(B) identifying grid cells covered by a plurality of filters each having at least one described attribute;
(C) updating the content-based routing (CBR) table with processing rules described by a plurality of filters covering the grid cells identified in step (b). Determining whether one or more new filters have been inserted into or removed from the node;
90. The communication method according to claim 88, further comprising: executing steps (b) and (c) at the node when it is determined that one or more new filters have been inserted into or deleted from the node. Receiving issued packets; and
The communication method according to claim 88, further comprising the step of plotting the received packet on the grid.   Inspecting a CBR table for a packet plotted in a grid cell completely covered by the calculated filter coverage information, and determining processing for the packet plotted in the fully covered grid cell The communication method according to claim 90, comprising: Determining whether network traffic exceeds a set threshold;
92. The communication method according to claim 91, further comprising the step of routing the issued packet when the network traffic exceeds a set threshold value.   The communication method according to claim 92, further comprising a step of executing leakage reduction processing when the network traffic is equal to or less than a set threshold value.   94. The communication method according to claim 93, wherein the step of executing the leakage reduction process includes a step of performing one-dimensional accurate matching.   94. The communication method according to claim 93, wherein the step of executing the leakage reduction process includes a process of calculating an outer product.   94. The communication of claim 93, wherein performing the leakage reduction process comprises constructing a plurality of grids and CBR tables based on a complete set of permuted subsets of a limited number and various attributes. Method. Determining if there is a reduced filter predicate or packet plotted in a partially covered grid cell if the network traffic is below a set threshold;
If there is a reduced filter predicate or packet plotted in the partially covered grid cell, perform a filter match for each predicate and any reduced filter plotted in the partially covered grid cell Determining a predicate and a routing process for every packet;
94. The communication method according to claim 92, further comprising the step of routing the packet based on the determined routing process.   The communication method according to claim 88, further comprising the step of reducing the filter expression.   The communication method according to claim 88, wherein said step (a) includes a step of adaptively quantizing the attribute space.   The communication method according to claim 99, wherein the step of adaptive quantization includes a step of adaptively quantizing the attribute space based on packet distribution information.   The communication method according to claim 88, wherein the spatial quantization step includes a step of spatially quantizing the attribute space based on packet distribution information.   The communication method according to claim 88, wherein the spatial quantization step includes a step of updating a grid based on packet distribution information.   The communication method according to claim 88, wherein the step (c) includes a step of generating or updating an operation bit mask for the grid cell covered by the calculated filter application range information.   The communication method according to claim 88, wherein the step (c) includes a step of generating or updating a list of linked processes for the grid cells covered by the calculated filter coverage information. .   90. The communication method according to claim 88, further comprising the step of updating a parameter including statistical information regarding a filter distribution of a cell covered in the CBR table.   The communication method according to claim 88, further comprising the step of collecting statistical information relating to the packet distribution used in step (a).   The communication method according to claim 88, wherein the steps (a) to (c) are performed offline.   The communication method according to claim 88, further comprising performing content-based routing (CBR) using only the CBR table and the rules in the CBR table.   90. A communication device that performs content-based routing of packets in an issue-subscribe network, comprising a module that implements the communication method of claim 88.   The communication apparatus according to claim 109, wherein the communication apparatus is a router.   90. A network comprising a plurality of routers comprising modules for realizing the communication method according to claim 88.   90. A computer readable medium having instructions for implementing the communication method of claim 88. A communication method for realizing query-response interaction in an issue-subscribe network, comprising:
Receiving an advertisement associated with the dataset;
Receiving a query representing a logical expression;
Mapping the advertisement to a corresponding subscription reservation;
Mapping the query to a corresponding notification;
Using the corresponding subscription above to realize the advertisement;
Using the corresponding notification to implement a query in the network.   114. The communication method according to claim 113, wherein the step of mapping the advertisement comprises performing general mapping on the advertisement.   114. The communication method according to claim 113, wherein the step of mapping the query includes the step of performing disjunctive query mapping for the query. A communication method for realizing query-response interaction in an issue-subscribe network, comprising:
Creating a subscription to register an ad;
Issuing a notification and delivering a query;
Performing content-based routing (CBR) using a subscription and pushing a notification to the advertiser;
Returning the response data to the requester.   117. The communication method according to claim 116, wherein the step of returning the response data is performed using a unicast connection between the advertiser and the requester.   In the step of returning the response data, the requester creates a new subscription reservation for the response data, executes CBR using the new subscription reservation, and pushes the response data to the requester. 117. The communication method according to claim 116, further comprising a step.   117. The communication method according to claim 116, wherein the creating step comprises creating a subscription for registering an advertisement related to a set of data.   119. An publish-subscribe network that implements a query-response interaction, the router comprising a module that implements the communication method of claim 116, an issuer device, and a subscriber device. A communication method for realizing query-response interaction in an issue-subscribe network, comprising:
Receiving a subscription to register an advertisement;
Receiving a notification to deliver the query;
Using a subscription to perform content-based routing and determining whether to forward notifications to adjacent nodes;
And a step of selectively transferring the notification based on the determination.   122. A communication device that implements query-response interaction in an issue-subscribe network, comprising a module that implements the communication method according to claim 121.   The communication device according to claim 122, wherein the communication device includes a router. A communication method for realizing query-response interaction in an issue-subscribe network, comprising:
Issuing a notification including an initial subject;
Issuing a notification containing past subjects about previously published ads;
Submitting a subscription to a past subject that is a query for a previously published ad;
Performing a content-based routing using a subscription reservation for a past subject and pushing previously published advertisements to a consumer.   129. The communication method according to claim 124, further comprising the step of submitting a subscription stop for the past subject.   129. The communication method of claim 124, further comprising the step of submitting a subscription subscription for the initial subject.   129. An publish-subscribe network that implements query-response interaction, comprising a router comprising a module that implements the communication method of claim 124, an issuer device, and a subscriber device. A communication method for realizing query-response interaction in an issue-subscribe network, comprising:
Receiving a notification including an initial subject and a notification including a past subject for a previously published advertisement;
Receiving a subscription reservation for an initial subject and a subscription reservation for a past subject that is a query for a previously published ad;
Performing a content-based routing using a subscription reservation for the past subject and determining whether to forward the previously published advertisement to an adjacent node;
Selectively transferring the previously issued advertisement based on the determination.   129. A communication device that implements query-response interaction in an issue-subscribe network, comprising a module that implements the communication method of claim 128.   131. The communication device according to claim 129, wherein the communication device comprises a router. A communication method for realizing query-response interaction in an issue-subscribe network, comprising:
Caching issued notifications;
Issuing a notification advertising the cached content in the issued notification;
Submitting a subscription that is a query to the issued notification;
Replaying the cached issued notification from the cache;
Performing a content-based routing (CBR) based on the subscription and pushing the reproduced issued notification to a consumer.   132. An publish-subscribe network that implements a query-response interaction, the router comprising a module that implements the communication method of claim 131, an issuer device, and a subscriber device. A communication method for realizing query-response interaction in an issue-subscribe network, comprising:
Caching issued notifications;
Receiving a subscription requesting the issued notification;
Replaying the cached issued notification from the cache;
Performs content-based routing (CBR) using a subscription that is a query to the issued notification to determine whether to forward the cached issued notification to an adjacent node Steps,
Selectively forwarding notifications to the adjacent nodes using a reliable transmission protocol based on the determination.   134. A communication device that implements query-response interaction in an issue-subscribe network, comprising a module that implements the communication method of claim 133.   135. The communication device according to claim 134, wherein the communication device comprises a router. A communication method for realizing query-response interaction in an issue-subscribe network, comprising:
Receiving a data ad;
Converting the data ad into a subscription,
Receiving a query;
Converting the above query into a notification;
Performing content-based routing (CBR) based on the subscription and pushing a notification to the advertiser;
Returning a response data to the requester based on the notification.   The communication method according to claim 136, wherein the step of converting comprises the step of performing ontological conversion.   The communication method according to claim 136, wherein the step of returning the response data is performed using a unicast connection between the advertiser and the requester.   In the step of returning the response data, the requester creates a new subscription reservation for the response data, executes CBR using the new subscription reservation, and pushes the response data to the requester. The communication method according to claim 136, comprising steps.   The communication method according to claim 136, wherein the step of receiving a query receives a query related to a data set.   The communication method according to claim 136, wherein the step of receiving the data advertisement includes receiving a data advertisement related to a data set.   The communication method of claim 136, wherein the step of converting the data advertisement comprises performing general mapping on the advertisement.   The communication method according to claim 136, wherein the step of converting the data advertisement comprises performing disjunctive query mapping for the advertisement.   The communication method according to claim 136, wherein the step of converting the data advertisement includes the step of defining an attribute representing a combination of an attribute and an expression operator in the data advertisement in the subscription.   The communication method according to claim 136, wherein the step of converting the query includes a step of defining an attribute in a notification representing a combination of an attribute and an expression operator in the query.   137. An publish-subscribe network that implements a query-response interaction, the router comprising a module that implements the communication method of claim 136, an issuer device, and a subscriber device. A communication method for realizing query-response interaction in an issue-subscribe network, comprising:
Receiving a query;
Converting the above query into a notification;
Receiving response data based on content-based routing performed on the data advertisement converted into the notification and subscription.   148. A communication device that implements query-response interaction in an issue-subscribe network, comprising a module that implements the communication method of claim 147.   149. The communication device of claim 148, wherein the communication device includes a subscriber device. A communication method for realizing query-response interaction in an issue-subscribe network, comprising:
Receiving a data ad;
Converting the data ad into a subscription,
Receiving a query transformed into a notification based on content-based routing performed on the subscription and notification;
Transferring the response data to the requester based on the notification.   161. A communication device that implements query-response interaction in an issue-subscribe network, comprising a module that implements the communication method of claim 150.   The communication apparatus according to claim 151, wherein the communication apparatus includes an issuer apparatus. A communication method for realizing query-response interaction in an issue-subscribe network, comprising:
Receiving a changed notification from the query;
Content-based routing (CBR) and determining whether to forward notifications to adjacent nodes;
And selectively transferring the notification based on the determination. Receiving a converted subscription from a data advertisement;
154. The communication method according to claim 153, wherein the step of executing CBR executes CBR based on the subscription.   154. A communication device that implements query-response interaction in an issue-subscribe network, comprising a module that implements the communication method of claim 153.   The communication apparatus according to claim 155, wherein the communication apparatus includes a router. A communication method for realizing persistent caching of messages provided via a publish-subscribe network, comprising:
(A) receiving a message including data via a network at a first node;
(B) adding a mark indicating time to the data;
(C) caching the data in the cache memory of the first node;
And (d) routing the message to the second node using content-based routing.   158. The communication method according to claim 157, further comprising: (e) repeating steps (a) to (d) at the second node except that data is cached in the cache memory of the second node. Determining whether a time grain period G has elapsed since step (c) of caching the data;
159. The communication method according to claim 157, further comprising a step of transferring the cached data from a cache memory to a disk when it is determined that the time grain period G has elapsed. Determining whether the last time frame period T has elapsed since step (c) of caching the data for the last block of the cached data;
158. The communication method according to claim 157, further comprising: erasing the cached data when it is determined that the continuous time frame period T has elapsed.   The communication method according to claim 160, wherein the duration time frame period T is determined based on a channel associated with data.   158. The communication method according to claim 157, wherein said first node includes an intelligent router and said receiving step (a) comprises receiving a message at said intelligent router.   The communication method according to claim 162, wherein the intelligent router includes a cache manager, and the cache manager executes steps (b) and (c).   The communication method according to claim 157, wherein the second node includes a subscriber device.   158. A router that implements persistent caching of messages delivered over an publish-subscribe network, comprising a module that implements the communication method of claim 157.   158. An publish-subscribe network that implements persistent caching of messages, comprising a node comprising a module that implements the communication method of claim 158.   158. A computer readable medium having instructions for implementing the communication method of claim 157. A communication method that enables persistent caching of messages delivered over a publish-subscribe network, comprising:
Receiving a message containing data via a publish-subscribe network;
Adding a mark indicating time to the data;
Caching the data in a cache memory;
Determining whether a time grain period G has elapsed from the step of caching the data;
If it is determined that the time grain period G has elapsed, the step of transferring the cached data from the cache memory to a disk;
Determining whether a last time frame period T has elapsed from the step of caching the data for the last block of the cached data;
And erasing the cached data when it is determined that the duration time frame period T has elapsed.   169. The communication method of claim 168, comprising determining whether a request for the cached data has been received.   169. The communication method of claim 169, wherein the request for cached data is received from a user device.   169. The communication method of claim 169, comprising the step of reading the cached data when a request for the cached data is received.   The communication method according to claim 171, wherein the step of reading the cached data includes a step of calling an upstream cache and a step of reading the cached data from the upstream cache.   173. The communication method according to claim 172, further comprising a step of executing a duplication suppression process on the cached data read from the upstream cache.   173. The communication method of claim 172, wherein the step of reading the cached data comprises the step of obtaining a list of upstream caches from an agent of a user device.   The communication method according to claim 171, further comprising the step of transferring the cached data to a backbone router for routing to a user equipment.   169. The communication method according to claim 168, further comprising the step of attaching an index to the cached data.   177. The communication method according to claim 176, wherein the step of indexing the cached data includes the step of indexing the data by ID, subject and / or time stamp.   168. The communication method according to claim 168, wherein the steps of receiving, adding time information, and caching are executed in an upstream routing node.   168. The communication method according to claim 168, wherein the receiving step is repeated.   169. A router that implements persistent caching of messages delivered over an publish-subscribe network, comprising a module that implements the communication method of claim 168.   170. A publish-subscribe network that implements persistent caching of messages, each comprising a first node comprising a module for performing the communication method of claim 168 and a second downstream side of the first node. Publish-subscribe network with nodes.   170. A computer readable medium having instructions for implementing the communication method of claim 168. A communication method that enables persistent caching of messages delivered over a publish-subscribe network, comprising:
(A) receiving a plurality of messages having data via a network at a plurality of upstream nodes;
(B) adding time information to the data;
(C) caching the data in a cache at one upstream node of the plurality of upstream nodes;
(D) routing the message to the edge node using content-based routing;
(E) A data communication method comprising the steps of repeating steps (a) to (c) at the edge node except that data is cached in the cache memory of the edge node.   184. The communication method of claim 183, comprising routing at least one of the messages from an edge node to a subscriber device using content-based routing.   185. The communication method according to claim 184, further comprising the step of receiving the cached data request from the subscriber device at the edge node. Determining the amount of time for the cached data in response to the request;
186. The communication method according to claim 185, further comprising: reading the cached data corresponding to the determined amount of time based on time information attached to the cached data.   187. The communication method of claim 186, wherein the amount of time is determined based on a request that includes a time amount of cached data to be read.   187. The communication method of claim 186, wherein the amount of time is determined based on a subscriber device disconnect time, which is a time that the subscriber device disconnects from the issue-subscribe network.   186. The communication method of claim 185, comprising determining that at least some of the requested data is stored in one or more of the plurality of upstream nodes.   189. The communication method of claim 189, comprising the step of receiving at the edge node a list of upstream nodes from the subscriber device. Sending a request for the cached data to one or more upstream nodes of the plurality of upstream nodes;
Reading the requested cached data from one or more upstream nodes of the plurality of upstream nodes;
190. The communication method according to claim 189, further comprising the step of routing the read data to the subscriber device.   191. The communication method according to claim 191, further comprising a step of executing a duplication suppression process on the read cached data.   184. A publish-subscribe network that implements persistent caching of messages, comprising a plurality of upstream nodes and edge nodes comprising modules that implement the communication method of claim 183.

Images