US20190335010A1 - Systems and methods for providing messages to multiple subscribers - Google Patents
Systems and methods for providing messages to multiple subscribers Download PDFInfo
- Publication number
- US20190335010A1 US20190335010A1 US16/506,890 US201916506890A US2019335010A1 US 20190335010 A1 US20190335010 A1 US 20190335010A1 US 201916506890 A US201916506890 A US 201916506890A US 2019335010 A1 US2019335010 A1 US 2019335010A1
- Authority
- US
- United States
- Prior art keywords
- messages
- node
- channel
- streamlet
- data
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Abandoned
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/50—Network services
- H04L67/55—Push-based network services
-
- H04L67/26—
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L47/00—Traffic control in data switching networks
- H04L47/10—Flow control; Congestion control
- H04L47/28—Flow control; Congestion control in relation to timing considerations
- H04L47/286—Time to live
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L47/00—Traffic control in data switching networks
- H04L47/70—Admission control; Resource allocation
- H04L47/72—Admission control; Resource allocation using reservation actions during connection setup
- H04L47/722—Admission control; Resource allocation using reservation actions during connection setup at the destination endpoint, e.g. reservation of terminal resources or buffer space
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L51/00—User-to-user messaging in packet-switching networks, transmitted according to store-and-forward or real-time protocols, e.g. e-mail
- H04L51/21—Monitoring or handling of messages
- H04L51/226—Delivery according to priorities
-
- H04L51/26—
-
- H04L65/4076—
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L65/00—Network arrangements, protocols or services for supporting real-time applications in data packet communication
- H04L65/60—Network streaming of media packets
- H04L65/61—Network streaming of media packets for supporting one-way streaming services, e.g. Internet radio
- H04L65/611—Network streaming of media packets for supporting one-way streaming services, e.g. Internet radio for multicast or broadcast
-
- H04L67/2842—
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/50—Network services
- H04L67/56—Provisioning of proxy services
- H04L67/568—Storing data temporarily at an intermediate stage, e.g. caching
-
- H04L51/14—
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L51/00—User-to-user messaging in packet-switching networks, transmitted according to store-and-forward or real-time protocols, e.g. e-mail
- H04L51/21—Monitoring or handling of messages
- H04L51/214—Monitoring or handling of messages using selective forwarding
Definitions
- This specification relates to a data communication system and, in particular, a system that implements real-time, scalable publish-subscribe messaging.
- the publish-subscribe pattern is a data communication messaging arrangement implemented by software systems where so-called publishers publish messages to topics and so-called subscribers receive the messages pertaining to particular topics to which they are subscribed. There can be one or more publishers per topic and publishers generally have no knowledge of what subscribers, if any, will receive the published messages. Some PubSub systems do not cache messages or have small caches meaning that subscribers may not receive messages that were published before the time of subscription to a particular topic. PubSub systems can be susceptible to performance instability during, surges of message publications or as the number of subscribers to a particular topic increases.
- the systems and methods described herein utilize a shared structure (i.e., a “fast buffer”) to store message data that can be read by multiple subscribers simultaneously.
- a shared structure i.e., a “fast buffer”
- the systems and methods utilize a “lock-free” or non-blocking approach. Subscribers are able to grab many pieces of data without blocking or figuring out where to read from and up to which point in the fast buffer.
- the systems and methods provide many benefits compared to prior approaches. For example, publishers are not limited by subscribers or associated subscription logic and have a stable publishing speed, regardless of how many subscribers exist, if any. This property of the system can be achieved, for example, by making publishing and subscribing components and/or processes independent from each other. In some instances, the publishing and subscribing components can share information through lock-free memory writes and reads, for example, using the fast buffer approach described herein. Alternatively or additionally, publishing and subscribing components and/or processes can be configured to utilize different physical cores, such that, for example, publishing and subscribing components interact solely through a memory controller and bus, which is typically fast and/or does not present a bottleneck.
- the fast buffer has or utilizes loosely coupled publishing and subscribing loops. This allows a publishing speed to be different from a subscribing speed, if desired, and/or for each speed to be chosen or prioritized, for example, according system load.
- the publishing loop can be prioritized over the subscribing loop. To achieve such prioritization, the publishing loop can be made to spin as if there were no subscribers (e.g., the publishing loop can be processed faster than the subscribing loop is processed).
- Such prioritization can be achieved, for example, by creating two worker pools (e.g., one pool for publishers and one pool for subscribers) and using an operating system kernel scheduler prioritization mechanism to set different priorities to threads in the two pools.
- Other approaches for prioritizing or deprioritzing publishers and subscribers are contemplated.
- a further advantage of the systems and methods is that subscribers can utilize data aggregation and can define how much data to aggregate.
- use of data aggregation and/or aggregation size optimization can help achieve better performance characteristics and/or reduce memory usage.
- the messages can be grouped into blocks or hatches (e.g., a comma-separated list of messages in the fast buffer).
- the systems and methods can decide how many messages to aggregate for the next block.
- the sizes of packets exchanged in a connection can depend on connection properties (e.g., bandwidth and/or latency) and/or on a state of a TCP connection. In some instances, this allows the process of collecting messages for the connection to aggregate more messages in a block to produce a bigger packet, thereby better utilizing network capacity and providing higher throughput with less overhead.
- Another benefit of the systems and methods is that, depending on message activity, subscribers can change the frequency of data reading by finding an optimal balance between performance and latency. For example, using the notifier approach described herein, channels with high message activity can be actively monitored for new messages, while channels with low message activity can receive less active monitoring. Subscribers are preferably isolated and do not affect one other. If one subscriber is slower than a message stream, other subscribers are generally unaffected by the slow subscriber and can keep up with the message stream.
- a further benefit of the systems and methods is that modularity is increased, which allows system components to be implemented independently. Interaction is preferably done through memory, so subscribers can be placed on separate individual processes to access the fist buffer.
- multiple logical parts of a system can be coupled tightly to reduce interface overhead (e.g., preparing data to be exported and parsing on input) and/or to make changes more granular and precise.
- interface overhead e.g., preparing data to be exported and parsing on input
- the drawback of such approaches is reduced isolation, reduced reliability and, in general, less efficiency (e.g., due to load of one component influencing other components).
- the opposite approach, used in examples of the systems and methods described herein, is to isolate components from each other and to achieve increased reliability and efficiency and better management and prioritization of system components.
- isolation is more effective when the cost of overhead is low or at least acceptable, in the systems and methods, for example, the cost of overhead is generally low because processes can communicate by writing to and reading from the fast buffer.
- the systems and methods can benefit from such modularity with little or no added cost.
- when shared memory at the operating system level is used publishers and subscribers can be implemented in different processes and/or can start using implementations written in different languages.
- interaction through use of the fast buffer limits communication to only two operations: writing to the memory and reading from the memory. The use of messages, signals, locks, and/or synchronization primitives can be avoided. As discussed herein, test results indicate that the systems and methods greatly improve message transfer rates from publishers to subscribers.
- one aspect of the subject matter described in this specification can be embodied in computer-implemented methods that include the actions of: retrieving messages of a particular channel of a plurality of channels from respective storage buffers, wherein each storage buffer comprises a respective first time-to-live.
- the method may also include storing the retrieved messages of the particular channel in a delivery buffer, wherein the delivery buffer comprises one or more data blocks, and wherein each data block comprises a respective second time-to-live.
- the method may also include transmitting, by one or more computer processors, the messages stored in the delivery buffer to subscribers of the particular channel.
- Other embodiments of this aspect include corresponding systems, apparatus, storage devices, and computer programs.
- FIG. 1A illustrates an example system that supports the PubSub communication pattern.
- FIG. 1B illustrates functional layers of software on an example client device.
- FIG. 2 is a diagram of an example messaging system.
- FIG. 3A is a data flow diagram of an example method for writing data to a streamlet.
- FIG. 3B is a data flow diagram of an example method for reading data from a streamlet.
- FIG. 4A is a data flow diagram of an example method for publishing messages to a channel of a messaging system.
- FIG. 4B is a data flow diagram of an example method for subscribing to a channel of a messaging system.
- FIG. 4C is an example data structure for storing messages of a channel of a messaging system.
- FIG. 5 is a schematic diagram of an MX Node retrieving messages from a Q node and providing the retrieved messages to a subscriber of a messaging system.
- FIG. 6 shows example data blocks of a fast buffer in an MX node of a messaging system at different time instances.
- FIG. 7 is a flowchart of an example method for sending message data to a plurality of subscribers.
- FIG. 1A illustrates an example system 100 that supports the PubSub communication pattern.
- Publishers e.g., Publishers 1 -N
- a message can include any type of information including one or more of the following: text, image content, sound content, multimedia content, video content, binary data, and so on. Other types of message data are possible.
- Subscribers e.g., Subscribers 1 -N
- a client can be both a publisher and a subscriber.
- a PubSub system can be categorized as follows:
- M:N Many to Many
- M:N Many to Many
- Example use cases are map applications.
- channel names can be qualified by a name space.
- a name space includes one or more channel names. Different name spaces can have the same channel names without causing ambiguity.
- the name space name can be a prefix of a channel name where the name space and channel name are separated by a dot or other suitable separator.
- name spaces can be used when specifying channel authorization settings.
- the system 100 may have app1.foo and app1.system.notifications channels where “app1” is the name of the name space. The system can allow clients to subscribe and publish to the app1.foo channel. However, clients can subscribe, but not publish, to the app1.system.notiftcations channel.
- FIG. 1B illustrates functional layers of software on an example client device.
- a client device e.g., client 102
- the application layer 104 includes the end-user application(s) that will integrate with the system 100 .
- the messaging layer 106 is a programmatic interface for the application layer 104 to utilize services of the system 100 such as channel subscription, message publication, message retrieval, user authentication, and user authorization.
- the messages passed to and from the messaging layer 106 are encoded as JavaScript Object Notation (JSON) objects. Other message encoding schemes are possible.
- JSON JavaScript Object Notation
- the operating system layer 108 includes the operating system software on the client 102 .
- messages can be sent and received to/from the system 100 using persistent or non-persistent connections. Persistent connections can be created using, for example, network sockets.
- a transport protocol such as TCP/IP layer 112 implements the Transport Control Protocol Internet Protocol communication with the system 100 that can be used by the messaging layer 106 to send messages over connections to the system 100 .
- Other communication protocols are possible including, for example, User Datagram Protocol (UDP).
- UDP User Datagram Protocol
- an optional Transport Layer Security (TLS) layer 110 can be employed to ensure the confidentiality of the messages.
- TLS Transport Layer Security
- FIG. 2 is a diagram of an example messaging system 100 .
- the messaging system 100 provides functionality for implementing PubSub communication patterns.
- the messaging system 100 includes software components and storage that can be deployed at one or more data centers 122 in one or more geographic locations, for example.
- the messaging system 100 includes multiplexer (MX) nodes 202 , 204 and 206 , queue (Q) nodes 208 , 210 and 212 , one or more channel manager nodes (e,g., channel managers 214 , 215 ), and optionally one or more cache (C) nodes 220 and 222 .
- Each node can execute in a virtual machine or on a physical machine (e.g., a data processing apparatus).
- Each MX node serves as a termination point for one or more publisher and/or subscriber connections through the external network 216 .
- the internal communication among MX nodes, Q nodes, C nodes, and the channel managers, is conducted over an internal network 218 .
- MX node 204 can be the terminus of a subscriber connection from client 102 .
- Each Q node buffers channel data for consumption by the MX nodes.
- An ordered sequence of messages published to a channel is a logical channel stream. For example, if three clients publish messages to a given channel, the combined messages published by the clients include a channel stream. Messages can be ordered in a channel stream.
- the messages may be ordered by time of publication by the client, by time of receipt by an M X node, or by time of receipt by a Q node.
- Other ways for ordering messages in a channel stream are possible. In the case where more than one message would be assigned to the same position in the order, one of the messages can be chosen (e.g., randomly) to have a later sequence in the order.
- Each channel manager node is responsible for managing Q node load by splitting channel streams into streamlets, as will be discussed in further detail below.
- the optional C nodes provide caching and load removal from the Q nodes.
- one or more client devices establish respective persistent connections (e.g., TCP connections) to an MX node (e.g., MX nodes 202 , 204 and/or 206 ).
- the MX node serves as a termination point for these connections.
- external messages e.g., between respective client devices and the MX node
- the MX node terminates the external protocol and translates the external messages to internal communication, and vice versa.
- the MX nodes 202 , 204 and 206 publish and subscribe to streamlets on behalf of clients. In this way, an MX node can multiplex and merge requests of client devices subscribing for or publishing to the same channel, thus representing multiple client devices as one, instead of one by one.
- a Q node e.g., nodes 208 , 210 and/or 212
- a streamlet is a data buffer for a portion of a channel stream.
- a streamlet will close to writing when its storage is full.
- a streamlet will close to reading and writing and be de-allocated when its time-to-live (TTL) has expired.
- TTL time-to-live
- a streamlet can have a maximum size of 1 MB and a TTL, of three minutes.
- Different channels can have streamlets limited by different sizes and/or by different TTLs. For example, streamlets in one channel can exist for up to three minutes, while streamlets in another channel can exist for up to 10 minutes.
- a streamlet corresponds to a computing process running on a Q node.
- the computing process can be terminated after the streamlet's TTL has expired, thus freeing up computing resources (for the streamlet) back to the Q node, for example.
- an MX node When receiving a publish request from client 102 , an MX node (e.g., MX node 204 ) makes a request to a channel manager (e.g., channel manager 214 ) to grant access to a streamlet to write the message being published. However, if the MX node has already been granted write access to a streamlet for the channel (and the channel has not been closed to writing), the MX node can write the message to that streamlet without having to request a grant to access the streamlet. Once a message is written to a streamlet for a channel, the message can be read by MX nodes and provided to subscribers of that channel.
- a channel manager e.g., channel manager 214
- an MX node when receiving a channel subscription request from a client device, an MX node makes a request to a channel manager to grant access to a streamlet for the channel from which messages are read. If the MX node has already been granted read access to a streamlet for the channel (and the channel's TTL has not been closed to reading) the MX node can read messages from the streamlet without having to request a grant to access the streamlet. The read messages can then be forwarded to client devices that have subscribed to the channel.
- messages read from streamlets are cached by MX nodes so that MX nodes can reduce the number of times messages are read from the streamlets.
- an MX node can request a grant from the channel manager that allows the MX node to store a block of data into a streamlet on a particular Q node that stores streamlets of a particular channel.
- Example streamlet grant request and grant data structures are as follows:
- StreamletGrantRequest ⁇ “channel”: string( ) “mode”; “read”
- StreamletGrantResponse ⁇ “streamlet-id”: “abcdef82734987”, “limit-size”: 2000000, # 2 megabytes max “limit-msgs”: 5000, # 5 thousand messages max “limit-life”: 4000, # the grant is valid for 4 seconds “q-node”: string( ) “position”: 0 ⁇
- the StreamletGrantRequest data structure stores the name of the stream channel and a mode indicating whether the MX node intends on reading from or writing to the streamlet.
- the MX node sends the StreamletGrantRequest to a channel manager node.
- the channel manager node in response, sends the MX node a StreamletGrantResponse data structure.
- the StreamletGrantResponse contains an identifier of the streamlet (streamlet-id), the maximum size of the streamlet (limit-sizes), the maximum number of messages that the streamlet can store (limit-msgs), the TTL (limit-life), and an identifier of a Q node (q-node) on which the streamlet resides.
- the StreamletGrantRequest and StreamletGrantResponse can also have a position field that points to a position in a streamlet (or a position in a channel) for reading from the streamlet.
- the MX node can request a new grant from the channel manager to read from or write to a streamlet.
- the new grant will reference a different streamlet and will refer to the same or a different Q node depending on where the new streamlet resides.
- FIG. 3A is a data flow diagram of an example method 300 for writing data to a streamlet in various embodiments.
- an MX node's e.g., MX node 202
- the MX node 202 establishes a Transmission Control Protocol (TCP) connection with the Q node (e.g., Q node 208 ) identified in the grant response received from the channel manager ( 302 ).
- TCP Transmission Control Protocol
- Q node e.g., Q node 208
- a streamlet can be written concurrently by multiple write grants (e.g., for messages published by, multiple publishers).
- Other types of connection protocols between the MX node 202 and the Q node 208 are possible.
- the MX node 202 sends a prepare-publish message with an identifier of a streamlet that the MX node 202 wants to write to the Q node 208 ( 304 ).
- the streamlet identifier and Q node identifier can be provided by the channel manager in the write grant as described earlier.
- the Q node 202 provides the message to a handler 301 (e.g., a computing process running on the Q node 208 ) for the identified streamlet ( 306 ).
- the handler 301 can send an acknowledgment to the MX node 202 ( 308 ).
- the MX node 202 After receiving the acknowledgement, the MX node 202 starts writing (publishing) messages (e.g., 310 , 312 , 314 , and 318 ) to the handler 301 , which stores the received data in the identified streamlet.
- the handler 301 can also send acknowledgements ( 316 , 320 ) to the MX node 202 for the received data.
- acknowledgements can be piggy-backed or cumulative.
- the handler 301 can send an acknowledgement to the MX node 202 for every predetermined amount of data received (e.g., for every 100 messages received) or for every predetermined time period (e.g., for every one millisecond).
- Other acknowledgement scheduling algorithms such as Nagle's algorithm, can be used.
- the handler 301 sends a Negative-Acknowledgement (NAK) message ( 330 ) indicating a problem, following by an EOF (end-of-file) message ( 332 ). In this way, the handler 301 closes the association with the MX node 202 for the publish grant.
- the MX node 202 can request a write grant for another streamlet from a channel manager if the MX node 202 has additional messages to store.
- FIG. 3B is a data flow diagram of an example method 350 for reading data from a streamlet in various embodiments.
- an MX node e.g., MX node 204
- a channel manager e.g., channel manager 214
- the channel manager returns a read grant to the MX node 204 including an identifier of a streamlet containing the particular message, a position in the streamlet corresponding to the particular message, and an identifier of a Q node (e.g., Q node 208 ) containing the particular streamlet.
- the MIX node 204 then establishes a TCP connection with the Q node 208 ( 352 ).
- Other types of connection protocols between the MX node 204 and the Q node 208 are possible.
- the MX node 204 then sends a subscribe message ( 354 ) to the Q node 208 with the identifier of the streamlet in the Q node and the position in the streamlet from which the MX node 204 wants to read ( 356 ).
- the Q node 208 provides the subscribe message to a handler 351 of the streamlet ( 356 ).
- the handier 351 can send an acknowledgement to the MX node 204 ( 358 ).
- the handler 351 sends messages ( 360 , 364 , 366 ), starting at the position in the streamlet, to the MX node 204 .
- the handler 351 can send all of the messages in the streamlet to the MX node 204 .
- the handler 351 can send a notification of the last message to the MX node 204 .
- the MX node 204 can send to the channel manager another request for another streamlet containing a next message in the particular channel.
- the handler 351 can send an unsubscribe message ( 390 ), followed by an EOF message ( 392 ), to close the association with the MX node 204 for the read grant.
- the MX node 204 can close the association with the handler 351 when the MX node 204 moves to another streamlet for messages in the particular channel (e.g., as instructed by the channel manager).
- the MX node 204 can also close the association with the handler 351 if the MX node 204 receives an unsubscribe message from a corresponding client device.
- a streamlet can be written into and read from at the same time instance. For example, there can be a valid read grant and a valid write grant at the same time instance.
- a streamlet can be read concurrently by multiple read grants (e.g., for channels subscribed to by multiple publisher clients).
- the handler of the streamlet can order messages from concurrent write grants based on, for example, time-of-arrival, and store the messages based on the order. In this way, messages published to a channel from multiple publisher clients can be serialized and stored in a streamlet of the channel.
- one or more C nodes can offload data transfers from one or more Q nodes. For example, if there are multiple: MX nodes requesting streamlets from Q nodes for a particular channel, the streamlets can be offloaded and cached in one or more C nodes.
- the MX nodes e.g., as instructed by read grants from a channel manager
- a channel manager (e.g., channel manager 214 ) splits the channel stream into fixed-sized streamlets that each reside on a respective Q node.
- each Q node stores a portion (one or more streamlets) of the channel stream.
- a streamlet can be stored in, for example, registers and for dynamic memory elements associated with a computing process on a Q node, thus avoiding the need to access persistent, slower storage devices such as hard disks. This results in faster message access.
- the channel manager can also balance loads among Q nodes in the messaging system 100 by monitoring respective workloads of the Q nodes and allocating streamlets in a way that avoids overloading any one Q node.
- a channel manager maintains a list identifying each active streamlet, the respective node on which the streamlet resides, an identification of the position of the first message in the streamlet, and whether the streamlet is closed for writing.
- Q nodes notify the Channel manager and any MX nodes that are publishing to a streamlet that the streamlet is closed due to being full or when the streamlet's TTL has expired.
- the streamlet remains on the channel manager's list of active streamlets until the streamlet's TTL has expired so that MX nodes can continue to retrieve messages from the streamlet.
- the channel manager allocates a new streamlet on one of the Q nodes and returns the identity of the streamlet and the Q node in the StreamletGrantResponse to the MX node. Otherwise, the channel manager returns the identity of the currently open for writing streamlet and corresponding Q node in the StreamletGrantResponse to the MX node.
- MX nodes can publish messages to the streamlet until the streamlet is full or the streamlet's TTL has expired, after which a new streamlet can be allocated by the channel manager.
- the channel manager allocates a new streamlet on one of the Q nodes and returns the identity of the streamlet and the Q node in the StreamletGrantResponse to the MX node. Otherwise, the channel manager returns the identity of the streamlet and Q node that contains the position from which the MX node wishes to read to the MX node.
- the Q node can then begin sending messages to the MX node from the streamlet beginning at the specified position until there are no more messages in the streamlet to send.
- MX nodes that have subscribed to that streamlet will receive the new message. If a streamlet's TTL has expired, the handier process 351 sends an Eat: message ( 392 ) to any MX nodes that are subscribed to the streamlet.
- the messaging system 100 can include multiple channel managers (e.g., channel managers 214 , 215 ). Multiple channel managers provide resiliency and prevent single point of failure. For instance, one channel manager can replicate lists of streamlets and current grants it maintains to another “slave” channel manager. As for another example, multiple channel managers can coordinate operations between them using distributed consensus protocols, such as, for example, Paxos or Raft protocols.
- distributed consensus protocols such as, for example, Paxos or Raft protocols.
- FIG. 4A is a data flow diagram of an example method 400 for publishing messages to a channel of a messaging system.
- publishers e.g., publishers 402 , 404 , 406
- publishers 402 respectively establish connections 411 and send publish requests to the MX node 202 .
- Publishers 404 respectively establish connections 413 and send publish requests to the MX node 206 .
- Publishers 406 respectively establish connections 415 and send publish requests to the MX node 204 .
- the MX nodes can communicate ( 417 ) with a channel manager (e.g., channel manager 214 ) and one or more Q nodes (e.g., nodes 212 and 208 ) in the messaging system 100 via the internal network 218 .
- a channel manager e.g., channel manager 214
- Q nodes e.g., nodes 212 and 208
- each publish request (e.g., in JSON key/value pairs) from a publisher to an MX node includes a channel name and a message.
- the MX node e.g., MX node 202
- the MX node can confirm the assigned channel with the channel manager 214 . If the channel (specified in the subscribe request) does not yet exist in the messaging system 100 , the channel manager can create and maintain a new channel in the messaging system 100 . For instance, the channel manager can maintain a new channel by maintaining a list identifying each active streamlet of the channel's stream, the respective Q node on which the streamlet resides, and identification of the positions of the first and last messages in the streamlet as described earlier.
- the MX node can store the messages in one or more buffers or streamlets in the messaging system 100 .
- the MX node 202 receives, from the publishers 402 , requests to publish messages M 11 , M 12 , M 13 , and M 14 to a channel foo.
- the MX node 206 receives, from the publishers 404 , requests to publish messages M 78 and M 79 to the channel foe.
- the MX node 204 receives, from the publishers 406 , requests to publish messages M 26 , M 27 , M 28 , M 29 , M 30 , and M 31 to the channel foo.
- the MX nodes can identify one or more streamlets for storing messages for the channel foo.
- each MX node can request a write grant from the channel manager 214 that allows the MX node to store the messages in a streamlet of the channel foo.
- the MX node 202 receives a grant from the channel manager 214 to write messages M 11 , M 12 , M 13 , and M 14 to a streamlet 4101 on the Q node 212 .
- the MX node 206 receives a grant from the channel manager 214 to write messages M 78 and M 79 to the streamlet 4101 .
- the streamlet 4101 is the last streamlet of a sequence of streamlets of the channel stream 430 storing messages of the channel foo.
- the streamlet 4101 has messages 421 of the channel foo that were previously stored in the streamlet 4101 , but is still open (e.g., the streamlet 4101 still has space for storing more messages and the streamlet's TTL has not expired.)
- the MX node 202 can arrange the messages for the channel foo based on the respective time that each of the messages 422 was received by the MX node 202 (e.g., M 11 , M 13 , M 14 , M 12 ) and store the received messages as arranged in the streamlet 4101 . That is, the MX node 202 receives M 11 first, followed by M 13 , M 14 , and M 12 .
- the MX node 206 can arrange the messages for the channel foo based on their respective time that each of the messages 423 was received by the MX node 206 (e.g., M 78 , M 79 ) and store the received messages 423 as arranged in the streamlet 4101 .
- Other arrangements or ordering of the messages for the channel are possible.
- the MX node 202 (or MX node 206 ) can store the received messages using the method for writing data to a streamlet described earlier in reference to FIG. 3A , for example.
- the MX node 202 (or MX node 206 ) can buffer (e.g., in a local data buffer) the received messages for the channel foo and store the received messages in a streamlet for the channel foo (e.g., streamlet 4101 ) when the buffered messages reach a predetermined number or size (e.g., 100 messages) or when a predetermined time (e.g., 50 milliseconds) has elapsed.
- the MX node 202 can store in the streamlet 100 messages at a time or in 50 millisecond increments.
- Other acknowledgement scheduling algorithms such as Nagle's algorithm, can be used.
- the Q node 212 (e.g., a handler) stores the messages of the channel foo in the streamlet 4101 in the order as arranged by the MX node 202 and MX node 206 .
- the Q node 212 stores the messages of the channel foo in the streamlet 4101 in the order the Q node 212 receives the messages. For instance, assume that the Q node 212 receives message M 78 (from the MX node 206 ) first, followed by messages M 11 and M 13 (from the MX node 202 ), M 79 (from the MX node 206 ), and M 14 and M 12 (from the MX node 202 ).
- the Q node 212 stores in the streamlet 4101 the messages in the order as received (e.g., M 78 , M 11 , M 13 , M 79 , M 14 , and M 12 ) immediately after the messages 421 that are already stored in the streamlet 4101 .
- messages published to the channel too from multiple publishers e.g., MX nodes 402 , 404
- the NIX node 204 requests a grant from the channel manager 214 to write to the channel foo.
- the channel manager 214 provides the MX node 204 a grant to write messages to the streamlet 4101 , as the streamlet 4101 is still open for writing.
- the MX node 204 arranges the messages for the channel foo based on the respective time that each message 424 was received by the MX node 204 (e.g., M 26 , M 27 , M 31 , M 29 , M 30 , M 28 ) and stores the messages as arranged for the channel foo.
- the Q node 212 sends to the MX node 204 a NAK message, following by an EOF message, to close the association with the MX node 204 for the write grant, as described earlier in reference to FIG. 3A .
- the MX node 204 then requests another write grant from the channel manager 214 for additional messages (e.g., M 27 , M 31 , and so on) for the channel too.
- the channel manager 214 can monitor available Q nodes in the messaging system 100 for the Q nodes respective wort loads (e.g., how many streamlets are residing in each Q node). The channel manager 214 can allocate a streamlet for the write request from the MX node 204 such that overloading (e.g., too many streamlets or too many read or write grants) can be avoided for any given Q node. For example, the channel manager 214 can identify a least loaded Q node in the messaging system 100 and allocate a new streamlet on the least loaded Q node for write request's from the MX node 204 . In the example of FIG.
- the channel manager 214 allocates a new streamlet 4102 on the Q node 208 and provides a write grant to the MX node 204 to write messages for the channel foo to the streamlet 4102 .
- the Q node 208 stores in the streamlet 4102 the messages from the MX node 204 in an order as arranged by the MX node 204 : M 27 , M 31 , M 29 , M 30 , and M 28 (assuming that there is no other concurrent write grants for the streamlet 4102 at the moment).
- the channel manager 214 allocates a new streamlet (e.g., streamlet 4102 ) for a request for a grant from an MX node (e.g., MX node 204 ) to write to a channel (e.g., too), the channel manager 214 assigns to the streamlet its TTL, which will expire after TTLs of other streamlets that are already in the channel's stream. For instance, the channel manager 214 can assign to each streamlet of the channel foo's channel stream a TTL of 3 minutes when allocating the streamlet. That is, each streamlet will expire 3 minutes after it is allocated (created) by the channel manager 214 .
- a new streamlet e.g., streamlet 4102
- MX node e.g., MX node 204
- the channel foo's channel stream includes streamlets that each expire sequentially after the previous streamlet expires. For example, as shown in an example channel stream 430 of the channel foo in FIG. 4A , streamlet 4098 and streamlets before 4098 (e.g., streamlet 4097 ) have expired (as indicated by the dotted-lined gray-out boxes). Messages stored in these expired streamlets are not available for reading for subscribers of the channel foo. Streamlets 4099 , 4100 , 4101 , and 4102 are still active (not expired).
- the streamlets 4099 , 4100 , and 4101 are closed for writing, but still are available for reading.
- the streamlet 4102 is available for reading and writing, at the moment when the message M 28 was stored in the streamlet 4102 .
- the streamlet 4099 will expire, following by the streamlets 4100 , 4101 , and so on.
- FIG. 4B is a data flow diagram of an example method 450 for subscribing to a channel of a messaging system.
- a subscriber 480 establishes a connection 462 with an MX node 461 of the messaging system 100 .
- Subscriber 482 establishes a connection 463 with the MX node 461 .
- Subscriber 485 establishes a connection 467 with an MX node 468 of the messaging system 100 .
- the MX nodes 461 and 468 can respectively communicate 464 with the channel manager 214 and one or more Q nodes in the messaging system 100 via the internal network 218 .
- a subscriber e.g., subscriber 480
- MX node e.g., MX node 461
- the request (e.g., in JSON key/value pairs) can include a channel name, such as, for example, “foo.”
- the MX node 461 can send a request to the channel manager 214 for a read grant for a streamlet in the channel foo's channel stream.
- the channel foo's channel stream 431 includes active streamlets 4102 , 4103 , and 4104 , as shown in FIG. 4B .
- the streamlets 4102 and 4103 each are full.
- the streamlet 4104 stores messages of the channel foo, including the last message stored at a position 47731 .
- Streamlets 4101 and streamlets before 4101 are invalid, as their respective TTLs have expired.
- the messages M 78 , M 11 , M 13 , M 79 , M 14 , M 12 , and M 26 stored in the streamlet 4101 , described earlier in reference to FIG.
- each streamlet in the channel foo's channel stream has a TTL of 3 minutes, thus only messages (as stored in streamlets of the channel foo) that are published to the channel foo (i.e., stored into the channel's streamlets) no earlier than 3 minutes from the current time can be available for subscribers of the channel foo.
- the MX node 461 can request a read grant for all available messages in the channel foo, for example, when the subscriber 480 is a new subscriber to the channel foo. Based on the request, the channel manager 214 provides the MX node 461 a read grant to the streamlet 4102 (on the Q node 208 ) that is the earliest streamlet in the active streamlets of the channel foo (e.g., the first in the sequence of the active streamlets). The MX node 461 can retrieve messages in the streamlet 4102 from the Q node 208 , using the method for reading data from a streamlet described earlier in reference to FIG. 3B , for example. Note that the messages retrieved from the streamlet 4102 maintain the same order as stored in the streamlet 4102 .
- the Q node 208 when providing messages stored in the streamlet 4102 to the MX node 461 , the Q node 208 can buffer (e.g., in a local data buffer) the messages and send the messages to the MX node 461 when the buffer messages reach a predetermined number or size (e.g., 200 messages or a predetermined time (e.g., 50 milliseconds) has elapsed. For instance, the Q node 208 can send the channel foo's messages (from the streamlet 4102 ) to the MX node 461 200 messages at a time or in 50 millisecond increments. Other acknowledgement scheduling algorithms, such as Nile's algorithm, can be used.
- the MX node 461 can send an acknowledgement to the Q node 208 , and send to the channel manager 214 another request (e.g., for a read grant) for the next streamlet in the channel stream of the channel foo.
- the channel manager 214 provides the MX node 461 a read grant to the streamlet 4103 (on Q node 472 ) that logically follows the streamlet 4102 in the sequence of active streamlets of the channel foo.
- the MX node 461 can retrieve messages stored in the streamlet 4103 using the method for reading data from a streamlet described earlier in reference to FIG. 3B , until MX node 601 retrieves the last message stored in the streamlet 4103 .
- the MX node 461 can send to the channel manager 214 yet another request for a read grant for messages in the next streamlet 4104 (on Q node 474 ). After receiving the read grant, the MX node 461 retrieves message of the channel foo stored in the streamlet 4104 , until the last message at the position 47731 is retrieved by MX node 461 . Similarly, the MX node 468 can retrieve messages from the streamlets 4102 , 4103 , and 4104 (as shown with dotted arrows in FIG. 4B ), and provide the messages to the subscriber 485 .
- the MX node 461 can send the retrieved messages of the channel foo to the subscriber 480 (via the connection 462 ) while receiving the messages from the Q node 208 , 472 , or 474 .
- the MX node 461 can store the retrieved messages in a local buffer. In this way, the retrieved messages can be provided to another subscriber (e.g., subscriber 482 ) when the other subscriber subscribes to the channel too and requests the channel's messages.
- the MX node 461 can remove messages stored in the local buffer that each has a time of publication that has exceeded a predetermined time period. For instance, the MX node 461 can remove messages stored in the local buffer with respective times of publication exceeding 3 minutes.
- the predetermined time period for keeping messages in the local buffer on M X node 461 can be the same as or similar to the time-to-live duration of a streamlet in the channel foo's channel stream, since at a given moment, messages retrieved from the charmers stream do not include the messages i r streamlets having respective time-to-lives that have already expired.
- the messages retrieved from the channel stream 431 and sent to the subscriber 480 (by the MX node 461 ) are arranged in the same order as the messages were stored in the channel stream, although other arrangements or ordering of the messages are possible.
- messages published to the channel foo are serialized and stored in the streamlet 4102 in a particular order (e.g., M 27 , M 31 , M 29 , M 30 , and so on), then stored subsequently in the streamlet 4103 and the streamlet 4104 .
- the MX node 461 retrieves messages from the channel stream 431 and provides the retrieved messages to the subscriber 480 in the same order as the messages are stored in the channel stream: M 27 , M 31 , M 29 , M 30 , and so on, followed by ordered messages in the streamlet 4103 , and followed by ordered messages in the streamlet 4104 .
- the MX node 461 can request a read grant for messages stored in the channel stream 431 starting from a message at particular position (e.g., position 47202 .)
- the position 47202 can correspond to an earlier time instance (e.g., 10 seconds before the current time) when the subscriber 480 was last subscribing to the channel foo (e.g., via a connection to the MX node 461 or another MX node of the messaging system 100 ).
- the MX node 461 can send a request to the channel manager 214 for a read grant for messages starting at the position 47202 .
- the channel manner 214 Based on the request, the channel manner 214 provides the MX node 461 a read grant to the streamlet 4104 (on the Q node 474 ) and a position on the streamlet 4104 that corresponds to the channel stream position 47202 .
- the MX node 461 can retrieve messages in the streamlet 4104 starting from the provided position, and send the retrieved messages to the subscriber 480 .
- messages published to the channel foo are serialized and stored in the channel's streamlets in a particular order.
- the channel manager 214 maintains the ordered sequence of streamlets as they are created throughout their respective time-to-lives.
- Messages retrieved from the streamlets by an MX node e.g., MX node 461 and/or MX node 468
- MX node e.g., MX node 461 and/or MX node 468
- a subscriber can be, in some implementations, in the same order as the messages are stored in the ordered sequence of streamlets.
- messages sent to different subscribers e.g., subscriber 480 , subscriber 482 , or subscriber 485
- a streamlet stores messages in a set of blocks of messages.
- Each block stores a number of messages. For instance, a block can store two hundred kilobytes of messages.
- Each block has its own time-to-live, which can be shorter than the time-to-live of the streamlet holding the block. Once a block's TTL has expired, the block can be discarded from the streamlet holding the block, as described in more detail below in reference to FIG. 4C .
- FIG. 4C is an example data structure 490 for storing messages of a channel of a messaging system.
- the channel foo's channel stream 432 includes active streamlets 4104 and 4105 .
- Streamlet 4103 and streamlets before streamlet 4103 are invalid, as their respective TTLs have expired.
- the streamlet 4104 has reach maximum capacity (e.g., as determined by a corresponding write grant) and is closed for additional message writes.
- the streamlet 4104 is available for message reads.
- the streamlet 4105 is open and is available for message writes and reads.
- the streamlet 4104 (e.g., a computing process running on the Q node 474 shown in FIG. 4B ) currently holds two blocks of messages.
- Block 494 bolds messages from channel positions 47301 to 47850 .
- Block 495 holds messages from channel positions 47851 to 48000 .
- the streamlet 4105 (e.g., a computing process running on another Q node in the messaging system 100 ) currently holds two blocks of messages.
- Block 496 holds messages from channel positions 48001 to 48200 .
- Block 497 holds messages starting from channel position 48201 , and still accepts additional messages of the channel foo.
- block 492 was created to store messages from channel positions 47010 to 47100 . Later on, after the block 492 had reached its capacity, another block 493 was created to store messages (e.g., from channel positions 47111 to 47300 .) Blocks 494 and 495 were subsequently created to store additional messages. Afterwards, the streamlet 4104 was closed for additional message writes, and the streamlet 4105 was created with additional blocks for storing additional messages of the channel foo.
- the respective TTL's of blocks 492 and 493 have expired.
- the messages stored in these two blocks are no longer available for reading by subscribers of the channel foo.
- the streamlet 4104 can discard these two expired blocks.
- the streamlet 4104 can de-allocate the memory space for the blocks 492 and 493 .
- the blocks 494 or 495 could become expired and be discarded by the streamlet 4104 , before the streamlet 4104 itself becomes invalid.
- streamlet 4104 itself could become invalid before the blocks 494 or 495 become expired.
- a streamlet can hold one or more blocks of messages, or contain no block of messages, depending on respective TTLs of the streamlet and blocks.
- a streamlet, or a computing process running on a (node in the messaging system 100 can create a block for storing messages of a channel by allocating a certain size of memory space from the Q node.
- the streamlet can receive, from an MX node in the messaging system 100 , one message at a time and store the received message in the block.
- the MX node can assemble (i.e., buffer) a group of messages and send the group of messages to the Q node.
- the streamlet can allocate a block of memory space (from the Q node) and store the group of messages in the block.
- the MX node can also perform compression on the group of messages, e.g., by removing a common header from each message or performing other suitable compression techniques.
- an MX node can retrieve messages from Q nodes, store (cache) the retrieved messages in a local buffer, and provide the retrieved messages to multiple subscribers from the local buffer.
- the MX node can support a large number (e.g., thousands or more) of subscribers (e.g., of a particular channel), without creating a network bottleneck at Q nodes where messages are stored (in streamlets).
- the messaging system 100 can further scale up the number of subscribers by adding additional MX nodes that each can cache messages and support additional subscribers.
- a sender process running on the MX node can send copies of the cached message to destination processes (running on the MX node) that each corresponds to a connection to one of the subscribers.
- destination processes running on the MX node
- copies of the cached message e.g., created by the sender process for different destination processes
- blocking for synchronization between the destination processes can cause computing resource contention on the MX node.
- a delivery buffer i.e., a “fast buffer” in an MX Node that can be used to cache messages retrieved from a Q node.
- the fast buffer includes data blocks for storing the retrieved messages.
- each data block has a respective TTL that is longer than a read operation time associated with reading data from the fast buffer. This may prevent expiration of the data block while data is being read from the buffer and/or sent to subscribers.
- the TTL of a data block is longer than a corresponding streamlet's TTL.
- a process running on the MX node) for a subscriber can read messages stored in one or more of the data blocks of the fast buffer, depending on the subscriber's connection status to the MX node.
- Processes for multiple subscribers can read messages from a single shared data structure (the fast buffer), without creating multiple copies for each message.
- multiple MX nodes are able to read from the fast buffer (e.g., the same data) simultaneously, without blocking or interfering with each other and without modifying the data in the fast buffer.
- a first read process on the MX node can read messages from the fast buffer without blocking (e.g., causing suspension of) a second read process that is on the same MX node but may reside on a different processor core of the MX node.
- FIG. 5 is an example system 500 in which an MX node 502 retrieves messages from a Q node 504 and provides the messages to a client device 506 of a subscriber.
- the MX node 502 can include an MX subscribe process 508 and a fast buffer 510 .
- the MX subscribe process 508 can use a connection 512 to read messages stored on the Q node 504 (e.g., in a streamlet) and can use a connection 514 to provide the messages to the fast buffer 510 for storage.
- the fast buffer 510 may store the messages in one or more blocks, with each block having a TTL.
- the fast buffer 510 can be a shared, lock-free structure that holds the data blocks and expires old data periodically (e.g., when the old data is no longer being read by MX processes or requested by subscribers or after a specified or predetermined time period).
- the MX node 502 can also include an MX connect process 516 and an MX send process 518 for establishing communications with one or more client devices, such as client device 506 .
- the client device 506 connects to the MX connect process 516 using a connection 520 and requests messages from the MX node 502 for one or more channels.
- the client device 506 can request, for example, messages of the channel foo starting at a particular position, such as the position 47202 .
- An example JSON version of this specific request is as follows:
- the MX connect process 516 sends an instruction 524 to the MX send process 518 to retrieve message data stored in the fast bullet 510 (e.g., starting at the position 47202 of the channel fob).
- the MX send process 518 then retrieves the message data from the fast buffer 510 using a connection 526 and sends the messages to the client device 506 through a connection 522 .
- the MX subscribe process 508 and the fast buffer 510 are preferably instantiated on a per-channel basis and/or devoted to retrieving and storing messages for a single channel. Additional MX subscribe processes and/or Fast buffers may be included in the MX node 502 to retrieve and store messages for additional channels. In alternative implementations, the MX subscribe process 508 and the fast buffer 510 can be configured to retrieve and store messages for multiple channels. For example, the MX node 502 may include one MX subscribe process 508 and one fast buffer 510 for retrieving and storing messages for the multiple channels.
- the MX connect process 516 and the MX send process 518 are preferably instantiated on a per-client connection basis and/or devoted to establishing communications with a single client device (e.g., the client device 506 ). Additional MX connect processes and MX send processes may be included on the MX node 502 to establish communications with additional client devices and retrieve and send messages from one or more fast buffers.
- the MX connect process 516 and MX send process 518 can be configured to connect with and serve multiple client devices.
- the MX node 502 may include one MX connect process 516 and one MX send process 518 for receiving requests from client devices (including client device 506 ) and sending messages to the client devices (e.g., for one or more channels) respectively.
- the blocks of data delivered to the MX subscribe process 508 from the Q node 504 can be stored in the fast buffer 510 in the same order as the blocks were received by the fist buffer 510 , although other arrangements or ordering of the data blocks are possible.
- the blocks are preferably given an offset and a timestamp. This allows the blocks to be located and retrieved from the fast buffer 510 and/or allows the MX subscribe process 508 to identify any blocks that are older than their TTL and/or otherwise unavailable for reading. Offsets are generally assigned in ascending order, with blocks added one-by-one to a back portion of a single queue or linked list (e.g., a forward list) in the fast buffer 510 .
- the MX subscribe process 508 may categorize each block as being, for example, active, inaccessible, or expired, although other designations or categorizations are possible.
- active blocks are “young” blocks whose TTLs have not expired. Active blocks reside in a visible or otherwise accessible area of the fast buffer 510 and may be accessed with new read requests.
- Inaccessible blocks are “middle-aged” blocks whose TTLs are about to expire. Inaccessible blocks reside in an invisible or otherwise restricted or no access area of the fast buffer 510 and may be accessed only by read processes that have an established connection (e.g., a previously granted read request).
- Expired blocks reside in the invisible area of the fast buffer 510 and are deleted periodically from the fast buffer 510 .
- the MX subscribe process 508 may shrink the queue in the fast buffer 510 .
- the MX subscribe process 508 may shrink a head, a tail, or other suitable portion of the Past buffer 510 , so that the expired blocks are removed from the queue and are no longer accessible.
- expired blocks may remain available for reading for additional time to accommodate certain processes that were unable to read messages from the blocks within their TTLs.
- client devices (and associated MX send processes) having a current or pre-existing connection to the MX node 502 e.g., a pre-existing read request
- client devices and associated MX send processes having a new connection to the MX node 502 (e.g., with no pre-existing read request) may not be allowed to read data from inaccessible and/or expired blocks.
- client devices with new connections may be limited to reading only active blocks that are presently in a visible or accessible portion of the queue. This approach gives preference to pre-existing connections or pre-existing read requests and allows the corresponding client devices to avoid any gaps in received message data. Such gaps may occur, for example, due to slow connection speeds and/or temporary lapses in a connection.
- a “pre-existing connection” refers to a connection with which an MX send process has requested and read messages from the fast buffer 510 .
- a “new connection” refers to a connection with which an MX send process has not yet read any messages from the fast buffer 510 .
- this approach of categorizing blocks and periodically deleting expired blocks provides improved control over memory usage.
- the amount of memory used by channel data can be dynamically allocated and/or can change over time according to demand. For example, as message activity for a channel decreases, the amount of memory used for the channel can automatically decrease, as expired blocks are deleted and fewer new blocks are created. Also, expired blocks can be deleted after existing read processes have had a chance to read the message data and the message data in the blocks is no longer needed.
- the approach also provides flexibility to add one or more processing steps (e.g., encryption, compression, etc.), and ensures that such steps are not processing or attempting to access memory that has been freed.
- the M X subscribe process 508 can be the only process having authority to modify the structure of the fast buffer 510 .
- the MX subscribe process 508 may be configured and allowed to add blocks to the fast buffer 510 , remove blocks from the fast buffer 510 , and/or modify blocks on the fast buffer 510 , as needed.
- the MX subscribe process 508 may also move pointers that identify a head portion and a tail portion of a queue or linked list in the fast buffer 510 .
- Other processes or MIX node components are preferably not allowed to perform such modifications.
- the MX send process 518 can check the front (i.e., head) and back (i.e., tail) of the fast buffer 510 at the beginning of every read operation.
- the MX send process 518 may read the block from the fast buffer 510 and forward the message data in the block to the client device 506 .
- the MX send process 518 may read an amount of data (e.g., up to 64 Kb) from the fast buffer 510 (e.g., starting from a block having a requested offset or location in the fast buffer 510 ) and forward the message data to the client device 506 .
- the MX send process 518 may apply a separate logic that involves error reporting and/or waiting. The separate logic used by the MX send process 518 may depend on where the target block resides in comparison to accessible regions of the fast buffer 510 .
- the client may be unsubscribed from the channel and a message may be sent to inform that client that it is “out of sync” with the channel's data.
- the client may be fast forwarded to the oldest data available (e.g., the oldest active block) or to the newest data available (e.g., the newest active block).
- the systems and methods may identify one or more key active blocks (e.g., as indicated by one or more flags), forward the data from the key active blocks to the client, and then fast forward the client to the newest active block for further reading.
- the client may have certain preferences that define how error reporting and/or waiting are to be performed. For example, the client may have a preference that causes its state to be reloaded completely (e.g., when the client loses a connection and is able to reconnect).
- the MX send process 51 . 8 can read the accessible blocks in the fast buffer 510 and then can wait to be notified when additional blocks are added to the fast buffer 510 and available for reading. For example, when the MX send process 518 reaches a tail of the fast buffer 510 , the MX send process 518 may turn on or activate a notifier 528 and/or yield to other processes.
- the notifier 528 can be configured to wake up the MX send process 518 upon the arrival of new data into the fast buffer 510 .
- the MX send process 518 may send its process identifier (PID) or other suitable identifier to a waiting list 530 along a connection 532 .
- PID process identifier
- the MX send process 518 may then perform a second check of the fast buffer 510 for new data and, if no new data is present, the MX send process 518 may go to sleep and wait for a message from the notifier 528 .
- Checking the fast buffer 510 again after providing the PID to the waiting list 530 may allow the MX send process 518 to identify the presence of any new data that arrived just before or at around the same time as the PID was added to the waiting list 530 .
- the waiting list 530 includes a list of PIDs for one or more MX send processes that are waiting to be notified when new data arrives in the fast buffer 510 .
- the waiting list 530 may provide PIDs to the notifier 528 along a connection 534 , and the notifier 528 may wake up or notify the MX send processes in the waiting list 530 when new data arrives in the fast buffer 510 and is available for reading.
- the notifier 528 may wake up the MX send process 518 by sending a notification along a connection 536 .
- the waiting list 530 and the notifier 528 are preferably configured to perform operations for a single channel.
- the MX node 502 may include additional waiting lists and notifiers to perform operations for additional channels, in alternative implementations, the waiting list 530 and the notifier 528 can be configured to perform operations for multiple channels.
- the MX node 502 may include one waiting list 530 and one notifier 528 for receiving PIDs of MX send processes and waking up the corresponding MX send processes, respectively, when new data arrives in the fast buffer 510 for the multiple channels.
- the notifier 528 can work together and/or can be synchronized with the MX subscribe process 508 .
- the notifier 528 may receive a list of PIDs from the waiting list 530 for the one or more MX send processes waiting to be notified.
- the notifier 52 $ then delivers a notification to the one or more MX send processes.
- the notification preferably includes the offset or location of the new block in the fast buffer 510 .
- the PIDs for the one or more MX send processes may be removed from the waiting list 530 .
- the notifier 528 can use a two-level model of synchronization with the M X subscribe process 508 .
- the notifier 528 may contact the MX subscribe process 508 at short time intervals (e.g., every 10 ms) to determine if any new blocks have been added to the fast buffer 510 .
- the notifier 528 may, for example, read from a dictionary or log of the MX subscribe process 508 to determine the offset of the block most recently added to the fast buffer 510 .
- the notifier 528 may switch to an outer loop 540 where the notifier 528 goes to sleep and/or waits for a notification from the MX subscribe process 508 .
- the MX subscribe process 508 preferably checks the state of the notifier 528 and wakes up the notifier 528 by sending a notification when the notifier 528 is sleeping and new data has been sent to the fast buffer 510 .
- this approach allows the MX subscribe process 508 to look for new data on the Q node more frequently and to add any such new data to the fast buffer 510 .
- the MX subscribe process 508 informs the notifier 528 that a new block has been added to the fast buffer 510 , the notifier 528 may switch back to the inner loop 538 .
- the two-loop approach allows the MX node 502 to conserve resources and operate more efficiently.
- the notifier 528 actively looks for new message data only when message data is being added to the fast buffer frequently. Otherwise, the notifier 528 is able to sleep and is woken up by the MX subscribe process 508 when new data arrives.
- a primary benefit of the two-loop approach is that resources are used only when there is work to do or when work is expected in the near future. If new channel data is not expected soon, the system switches from the inner loop 538 to the use of push notifications in the outer loop 540 .
- the two-loop approach achieves a good balance between latency and central processing unit (CPU) overhead.
- the outer loop 540 allows the system to maintain millions of channels that are rarely updated and not waste a single CPU cycle on checking the status of such channels.
- the inner loop 538 allows the system to maintain with minimal latency other channels that are updated more frequently.
- the notifier 528 when the notifier 528 becomes aware of new data added to the fast buffer 510 , the notifier 528 sends a notification to any MX send processes waiting for the new data. For example, the notifier 528 may obtain from the waiting list 530 the PIDs of MX send processes that are sleeping and waiting for new data and may send notifications to those MX send processes. The notifications preferably include the offset or other location information for any new blocks of data in the fast butler 510 .
- the systems and methods allow publishers to be prioritized (or deprioritized) with respect to subscribers.
- the fast buffer 510 can be considered to (i) receive data from publishers in a publishing loop and (ii) provide data to subscribers in a subscribing loop.
- the publishing loop can include, for example, a step of obtaining new message data from one or more publishers and a step of writing the new message data to the fast buffer 510 .
- the subscribing loop can include, for example, a step of checking the fast buffer 510 for new data and a step of providing the new data to one or more subscribers (e.g., using the MX send process 518 ).
- the publishing loop and the subscribing loop are loosely coupled or not coupled, such that publishing and subscribing speeds can be independently controlled and/or prioritized.
- the steps of the publishing loop can be performed taster, so that message data can be written more frequently to the fast buffer 510 .
- the steps of the subscribing loop can be performed faster, so that message data can be read more frequently from the fast buffer 510 .
- the steps of one loop when the steps of one loop are performed faster, the steps of the other loop can be performed slower, for example, to maintain overall system load at or near a constant value and/or to avoid excessive system load.
- Such prioritization can be achieved, for example, by creating two worker pools (e.g., one pool for publishers and one pool for subscribers) and using an operating system kernel scheduler prioritization mechanism to set different priorities to threads in the two pools. Other approaches for prioritizing or deprioritzing publishers and subscribers are contemplated.
- FIG. 6 shows example data blocks of the fast butler 510 at different instances in time.
- the fast buffer 510 includes a sequence of data blocks 601 , 602 , . . . , 608 , that were retrieved from one or more streamlets in the Q node 504 .
- the sequence of message data in the data blocks is preferably consistent with or the same as the sequence of message data from the one or more streamlets in the Q node 504 .
- Each data block 601 , 602 , . . . . , 608 in the fast butler 510 is given a respective TTL, in preferred implementations, the TTLs for the blocks are sufficient in duration to give read processes enough time to read data from the blocks. In some instances, the TTLs for one or more blocks are longer than the TTLs for one or more streamlets in the Q node 504 . In general, the TTLs for the blocks are chosen to give the MX send processes adequate time to read from the fast buffer 510 .
- the TTL for the blocks in the fast buffer 510 is about 30 seconds or about 60 seconds, while the TTL for the streamlets in the Q node 504 is about 10 seconds, although other TTLs for the blocks and the streamlets are possible.
- the fast buffer 510 (on the MX node 502 ) can hold more messages (e.g., messages for the past 30 seconds) than the streamlets can hold (e.g., messages for the past 10 seconds) for the particular channel.
- the MX node 502 can thus provide a longer history of messages to client devices of subscribers (e.g., client device 506 ). This is beneficial, for example, when a client device has a slower connection and/or otherwise takes longer (e.g., exceeding to seconds) to read messages that may no longer be accessible or stored in streamlets of the Q node 504 .
- All blocks referenced by pointers represent a single linked list (e.g., a forward list) so all such pointers can be reached by iterating from a first block in the list (e.g., block 601 at time 650 ) using the block's next pointer.
- the visible area 662 includes blocks that are active and the invisible area 660 includes blocks that are inaccessible and/or expired.
- the sequence of data blocks of the fast buffer 510 can be arranged as a linked list 620 .
- each data block has a pointer pointing to the next data block in the sequence of the list.
- data block 604 has a pointer pointing to data block 605 , which has a pointer pointing to data block 606 .
- Data block 608 located at a tail end of the linked list 620 , may have a pointer (e.g., NUL) indicating that it is at the tail end of the linked list 620 .
- the linked list 620 includes the head pointer 640 identifying the block at the head of the linked list 620 .
- the linked list 620 may have a head pointer 640 identifying data block 603 , which has not reached its TTL but will be the next block in the linked list 620 to expire.
- the MX subscribe process 508 can do two things. First, the MX subscribe process 508 can determine the age of the oldest block(s) in the visible area 662 (i.e., block 603 at time 650 ), to determine whether the oldest block(s) should be moved to the invisible area 660 . If one or more blocks in the visible area 662 are inaccessible (i.e., because their TTLs will expire soon) or have expired, the MX subscribe process 508 may move the one or more blocks to the invisible area 660 or delete the one or more blocks from the fast buffer 510 . In one example, blocks that are inaccessible are moved to the invisible area 660 and blocks that have expired are deleted. The process of deleting or removing one or more blocks from the fast buffer 510 may be referred to herein as “shrinking”. Next, the MX subscribe process 508 can add one or more blocks of the data to a tail of the fast buffer 510 .
- the shrinking and adding processes are done without the use of locks or atomic operations, and resulting changes are preferably made in a conflict free order.
- locks or atomic operations are a significant source of poor performance on symmetric multiprocessing (i.e., SMP or multi core) systems. Every such operation must ensure that core or non-uniform memory access (NUMA) node caches do not have a cached value that can be read right after a change was made.
- NUMA non-uniform memory access
- Locks and/or atomic operations are expensive on modern CPUs and can cost hundreds of CPU cycles.
- the fast buffer 510 performs truly in parallel, such that multiple read processes can read from the same memory simultaneously. With the fast buffer approach described herein, preferably nothing other than boundary pointers requires a synchronization during the shrinking and adding processes.
- shrinking is preferably done with the following technique.
- expired blacks in the invisible area 660 are deleted.
- expired or inaccessible blocks in the invisible area 560 that are younger than some threshold time e.g., TTL*1.5
- the previous head pointer 644 is then moved to the current location of the head pointer 640 , and the head pointer 640 is moved toward the tail pointer 642 until a block is found that is active e.g., block 605 at time 652 ).
- the head pointer 640 is placed at the front edge of that first active block. If no active blocks are found, the head pointer 640 and the tail pointer 642 may mark the same location, indicating that the fast buffer 510 includes no active blocks.
- the shrinking process is further illustrated at times 652 , 654 , and 658 for the fast buffer 510 in FIG. 6 .
- the head pointer 640 and previous head pointer 644 have been moved, such that expired blocks 601 and 602 have been removed and no longer reside in the invisible area 660 , and the invisible area 660 now includes blocks 603 and 604 .
- the tail pointer 542 has also been moved to define a new tail at block 611 .
- the head painter 640 and the previous head pointer 644 have been moved, such that expired blocks 603 and 604 have been removed and no longer reside in the invisible area 660 .
- the invisible area 660 includes no blocks and the head pointer 640 and the previous head pointer 644 mark the same location (i.e., block 605 at time 654 ).
- Each block in the fast buffer 510 at time 554 is active and resides in the visible area 662 .
- the head pointer 640 and the previous head pointer 644 have been moved, such that expired blocks 605 and 606 have been removed and no longer reside in the invisible area 560 .
- the invisible area 660 at that time includes blocks 607 and 608 , and the previous head pointer 644 has been moved to identify the leftmost block of the invisible area 660 (i.e., block 607 at time 658 ).
- the tail pointer 642 at time 658 has been moved to define a new tail at block 615
- the head pointer 640 at time 658 has been moved to define a new head at block 609 .
- shrinking the fast buffer 510 and repositioning the pointers preferably occurs while certain MX send processes continue to have access to blocks in the new invisible area 660 .
- the invisible area 660 (e.g., including blocks 603 and 604 at time 652 ) may therefore be left accessible to existing MX send processes and not freed instantly. This allows existing read requests for MX send processes to continue accessing inaccessible and/or expired blocks after the blocks have been moved from the visible area 662 to the invisible area 660 .
- MX send processes that are new and/or have new read requests may be given access only to the active blocks in the visible area 662 of the fast buffer 510 .
- each MX send process e.g., MX send process 518
- MX send process 518 each MX send process having a new read request checks the boundaries of the visible area 662 and if the process requests access outside of the visible area 662 (e.g., to any blocks between the previous head and the head), the request fails as if there were no data available for the request.
- MX send processes having new requests are not allowed to access any inaccessible or expired blocks in the invisible area 660 .
- the invisible area 660 can be accessed only by MX send processes having old requests, and the MX send processes preferably finish the read operation from the invisible area 660 as soon as possible. Typical reading times take a few microseconds or milliseconds, because there are preferably no locks in the reading code.
- blocks reside in the visible area 662 for approximately one TTL and are then moved to the invisible area 660 where the blocks reside for some fraction (e.g., 0.25, 0.5, or 0.75) of the TTL. For example, after spending, approximately one TTL in the visible area 662 , a block may reside in the invisible area 660 for an additional TTL*0.5. This time in the invisible area 660 is generally many orders of magnitude longer than a typical access time and provides pre-existing MX send processes with additional time to obtain the requested message data. As discussed previously, a block resides in the visible area 662 when the block is active and is moved to the invisible area 660 when the block is inaccessible or expired.
- new blocks are combined into a separate single linked list 620 that is linked to the block identified by the current tail pointer 642 .
- the tail pointer 642 is then moved to the last added block.
- This block addition process is illustrated in FIG. 6 at times 652 , 656 , and 658 .
- new blocks have been added to the fast buffer 510 , compared to the previous time, and the tail pointer 642 has been moved to identify the last added block. For example, at time 652 , new blocks 609 , 610 , and 611 have been added and the tail pointer 642 points to the right-hand edge of block 611 , indicating that block 611 is the last added block.
- a data block contains zero messages and may serve as a gap block or a left boundary marker.
- Gap blocks may he used when there is a gap in block positions in the fast buffer 510 (e.g., when a previous block position plus a previous block length does not equal a next block position). Such gaps can be due to, for example, switching from one streamlet to another streamlet streamlets that are 16 m in size can be aligned by a 16 m boundary) or missing data, in which case clients reaching the gap in the buffer may need to be informed about anther provided with the missing data.
- a left boundary marker may be a zero message block or marker that indicates the leftmost available location in the fast buffer 510 where data was first written or intended to be written.
- the zero message block may contain the leftmost offset even when no data is available yet.
- the left boundary matter can help the MX node 502 make decisions about certain read requests and whether it will be possible to provide a client device with the requested data. Having the left boundary defined for every read request may make it possible to determine whether the read request can be satisfied in the future (e.g., when data is received) or not.
- fast buffers described herein are internal components inside a distributed PubSub system. While fast buffers can be included in or used by MX nodes, as shown in FIG. 5 , fast buffers can also be included in or used by other system components, including Q nodes.
- a fast buffer publisher is, in general, a component that writes data to a fast buffer
- a fast buffer subscriber is a component that reads data from a fast buffer.
- the MX subscribe process 508 is a fast buffer publisher and the MX send process 518 is a fast buffer subscriber, for the fast buffer 510 in MX node 502 . Similar fast buffer publishers and fast buffer subscribers can be used for Q nodes.
- a Q node fast buffer publisher is referred to as a Q streamlet process and a Q node fast buffer subscriber is referred to as a Q sender process.
- Fast butlers incorporated into or used by Q nodes can use the notifier and/or waiting list approach described herein with respect to MX node 502 .
- a Q sender process waiting to retrieve new message data from a Q node fast buffer can be put to sleep when no data is available and/or woken up when new message data arrives.
- FIG. 7 is a flowchart of example method 700 for providing messages to a plurality of subscribers.
- the method includes receiving (step 702 ) from a plurality of publishers a plurality of messages.
- Each of the messages is assigned (step 704 ) to one of a plurality of distinct channels wherein each channel includes an ordered plurality of messages.
- the messages of each of the channels are stored (step 706 ) in one or more respective storage butlers according to the order.
- Each storage buffer has a respective time-to-live of a first time duration.
- messages of the particular channel are retrieved (step 708 ) from respective storage buffers (e.g., those having times-to-live that have not expired and according to the order).
- the retrieved messages of the particular channel are stored (step 710 ) in a delivery buffer according to the order.
- the delivery buffer includes one or more data blocks, and each data block has a respective time-to-live of a second time duration (e.g., that is longer than the first time duration).
- the messages stored in the delivery buffer are provided (step 712 ) to a plurality of subscribers of the particular channel.
- Examples of the systems and methods described are able to send a stream of data from one source (e.g., an ERLANG process) to multiple destinations (e.g., ERLANG processes).
- the stream of data is delivered at a high rate (e.g., in messages per second) and with low latency (e.g., 99% of messages are delivered with a latency less than 50 ms.
- a first alternative approach involves sending messages sequentially from a sender process to all the destination processes, which may be represented as a list or other data structure. This process is generally slow and measured to be a bottleneck, for example, because locking on multicore processing can create a contention point around sending a message to processes running on other cores.
- the process also requires a lot of memory because every sending operation creates a copy of data to be sent (or a copy of reference to the data). The copy typically resides in the destination process message queue until the destination process handles the data. Using multiple senders does not solve the problem.
- ERLANG ETS a table storage
- ERLANG ETS e.g., an ERLANG ETS
- ETS electronic book reader
- This approach is also slow, however, because access to the table storage presents a bottleneck.
- ERLANG ETS tables utilize a set of locks to protect ranges of keys from multiple accesses and order these accesses.
- the fast buffer approach described herein tasks the receivers or subscribers themselves to grab data from a single shared structure (i.e., a fast buffer).
- a fast buffer i.e., a fast buffer
- the process utilizes a “lock-free” or non-blocking approach.
- Receivers are able to grab many pieces of data without blocking or figuring out where to read from and up to which point.
- Table 1 indicate, the rates of message transfer are improved considerably when compared to the alternative approaches.
- Embodiments of the subject matter and the operations described in this specification can be implemented in digital electronic circuitry, or in computer software, firmware, or hardware, including the structures disclosed in this specification and their structural equivalents, or in combinations of one or more of them.
- Embodiments of the subject matter described in this specification can be implemented as one or more computer programs, i.e., one or more modules of computer program instructions, encoded on computer storage medium for execution by, or to control the operation of, data processing apparatus.
- the program instructions can be encoded on an artificially-generated propagated signal, e.g., a machine-generated electrical, optical, or electromagnetic signal, that is generated to encode information for transmission to suitable receiver apparatus for execution by a data processing apparatus.
- a computer storage medium can be, or be included in, a computer-readable storage device, a computer-readable storage substrate, a random or serial access memory array or device, or a combination of one or more of them.
- a computer storage medium is not a propagated signal, a computer storage medium can be a source or destination of computer program instructions encoded in an artificially-generated propagated signal.
- the computer storage medium can also be, or be included in, one or more separate physical components or media (e.g., multiple CDs, disks, or other storage devices).
- the operations described in this specification can be implemented as operations performed by a data processing apparatus on data stored on one or more computer-readable storage devices or received from other sources.
- the term “data processing apparatus” encompasses all kinds of apparatus, devices, and machines for processing data, including by way of example a programmable processor, a computer, a system on a chip, or multiple ones, or combinations, of the foregoing.
- the apparatus can include special purpose logic circuitry, e.g., an FPGA (field programmable gate array) or an ASIC (application-specific integrated circuit).
- the apparatus can also include, in addition to hardware, code that creates an execution environment for the computer program in question, e.g., code that constitutes processor firmware, a protocol stack, a database management system, an operating system, a cross-platform runtime environment, a virtual machine, or a combination of one or more of them.
- the apparatus and execution environment can realize various different computing model infrastructures, such as web services, distributed computing and grid computing infrastructures.
- a computer program (also known as a program, software, software application, script, or code) can be written in any form of programming language, including compiled or interpreted languages, declarative, procedural, or functional languages, and it can be deployed in any form, including as a stand-alone program or as a module, component, subroutine, object, or other unit suitable for use in a computing environment.
- a computer program may, but need not, correspond to a tile in a file system.
- a program can be stored in a portion of a file that holds other programs or data (e.g., one or more scripts stored in a markup language resource), in a single file dedicated to the program in question, or in multiple coordinated tiles (e.g., files that store one or more modules, sub-programs, or portions of code).
- a computer program can be deployed to be executed on one computer or on multiple computers that are located at one site or distributed across multiple sites and interconnected by a communication network.
- the processes and logic flows described in this specification can be performed by one or more programmable processors executing one or more computer programs to perform actions by operating on input data and generating output.
- the processes and logic flows can also be performed by, and apparatus can also be implemented as, special purpose logic circuitry, e.g., an FPGA (field programmable gate array) or an ASIC (application-specific integrated circuit).
- processors suitable for the execution of a computer program include, by way of example, both general and special purpose microprocessors, and any one or more processors of any kind of digital computer.
- a processor will receive instructions and data from a read-only memory or a random access memory or both.
- the essential elements of a computer are a processor for performing actions in accordance with instructions and one or more memory devices for storing instructions and data.
- a computer will also include, or be operatively coupled to receive data from or transfer data to, or both, one or more mass storage devices fear storing data, e.g., magnetic disks, magneto-optical disks, optical disks, or solid state drives.
- mass storage devices fear storing data, e.g., magnetic disks, magneto-optical disks, optical disks, or solid state drives.
- a computer need not have such devices.
- a computer can be embedded in another device, e.g., a smart phone, a mobile audio or video player, a game console, a Global Positioning System (GPS) receiver, or a portable storage device (e.g., a universal serial bus (USB) flash drive), to name just a few.
- Devices suitable for storing computer program instructions and data include all forms of non-volatile memory, media and memory devices, including, by way of example, semiconductor memory devices, e.g., EPROM, EEPROM, and flash memory devices; magnetic disks, e.g., internal hard disks or removable disks; magneto-optical disks; and CD-ROM and DVD-ROM disks.
- the processor and the memory can be supplemented by, or incorporated in, special purpose logic circuitry.
- a computer having a display device, e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor, for displaying information to the user and a keyboard and a pointing device, e.g., a mouse, a trackball, a touchpad, or a stylus, by which the user can provide input to the computer.
- a display device e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor
- a keyboard and a pointing device e.g., a mouse, a trackball, a touchpad, or a stylus
- Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback, e.g., visual feedback, auditory feedback, or tactile feedback; and input from the user can be received in any form, including acoustic, speech, or tactile input.
- a computer can interact with a user by sending resources to
- Embodiments of the subject matter described in this specification can be implemented in a computing system that includes a back-end component, e.g., as a data server, or that includes a middleware component, e.g., an application server, or that includes a front-end component, e.g., a client computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the subject matter described in this specification, or any combination of one or more such back-end, middleware, or front-end components.
- the components of the system can be interconnected by any form or medium of digital data communication, e.g., a communication network.
- Examples of communication networks include a local area network (“LAN”) and a wide area, network (“WAN”), an inter-network (e.g., the Internet), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks).
- LAN local area network
- WAN wide area, network
- inter-network e.g., the Internet
- peer-to-peer networks e.g., ad hoc peer-to-peer networks.
- the computing system can include clients and servers.
- a client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other.
- a server transmits data (e.g., an HTML page) to a client device (e.g., for purposes of displaying data to and receiving user input from a user interacting with the client device).
- client device e.g., for purposes of displaying data to and receiving user input from a user interacting with the client device.
- Data generated at the client device e.g., a result of the user interaction
- a system of one or more computers can be configured to perform particular operations or actions by virtue of having software, firmware, hardware, or a combination of them installed on the system that in operation causes or cause the system to perform the actions.
- One or more computer programs can be configured to perform particular operations or actions by virtue of including instructions that, when executed by data processing apparatus, cause the apparatus to perform the actions.
Abstract
Description
- This application is a continuation of U.S. application Ser. No. 15/909,710, filed Mar. 1, 2018, which is a continuation of U.S. application Ser. No. 15/586,773, filed May 4, 2017 (now U.S. Pat. No. 9,942,339, issued Apr. 10, 2018), which is a continuation of U.S. application Ser. No. 15/274,281, filed Sep. 23, 2016 (now U.S. Pat. No. 9,667,681, issued May 30, 2017), the entire contents of each of which are hereby incorporated by reference.
- This specification relates to a data communication system and, in particular, a system that implements real-time, scalable publish-subscribe messaging.
- The publish-subscribe pattern (or “PubSub”) is a data communication messaging arrangement implemented by software systems where so-called publishers publish messages to topics and so-called subscribers receive the messages pertaining to particular topics to which they are subscribed. There can be one or more publishers per topic and publishers generally have no knowledge of what subscribers, if any, will receive the published messages. Some PubSub systems do not cache messages or have small caches meaning that subscribers may not receive messages that were published before the time of subscription to a particular topic. PubSub systems can be susceptible to performance instability during, surges of message publications or as the number of subscribers to a particular topic increases.
- In various implementations, the systems and methods described herein utilize a shared structure (i.e., a “fast buffer”) to store message data that can be read by multiple subscribers simultaneously. To avoid issues associated with locking, the systems and methods utilize a “lock-free” or non-blocking approach. Subscribers are able to grab many pieces of data without blocking or figuring out where to read from and up to which point in the fast buffer.
- The systems and methods provide many benefits compared to prior approaches. For example, publishers are not limited by subscribers or associated subscription logic and have a stable publishing speed, regardless of how many subscribers exist, if any. This property of the system can be achieved, for example, by making publishing and subscribing components and/or processes independent from each other. In some instances, the publishing and subscribing components can share information through lock-free memory writes and reads, for example, using the fast buffer approach described herein. Alternatively or additionally, publishing and subscribing components and/or processes can be configured to utilize different physical cores, such that, for example, publishing and subscribing components interact solely through a memory controller and bus, which is typically fast and/or does not present a bottleneck.
- Alternatively or additionally, system flexibility allows publishers to be prioritized (or deprioritized) over subscribers, if desired. In preferred examples, the fast buffer has or utilizes loosely coupled publishing and subscribing loops. This allows a publishing speed to be different from a subscribing speed, if desired, and/or for each speed to be chosen or prioritized, for example, according system load. When publishing activity is high and/or it is preferable to not limit the publishing activity, for example, the publishing loop can be prioritized over the subscribing loop. To achieve such prioritization, the publishing loop can be made to spin as if there were no subscribers (e.g., the publishing loop can be processed faster than the subscribing loop is processed). In other words, it can be desirable to not block or limit publishers by allowing the system to pass as much published message data as possible, even if subscribers are unable to keep up with or receive all of the message data. Alternatively or additionally, when subscribing activity is high, it can be desirable to not block or limit subscribers and/or to apply automatic back-pressure on publishers, such that publishing activity is reduced. Such prioritization can be achieved, for example, by creating two worker pools (e.g., one pool for publishers and one pool for subscribers) and using an operating system kernel scheduler prioritization mechanism to set different priorities to threads in the two pools. Other approaches for prioritizing or deprioritzing publishers and subscribers are contemplated.
- A further advantage of the systems and methods is that subscribers can utilize data aggregation and can define how much data to aggregate. In general, use of data aggregation and/or aggregation size optimization can help achieve better performance characteristics and/or reduce memory usage. For example, when delivering messages, the messages can be grouped into blocks or hatches (e.g., a comma-separated list of messages in the fast buffer). When preparing message data to be sent, the systems and methods can decide how many messages to aggregate for the next block. The sizes of packets exchanged in a connection can depend on connection properties (e.g., bandwidth and/or latency) and/or on a state of a TCP connection. In some instances, this allows the process of collecting messages for the connection to aggregate more messages in a block to produce a bigger packet, thereby better utilizing network capacity and providing higher throughput with less overhead.
- Another benefit of the systems and methods is that, depending on message activity, subscribers can change the frequency of data reading by finding an optimal balance between performance and latency. For example, using the notifier approach described herein, channels with high message activity can be actively monitored for new messages, while channels with low message activity can receive less active monitoring. Subscribers are preferably isolated and do not affect one other. If one subscriber is slower than a message stream, other subscribers are generally unaffected by the slow subscriber and can keep up with the message stream.
- A further benefit of the systems and methods is that modularity is increased, which allows system components to be implemented independently. Interaction is preferably done through memory, so subscribers can be placed on separate individual processes to access the fist buffer. In alternative approaches, multiple logical parts of a system can be coupled tightly to reduce interface overhead (e.g., preparing data to be exported and parsing on input) and/or to make changes more granular and precise. The drawback of such approaches is reduced isolation, reduced reliability and, in general, less efficiency (e.g., due to load of one component influencing other components). The opposite approach, used in examples of the systems and methods described herein, is to isolate components from each other and to achieve increased reliability and efficiency and better management and prioritization of system components. In general, isolation is more effective when the cost of overhead is low or at least acceptable, in the systems and methods, for example, the cost of overhead is generally low because processes can communicate by writing to and reading from the fast buffer. The systems and methods can benefit from such modularity with little or no added cost. In some implementations, when shared memory at the operating system level is used, publishers and subscribers can be implemented in different processes and/or can start using implementations written in different languages. In general, interaction through use of the fast buffer limits communication to only two operations: writing to the memory and reading from the memory. The use of messages, signals, locks, and/or synchronization primitives can be avoided. As discussed herein, test results indicate that the systems and methods greatly improve message transfer rates from publishers to subscribers.
- In general, one aspect of the subject matter described in this specification can be embodied in computer-implemented methods that include the actions of: retrieving messages of a particular channel of a plurality of channels from respective storage buffers, wherein each storage buffer comprises a respective first time-to-live. The method may also include storing the retrieved messages of the particular channel in a delivery buffer, wherein the delivery buffer comprises one or more data blocks, and wherein each data block comprises a respective second time-to-live. The method may also include transmitting, by one or more computer processors, the messages stored in the delivery buffer to subscribers of the particular channel. Other embodiments of this aspect include corresponding systems, apparatus, storage devices, and computer programs.
- The details of one or more embodiments of the subject matter described in this specification are set forth in the accompanying, drawings and the description below. Other features, aspects, and advantages of the subject matter will become apparent from the description, the drawings, and the claims.
-
FIG. 1A illustrates an example system that supports the PubSub communication pattern. -
FIG. 1B illustrates functional layers of software on an example client device. -
FIG. 2 is a diagram of an example messaging system. -
FIG. 3A is a data flow diagram of an example method for writing data to a streamlet. -
FIG. 3B is a data flow diagram of an example method for reading data from a streamlet. -
FIG. 4A is a data flow diagram of an example method for publishing messages to a channel of a messaging system. -
FIG. 4B is a data flow diagram of an example method for subscribing to a channel of a messaging system. -
FIG. 4C is an example data structure for storing messages of a channel of a messaging system. -
FIG. 5 is a schematic diagram of an MX Node retrieving messages from a Q node and providing the retrieved messages to a subscriber of a messaging system. -
FIG. 6 shows example data blocks of a fast buffer in an MX node of a messaging system at different time instances. -
FIG. 7 is a flowchart of an example method for sending message data to a plurality of subscribers. - Like reference numbers and designations in the various drawings indicate like elements.
-
FIG. 1A illustrates anexample system 100 that supports the PubSub communication pattern. Publishers e.g., Publishers 1-N) can publish messages to named channels (e.g., Channels 1-N) by way of the system 100 (also referred to as “messaging system” hereafter). A message can include any type of information including one or more of the following: text, image content, sound content, multimedia content, video content, binary data, and so on. Other types of message data are possible. Subscribers (e.g., Subscribers 1-N) can subscribe to a named channel using thesystem 100 and start receiving messages which occur after the subscription request or from a given position (e.g., a message number or time offset). A client can be both a publisher and a subscriber. - Depending on the configuration, a PubSub system can be categorized as follows:
- One to One (1:1). In this configuration there is one publisher and one subscriber per channel. An example use case is private messaging.
- One to Many (1:N). In this configuration there is one publisher and multiple subscribers per channel. Example use cases are broadcasting messages (e.g., stock prices).
- Many to Many (M:N). In this configuration there are multiple publishers publishing to a single channel. The messages are then delivered to multiple subscribers. Example use cases are map applications.
- There is no separate operation needed to create a named channel. A channel is created implicitly when the channel is subscribed to or when a message is published to the channel. In some implementations, channel names can be qualified by a name space. A name space includes one or more channel names. Different name spaces can have the same channel names without causing ambiguity. The name space name can be a prefix of a channel name where the name space and channel name are separated by a dot or other suitable separator. In some implementations, name spaces can be used when specifying channel authorization settings. For instance, the
system 100 may have app1.foo and app1.system.notifications channels where “app1” is the name of the name space. The system can allow clients to subscribe and publish to the app1.foo channel. However, clients can subscribe, but not publish, to the app1.system.notiftcations channel. -
FIG. 1B illustrates functional layers of software on an example client device. A client device (e.g., client 102) is a data processing apparatus such as, for example, a personal computer, a laptop computer, a tablet computer, a smart phone, a smart watch, or a server computer. Other types of client devices are possible. Theapplication layer 104 includes the end-user application(s) that will integrate with thesystem 100. Themessaging layer 106 is a programmatic interface for theapplication layer 104 to utilize services of thesystem 100 such as channel subscription, message publication, message retrieval, user authentication, and user authorization. In some implementations, the messages passed to and from themessaging layer 106 are encoded as JavaScript Object Notation (JSON) objects. Other message encoding schemes are possible. - The
operating system layer 108 includes the operating system software on theclient 102. In various implementations, messages can be sent and received to/from thesystem 100 using persistent or non-persistent connections. Persistent connections can be created using, for example, network sockets. A transport protocol such as TCP/IP layer 112 implements the Transport Control Protocol Internet Protocol communication with thesystem 100 that can be used by themessaging layer 106 to send messages over connections to thesystem 100. Other communication protocols are possible including, for example, User Datagram Protocol (UDP). In further implementations, an optional Transport Layer Security (TLS)layer 110 can be employed to ensure the confidentiality of the messages. -
FIG. 2 is a diagram of anexample messaging system 100. Themessaging system 100 provides functionality for implementing PubSub communication patterns. Themessaging system 100 includes software components and storage that can be deployed at one ormore data centers 122 in one or more geographic locations, for example. Themessaging system 100 includes multiplexer (MX)nodes nodes channel managers 214, 215), and optionally one or more cache (C)nodes external network 216. The internal communication among MX nodes, Q nodes, C nodes, and the channel managers, is conducted over aninternal network 218. For example,MX node 204 can be the terminus of a subscriber connection fromclient 102. Each Q node buffers channel data for consumption by the MX nodes. An ordered sequence of messages published to a channel is a logical channel stream. For example, if three clients publish messages to a given channel, the combined messages published by the clients include a channel stream. Messages can be ordered in a channel stream. For example, the messages may be ordered by time of publication by the client, by time of receipt by an M X node, or by time of receipt by a Q node. Other ways for ordering messages in a channel stream are possible. In the case where more than one message would be assigned to the same position in the order, one of the messages can be chosen (e.g., randomly) to have a later sequence in the order. Each channel manager node is responsible for managing Q node load by splitting channel streams into streamlets, as will be discussed in further detail below. The optional C nodes provide caching and load removal from the Q nodes. - In the
example messaging system 100, one or more client devices (publishers and/or subscribers) establish respective persistent connections (e.g., TCP connections) to an MX node (e.g.,MX nodes MX nodes - In the example messaging system 200, a Q node (e.g.,
nodes - When receiving a publish request from
client 102, an MX node (e.g., MX node 204) makes a request to a channel manager (e.g., channel manager 214) to grant access to a streamlet to write the message being published. However, if the MX node has already been granted write access to a streamlet for the channel (and the channel has not been closed to writing), the MX node can write the message to that streamlet without having to request a grant to access the streamlet. Once a message is written to a streamlet for a channel, the message can be read by MX nodes and provided to subscribers of that channel. - Similarly, when receiving a channel subscription request from a client device, an MX node makes a request to a channel manager to grant access to a streamlet for the channel from which messages are read. If the MX node has already been granted read access to a streamlet for the channel (and the channel's TTL has not been closed to reading) the MX node can read messages from the streamlet without having to request a grant to access the streamlet. The read messages can then be forwarded to client devices that have subscribed to the channel. In various implementations, messages read from streamlets are cached by MX nodes so that MX nodes can reduce the number of times messages are read from the streamlets.
- In implementations, an MX node can request a grant from the channel manager that allows the MX node to store a block of data into a streamlet on a particular Q node that stores streamlets of a particular channel. Example streamlet grant request and grant data structures are as follows:
-
StreamletGrantRequest = { “channel”: string( ) “mode”; “read” | “write” “position”: 0 } StreamletGrantResponse = { “streamlet-id”: “abcdef82734987”, “limit-size”: 2000000, # 2 megabytes max “limit-msgs”: 5000, # 5 thousand messages max “limit-life”: 4000, # the grant is valid for 4 seconds “q-node”: string( ) “position”: 0 } - The StreamletGrantRequest data structure stores the name of the stream channel and a mode indicating whether the MX node intends on reading from or writing to the streamlet. The MX node sends the StreamletGrantRequest to a channel manager node. The channel manager node, in response, sends the MX node a StreamletGrantResponse data structure. The StreamletGrantResponse contains an identifier of the streamlet (streamlet-id), the maximum size of the streamlet (limit-sizes), the maximum number of messages that the streamlet can store (limit-msgs), the TTL (limit-life), and an identifier of a Q node (q-node) on which the streamlet resides. The StreamletGrantRequest and StreamletGrantResponse can also have a position field that points to a position in a streamlet (or a position in a channel) for reading from the streamlet.
- A grant becomes invalid once the streamlet has closed. For example, a streamlet is closed to reading and writing once the streamlet's TTL has expired or when the streamlet's storage is full. When a grant becomes invalid, the MX node can request a new grant from the channel manager to read from or write to a streamlet. The new grant will reference a different streamlet and will refer to the same or a different Q node depending on where the new streamlet resides.
-
FIG. 3A is a data flow diagram of anexample method 300 for writing data to a streamlet in various embodiments. InFIG. 3A , when an MX node's (e.g., MX node 202) request to write to a streamlet is granted by a channel manager (e.g., channel manager 214), theMX node 202 establishes a Transmission Control Protocol (TCP) connection with the Q node (e.g., Q node 208) identified in the grant response received from the channel manager (302). A streamlet can be written concurrently by multiple write grants (e.g., for messages published by, multiple publishers). Other types of connection protocols between theMX node 202 and theQ node 208 are possible. - The
MX node 202 sends a prepare-publish message with an identifier of a streamlet that theMX node 202 wants to write to the Q node 208 (304). The streamlet identifier and Q node identifier can be provided by the channel manager in the write grant as described earlier. TheQ node 202 provides the message to a handler 301 (e.g., a computing process running on the Q node 208) for the identified streamlet (306). Thehandler 301 can send an acknowledgment to the MX node 202 (308). After receiving the acknowledgement, theMX node 202 starts writing (publishing) messages (e.g., 310, 312, 314, and 318) to thehandler 301, which stores the received data in the identified streamlet. Thehandler 301 can also send acknowledgements (316, 320) to theMX node 202 for the received data. In some implementations, acknowledgements can be piggy-backed or cumulative. For example, thehandler 301 can send an acknowledgement to theMX node 202 for every predetermined amount of data received (e.g., for every 100 messages received) or for every predetermined time period (e.g., for every one millisecond). Other acknowledgement scheduling algorithms, such as Nagle's algorithm, can be used. - If the streamlet can no longer accept published data (e.g., the streamlet is full), the
handler 301 sends a Negative-Acknowledgement (NAK) message (330) indicating a problem, following by an EOF (end-of-file) message (332). In this way, thehandler 301 closes the association with theMX node 202 for the publish grant. TheMX node 202 can request a write grant for another streamlet from a channel manager if theMX node 202 has additional messages to store. -
FIG. 3B is a data flow diagram of anexample method 350 for reading data from a streamlet in various embodiments. InFIG. 3B , an MX node (e.g., MX node 204) sends a request to a channel manager (e.g., channel manager 214) for reading a particular channel starting from a particular message or time offset in the channel. The channel manager returns a read grant to theMX node 204 including an identifier of a streamlet containing the particular message, a position in the streamlet corresponding to the particular message, and an identifier of a Q node (e.g., Q node 208) containing the particular streamlet. TheMIX node 204 then establishes a TCP connection with the Q node 208 (352). Other types of connection protocols between theMX node 204 and theQ node 208 are possible. - The
MX node 204 then sends a subscribe message (354) to theQ node 208 with the identifier of the streamlet in the Q node and the position in the streamlet from which theMX node 204 wants to read (356). TheQ node 208 provides the subscribe message to ahandler 351 of the streamlet (356). The handier 351 can send an acknowledgement to the MX node 204 (358). Thehandler 351 sends messages (360, 364, 366), starting at the position in the streamlet, to theMX node 204. In some implementations, thehandler 351 can send all of the messages in the streamlet to theMX node 204. After sending the last message in a particular streamlet, thehandler 351 can send a notification of the last message to theMX node 204. TheMX node 204 can send to the channel manager another request for another streamlet containing a next message in the particular channel. - If the particular streamlet is closed (e.g., after its has expired), the
handler 351 can send an unsubscribe message (390), followed by an EOF message (392), to close the association with theMX node 204 for the read grant. TheMX node 204 can close the association with thehandler 351 when theMX node 204 moves to another streamlet for messages in the particular channel (e.g., as instructed by the channel manager). TheMX node 204 can also close the association with thehandler 351 if theMX node 204 receives an unsubscribe message from a corresponding client device. - In various implementations, a streamlet can be written into and read from at the same time instance. For example, there can be a valid read grant and a valid write grant at the same time instance. In various implementations, a streamlet can be read concurrently by multiple read grants (e.g., for channels subscribed to by multiple publisher clients). The handler of the streamlet can order messages from concurrent write grants based on, for example, time-of-arrival, and store the messages based on the order. In this way, messages published to a channel from multiple publisher clients can be serialized and stored in a streamlet of the channel.
- In the
messaging system 100, one or more C nodes (e.g., C node 220) can offload data transfers from one or more Q nodes. For example, if there are multiple: MX nodes requesting streamlets from Q nodes for a particular channel, the streamlets can be offloaded and cached in one or more C nodes. The MX nodes (e.g., as instructed by read grants from a channel manager) can read the streamlets from the C nodes instead. - As described above, messages for a channel in the
messaging system 100 are ordered in a channel stream. A channel manager (e.g., channel manager 214) splits the channel stream into fixed-sized streamlets that each reside on a respective Q node. In this way, storing a channel stream can be shared among many Q nodes; each Q node stores a portion (one or more streamlets) of the channel stream. More particularly, a streamlet can be stored in, for example, registers and for dynamic memory elements associated with a computing process on a Q node, thus avoiding the need to access persistent, slower storage devices such as hard disks. This results in faster message access. The channel manager can also balance loads among Q nodes in themessaging system 100 by monitoring respective workloads of the Q nodes and allocating streamlets in a way that avoids overloading any one Q node. - In various implementations, a channel manager maintains a list identifying each active streamlet, the respective node on which the streamlet resides, an identification of the position of the first message in the streamlet, and whether the streamlet is closed for writing. In some implementations, Q nodes notify the Channel manager and any MX nodes that are publishing to a streamlet that the streamlet is closed due to being full or when the streamlet's TTL has expired. When a streamlet is closed, the streamlet remains on the channel manager's list of active streamlets until the streamlet's TTL has expired so that MX nodes can continue to retrieve messages from the streamlet.
- When an MX node requests a write grant for a given channel and there is not a streamlet for the channel that can be written to, the channel manager allocates a new streamlet on one of the Q nodes and returns the identity of the streamlet and the Q node in the StreamletGrantResponse to the MX node. Otherwise, the channel manager returns the identity of the currently open for writing streamlet and corresponding Q node in the StreamletGrantResponse to the MX node. MX nodes can publish messages to the streamlet until the streamlet is full or the streamlet's TTL has expired, after which a new streamlet can be allocated by the channel manager.
- When an MX node requests a read grant for a given channel and there is not a streamlet for the channel that can be read from, the channel manager allocates a new streamlet on one of the Q nodes and returns the identity of the streamlet and the Q node in the StreamletGrantResponse to the MX node. Otherwise, the channel manager returns the identity of the streamlet and Q node that contains the position from which the MX node wishes to read to the MX node. The Q node can then begin sending messages to the MX node from the streamlet beginning at the specified position until there are no more messages in the streamlet to send. When a new message is published to a streamlet, MX nodes that have subscribed to that streamlet will receive the new message. If a streamlet's TTL has expired, the
handier process 351 sends an Eat: message (392) to any MX nodes that are subscribed to the streamlet. - As described earlier in reference to
FIG. 2 , themessaging system 100 can include multiple channel managers (e.g.,channel managers 214, 215). Multiple channel managers provide resiliency and prevent single point of failure. For instance, one channel manager can replicate lists of streamlets and current grants it maintains to another “slave” channel manager. As for another example, multiple channel managers can coordinate operations between them using distributed consensus protocols, such as, for example, Paxos or Raft protocols. -
FIG. 4A is a data flow diagram of anexample method 400 for publishing messages to a channel of a messaging system. InFIG. 4A , publishers (e.g.,publishers FIG. 2 . For instance,publishers 402 respectively establishconnections 411 and send publish requests to theMX node 202.Publishers 404 respectively establish connections 413 and send publish requests to theMX node 206.Publishers 406 respectively establishconnections 415 and send publish requests to theMX node 204. Here, the MX nodes can communicate (417) with a channel manager (e.g., channel manager 214) and one or more Q nodes (e.g.,nodes 212 and 208) in themessaging system 100 via theinternal network 218. - By way of illustration, each publish request (e.g., in JSON key/value pairs) from a publisher to an MX node includes a channel name and a message. The MX node (e.g., MX node 202) can assign the message in the publish request to a distinct channel in the
messaging system 100 based on the channel name (e.g., “foo”) of the publish request. The MX node can confirm the assigned channel with thechannel manager 214. If the channel (specified in the subscribe request) does not yet exist in themessaging system 100, the channel manager can create and maintain a new channel in themessaging system 100. For instance, the channel manager can maintain a new channel by maintaining a list identifying each active streamlet of the channel's stream, the respective Q node on which the streamlet resides, and identification of the positions of the first and last messages in the streamlet as described earlier. - For messages of a particular channel, the MX node can store the messages in one or more buffers or streamlets in the
messaging system 100. For instance, theMX node 202 receives, from thepublishers 402, requests to publish messages M11, M12, M13, and M14 to a channel foo. TheMX node 206 receives, from thepublishers 404, requests to publish messages M78 and M79 to the channel foe. TheMX node 204 receives, from thepublishers 406, requests to publish messages M26, M27, M28, M29, M30, and M31 to the channel foo. - The MX nodes can identify one or more streamlets for storing messages for the channel foo. As described earlier, each MX node can request a write grant from the
channel manager 214 that allows the MX node to store the messages in a streamlet of the channel foo. For instance, theMX node 202 receives a grant from thechannel manager 214 to write messages M11, M12, M13, and M14 to astreamlet 4101 on theQ node 212. TheMX node 206 receives a grant from thechannel manager 214 to write messages M78 and M79 to thestreamlet 4101. Here, thestreamlet 4101 is the last streamlet of a sequence of streamlets of thechannel stream 430 storing messages of the channel foo. Thestreamlet 4101 hasmessages 421 of the channel foo that were previously stored in thestreamlet 4101, but is still open (e.g., thestreamlet 4101 still has space for storing more messages and the streamlet's TTL has not expired.) - The
MX node 202 can arrange the messages for the channel foo based on the respective time that each of the messages 422 was received by the MX node 202 (e.g., M11, M13, M14, M12) and store the received messages as arranged in thestreamlet 4101. That is, theMX node 202 receives M11 first, followed by M13, M14, and M12. Similarly, theMX node 206 can arrange the messages for the channel foo based on their respective time that each of the messages 423 was received by the MX node 206 (e.g., M78, M79) and store the received messages 423 as arranged in thestreamlet 4101. Other arrangements or ordering of the messages for the channel are possible. - The MX node 202 (or MX node 206) can store the received messages using the method for writing data to a streamlet described earlier in reference to
FIG. 3A , for example. In various implementations, the MX node 202 (or MX node 206) can buffer (e.g., in a local data buffer) the received messages for the channel foo and store the received messages in a streamlet for the channel foo (e.g., streamlet 4101) when the buffered messages reach a predetermined number or size (e.g., 100 messages) or when a predetermined time (e.g., 50 milliseconds) has elapsed. For example, theMX node 202 can store in thestreamlet 100 messages at a time or in 50 millisecond increments. Other acknowledgement scheduling algorithms, such as Nagle's algorithm, can be used. - In various implementations, the Q node 212 (e.g., a handler) stores the messages of the channel foo in the
streamlet 4101 in the order as arranged by theMX node 202 andMX node 206. TheQ node 212 stores the messages of the channel foo in thestreamlet 4101 in the order theQ node 212 receives the messages. For instance, assume that theQ node 212 receives message M78 (from the MX node 206) first, followed by messages M11 and M13 (from the MX node 202), M79 (from the MX node 206), and M14 and M12 (from the MX node 202). TheQ node 212 stores in thestreamlet 4101 the messages in the order as received (e.g., M78, M11, M13, M79, M14, and M12) immediately after themessages 421 that are already stored in thestreamlet 4101. In this way, messages published to the channel too from multiple publishers (e.g.,MX nodes 402, 404) can be serialized in a particular order and stored in thestreamlet 4101 of the channel foo. Different subscribers that subscribe to the channel foo will receive messages of the channel foo in the same particular order, as will he described in more detail in reference toFIG. 4B . - In the example of
FIG. 4A , at a time instance after the message M12 was stored in thestreamlet 4101, theNIX node 204 requests a grant from thechannel manager 214 to write to the channel foo. Thechannel manager 214 provides the MX node 204 a grant to write messages to thestreamlet 4101, as thestreamlet 4101 is still open for writing. TheMX node 204 arranges the messages for the channel foo based on the respective time that eachmessage 424 was received by the MX node 204 (e.g., M26, M27, M31, M29, M30, M28) and stores the messages as arranged for the channel foo. - By way of illustration, assume that the message M26 is stored to the last available position of the
streamlet 4101. As thestreamlet 4101 is now full, theQ node 212 sends to the MX node 204 a NAK message, following by an EOF message, to close the association with theMX node 204 for the write grant, as described earlier in reference toFIG. 3A . TheMX node 204 then requests another write grant from thechannel manager 214 for additional messages (e.g., M27, M31, and so on) for the channel too. - The
channel manager 214 can monitor available Q nodes in themessaging system 100 for the Q nodes respective wort loads (e.g., how many streamlets are residing in each Q node). Thechannel manager 214 can allocate a streamlet for the write request from theMX node 204 such that overloading (e.g., too many streamlets or too many read or write grants) can be avoided for any given Q node. For example, thechannel manager 214 can identify a least loaded Q node in themessaging system 100 and allocate a new streamlet on the least loaded Q node for write request's from theMX node 204. In the example ofFIG. 4A , thechannel manager 214 allocates anew streamlet 4102 on theQ node 208 and provides a write grant to theMX node 204 to write messages for the channel foo to thestreamlet 4102. As shown inFIG. 4A , theQ node 208 stores in thestreamlet 4102 the messages from theMX node 204 in an order as arranged by the MX node 204: M27, M31, M29, M30, and M28 (assuming that there is no other concurrent write grants for thestreamlet 4102 at the moment). - When the
channel manager 214 allocates a new streamlet (e.g., streamlet 4102) for a request for a grant from an MX node (e.g., MX node 204) to write to a channel (e.g., too), thechannel manager 214 assigns to the streamlet its TTL, which will expire after TTLs of other streamlets that are already in the channel's stream. For instance, thechannel manager 214 can assign to each streamlet of the channel foo's channel stream a TTL of 3 minutes when allocating the streamlet. That is, each streamlet will expire 3 minutes after it is allocated (created) by thechannel manager 214. Since a new streamlet is allocated after a previous streamlet is closed (e.g., filled entirely or expired), in this way, the channel foo's channel stream includes streamlets that each expire sequentially after the previous streamlet expires. For example, as shown in anexample channel stream 430 of the channel foo inFIG. 4A ,streamlet 4098 and streamlets before 4098 (e.g., streamlet 4097) have expired (as indicated by the dotted-lined gray-out boxes). Messages stored in these expired streamlets are not available for reading for subscribers of the channel foo.Streamlets streamlets streamlet 4102 is available for reading and writing, at the moment when the message M28 was stored in thestreamlet 4102. At a later time, thestreamlet 4099 will expire, following by thestreamlets -
FIG. 4B is a data flow diagram of anexample method 450 for subscribing to a channel of a messaging system. InFIG. 4B , asubscriber 480 establishes aconnection 462 with anMX node 461 of themessaging system 100.Subscriber 482 establishes aconnection 463 with theMX node 461.Subscriber 485 establishes aconnection 467 with anMX node 468 of themessaging system 100. Here, theMX nodes channel manager 214 and one or more Q nodes in themessaging system 100 via theinternal network 218. - A subscriber (e.g., subscriber 480) can subscribe to the channel foo of the
messaging system 100 by establishing aconnection 462 and sending a request for subscribing to messages of the channel foo to an MX node (e.g., MX node 461). The request (e.g., in JSON key/value pairs) can include a channel name, such as, for example, “foo.” When receiving the subscribe request, theMX node 461 can send a request to thechannel manager 214 for a read grant for a streamlet in the channel foo's channel stream. - By way of illustration, assume that at the current moment the channel foo's
channel stream 431 includesactive streamlets FIG. 4B . Thestreamlets streamlet 4104 stores messages of the channel foo, including the last message stored at aposition 47731.Streamlets 4101 and streamlets before 4101 are invalid, as their respective TTLs have expired. Note that the messages M78, M11, M13, M79, M14, M12, and M26 stored in thestreamlet 4101, described earlier in reference toFIG. 4A , are no longer available for subscribers of the channel foo, since thestreamlet 4101 is no longer valid, as the TTL ofstreamlet 4101 has expired. As described earlier, each streamlet in the channel foo's channel stream has a TTL of 3 minutes, thus only messages (as stored in streamlets of the channel foo) that are published to the channel foo (i.e., stored into the channel's streamlets) no earlier than 3 minutes from the current time can be available for subscribers of the channel foo. - The
MX node 461 can request a read grant for all available messages in the channel foo, for example, when thesubscriber 480 is a new subscriber to the channel foo. Based on the request, thechannel manager 214 provides the MX node 461 a read grant to the streamlet 4102 (on the Q node 208) that is the earliest streamlet in the active streamlets of the channel foo (e.g., the first in the sequence of the active streamlets). TheMX node 461 can retrieve messages in thestreamlet 4102 from theQ node 208, using the method for reading data from a streamlet described earlier in reference toFIG. 3B , for example. Note that the messages retrieved from thestreamlet 4102 maintain the same order as stored in thestreamlet 4102. However, other arrangements or ordering of the messages in the streamlet are possible. In various implementations, when providing messages stored in thestreamlet 4102 to theMX node 461, theQ node 208 can buffer (e.g., in a local data buffer) the messages and send the messages to theMX node 461 when the buffer messages reach a predetermined number or size (e.g., 200 messages or a predetermined time (e.g., 50 milliseconds) has elapsed. For instance, theQ node 208 can send the channel foo's messages (from the streamlet 4102) to theMX node 461 200 messages at a time or in 50 millisecond increments. Other acknowledgement scheduling algorithms, such as Nile's algorithm, can be used. - After receiving the last message in the
streamlet 4102, theMX node 461 can send an acknowledgement to theQ node 208, and send to thechannel manager 214 another request (e.g., for a read grant) for the next streamlet in the channel stream of the channel foo. Based on the request, thechannel manager 214 provides the MX node 461 a read grant to the streamlet 4103 (on Q node 472) that logically follows thestreamlet 4102 in the sequence of active streamlets of the channel foo. TheMX node 461 can retrieve messages stored in thestreamlet 4103 using the method for reading data from a streamlet described earlier in reference toFIG. 3B , untilMX node 601 retrieves the last message stored in thestreamlet 4103. TheMX node 461 can send to thechannel manager 214 yet another request for a read grant for messages in the next streamlet 4104 (on Q node 474). After receiving the read grant, theMX node 461 retrieves message of the channel foo stored in thestreamlet 4104, until the last message at theposition 47731 is retrieved byMX node 461. Similarly, theMX node 468 can retrieve messages from thestreamlets FIG. 4B ), and provide the messages to thesubscriber 485. - The
MX node 461 can send the retrieved messages of the channel foo to the subscriber 480 (via the connection 462) while receiving the messages from theQ node MX node 461 can store the retrieved messages in a local buffer. In this way, the retrieved messages can be provided to another subscriber (e.g., subscriber 482) when the other subscriber subscribes to the channel too and requests the channel's messages. TheMX node 461 can remove messages stored in the local buffer that each has a time of publication that has exceeded a predetermined time period. For instance, theMX node 461 can remove messages stored in the local buffer with respective times of publication exceeding 3 minutes. In some implementations, the predetermined time period for keeping messages in the local buffer onM X node 461 can be the same as or similar to the time-to-live duration of a streamlet in the channel foo's channel stream, since at a given moment, messages retrieved from the charmers stream do not include the messages i r streamlets having respective time-to-lives that have already expired. - The messages retrieved from the
channel stream 431 and sent to the subscriber 480 (by the MX node 461) are arranged in the same order as the messages were stored in the channel stream, although other arrangements or ordering of the messages are possible. For example, messages published to the channel foo are serialized and stored in thestreamlet 4102 in a particular order (e.g., M27, M31, M29, M30, and so on), then stored subsequently in thestreamlet 4103 and thestreamlet 4104. TheMX node 461 retrieves messages from thechannel stream 431 and provides the retrieved messages to thesubscriber 480 in the same order as the messages are stored in the channel stream: M27, M31, M29, M30, and so on, followed by ordered messages in thestreamlet 4103, and followed by ordered messages in thestreamlet 4104. - Instead of retrieving all available messages in the
channel stream 431, theMX node 461 can request a read grant for messages stored in thechannel stream 431 starting from a message at particular position (e.g.,position 47202.) For example, theposition 47202 can correspond to an earlier time instance (e.g., 10 seconds before the current time) when thesubscriber 480 was last subscribing to the channel foo (e.g., via a connection to theMX node 461 or another MX node of the messaging system 100). TheMX node 461 can send a request to thechannel manager 214 for a read grant for messages starting at theposition 47202. Based on the request, thechannel manner 214 provides the MX node 461 a read grant to the streamlet 4104 (on the Q node 474) and a position on thestreamlet 4104 that corresponds to thechannel stream position 47202. TheMX node 461 can retrieve messages in thestreamlet 4104 starting from the provided position, and send the retrieved messages to thesubscriber 480. - As described above in reference to
FIGS. 4A and 413 , messages published to the channel foo are serialized and stored in the channel's streamlets in a particular order. Thechannel manager 214 maintains the ordered sequence of streamlets as they are created throughout their respective time-to-lives. Messages retrieved from the streamlets by an MX node (e.g.,MX node 461 and/or MX node 468) and provided to a subscriber can be, in some implementations, in the same order as the messages are stored in the ordered sequence of streamlets. In this way, messages sent to different subscribers (e.g.,subscriber 480,subscriber 482, or subscriber 485) can be in the same order as the messages are stored in the streamlets, regardless which MX nodes the subscribers are connected to. - In various implementations, a streamlet stores messages in a set of blocks of messages. Each block stores a number of messages. For instance, a block can store two hundred kilobytes of messages. Each block has its own time-to-live, which can be shorter than the time-to-live of the streamlet holding the block. Once a block's TTL has expired, the block can be discarded from the streamlet holding the block, as described in more detail below in reference to
FIG. 4C . -
FIG. 4C is anexample data structure 490 for storing messages of a channel of a messaging system. As described with the channel foo in reference toFIGS. 4A and 4B , assume that at the current moment the channel foo'schannel stream 432 includesactive streamlets Streamlet 4103 and streamlets before streamlet 4103 (e,g.,streamlets 4101 and 4102) are invalid, as their respective TTLs have expired. Thestreamlet 4104 has reach maximum capacity (e.g., as determined by a corresponding write grant) and is closed for additional message writes. Thestreamlet 4104 is available for message reads. Thestreamlet 4105 is open and is available for message writes and reads. - By way of illustration, the streamlet 4104 (e.g., a computing process running on the
Q node 474 shown inFIG. 4B ) currently holds two blocks of messages.Block 494 bolds messages from channel positions 47301 to 47850.Block 495 holds messages from channel positions 47851 to 48000. The streamlet 4105 (e.g., a computing process running on another Q node in the messaging system 100) currently holds two blocks of messages.Block 496 holds messages from channel positions 48001 to 48200.Block 497 holds messages starting fromchannel position 48201, and still accepts additional messages of the channel foo. - When the
streamlet 4104 was created (e.g., by a write grant), block 492 was created to store messages from channel positions 47010 to 47100. Later on, after theblock 492 had reached its capacity, anotherblock 493 was created to store messages (e.g., from channel positions 47111 to 47300.)Blocks streamlet 4104 was closed for additional message writes, and thestreamlet 4105 was created with additional blocks for storing additional messages of the channel foo. - In this example, the respective TTL's of
blocks streamlet 4104 can discard these two expired blocks. For example, thestreamlet 4104 can de-allocate the memory space for theblocks blocks streamlet 4104, before thestreamlet 4104 itself becomes invalid. Alternatively,streamlet 4104 itself could become invalid before theblocks - A streamlet, or a computing process running on a (node in the
messaging system 100, can create a block for storing messages of a channel by allocating a certain size of memory space from the Q node. The streamlet can receive, from an MX node in themessaging system 100, one message at a time and store the received message in the block. Alternatively, the MX node can assemble (i.e., buffer) a group of messages and send the group of messages to the Q node. The streamlet can allocate a block of memory space (from the Q node) and store the group of messages in the block. The MX node can also perform compression on the group of messages, e.g., by removing a common header from each message or performing other suitable compression techniques. - As described earlier, an MX node can retrieve messages from Q nodes, store (cache) the retrieved messages in a local buffer, and provide the retrieved messages to multiple subscribers from the local buffer. In this way, the MX node can support a large number (e.g., thousands or more) of subscribers (e.g., of a particular channel), without creating a network bottleneck at Q nodes where messages are stored (in streamlets). The
messaging system 100 can further scale up the number of subscribers by adding additional MX nodes that each can cache messages and support additional subscribers. - When an MX node provides a locally cached message to multiple subscribers, a sender process running on the MX node can send copies of the cached message to destination processes (running on the MX node) that each corresponds to a connection to one of the subscribers. However, with a large number (e.g., thousands) of subscribers, copies of the cached message (e.g., created by the sender process for different destination processes) can cause a high demand of memory usage on the MX node. If the sender process sends the cached message sequentially to different destination processes, blocking (for synchronization) between the destination processes can cause computing resource contention on the MX node.
- Particular implementations described herein utilize a delivery buffer (i.e., a “fast buffer”) in an MX Node that can be used to cache messages retrieved from a Q node. The fast buffer includes data blocks for storing the retrieved messages. As described further below, each data block has a respective TTL that is longer than a read operation time associated with reading data from the fast buffer. This may prevent expiration of the data block while data is being read from the buffer and/or sent to subscribers. In certain instances, the TTL of a data block is longer than a corresponding streamlet's TTL. A process (running on the MX node) for a subscriber can read messages stored in one or more of the data blocks of the fast buffer, depending on the subscriber's connection status to the MX node. Processes for multiple subscribers can read messages from a single shared data structure (the fast buffer), without creating multiple copies for each message. In preferred implementations, multiple MX nodes are able to read from the fast buffer (e.g., the same data) simultaneously, without blocking or interfering with each other and without modifying the data in the fast buffer. For example, a first read process on the MX node can read messages from the fast buffer without blocking (e.g., causing suspension of) a second read process that is on the same MX node but may reside on a different processor core of the MX node.
-
FIG. 5 is anexample system 500 in which anMX node 502 retrieves messages from aQ node 504 and provides the messages to aclient device 506 of a subscriber. TheMX node 502 can include anMX subscribe process 508 and afast buffer 510. The MX subscribeprocess 508 can use aconnection 512 to read messages stored on the Q node 504 (e.g., in a streamlet) and can use aconnection 514 to provide the messages to thefast buffer 510 for storage. Thefast buffer 510 may store the messages in one or more blocks, with each block having a TTL. In preferred implementations, thefast buffer 510 can be a shared, lock-free structure that holds the data blocks and expires old data periodically (e.g., when the old data is no longer being read by MX processes or requested by subscribers or after a specified or predetermined time period). - In certain examples, the
MX node 502 can also include anMX connect process 516 and anMX send process 518 for establishing communications with one or more client devices, such asclient device 506. In the depicted example, theclient device 506 connects to theMX connect process 516 using aconnection 520 and requests messages from theMX node 502 for one or more channels. Theclient device 506 can request, for example, messages of the channel foo starting at a particular position, such as theposition 47202. An example JSON version of this specific request is as follows: -
{ “action”: “pubsub/subscribe”, “body” { “channel”: “foo”, “position”: 47202 } } - In response to the request, the
MX connect process 516 sends aninstruction 524 to the MX sendprocess 518 to retrieve message data stored in the fast bullet 510 (e.g., starting at theposition 47202 of the channel fob). The MX sendprocess 518 then retrieves the message data from thefast buffer 510 using a connection 526 and sends the messages to theclient device 506 through aconnection 522. - The MX subscribe
process 508 and thefast buffer 510 are preferably instantiated on a per-channel basis and/or devoted to retrieving and storing messages for a single channel. Additional MX subscribe processes and/or Fast buffers may be included in theMX node 502 to retrieve and store messages for additional channels. In alternative implementations, the MX subscribeprocess 508 and thefast buffer 510 can be configured to retrieve and store messages for multiple channels. For example, theMX node 502 may include one MX subscribeprocess 508 and onefast buffer 510 for retrieving and storing messages for the multiple channels. - Likewise, the
MX connect process 516 and the MX sendprocess 518 are preferably instantiated on a per-client connection basis and/or devoted to establishing communications with a single client device (e.g., the client device 506). Additional MX connect processes and MX send processes may be included on theMX node 502 to establish communications with additional client devices and retrieve and send messages from one or more fast buffers. In alternative implementations, theMX connect process 516 and MX sendprocess 518 can be configured to connect with and serve multiple client devices. For example, theMX node 502 may include oneMX connect process 516 and oneMX send process 518 for receiving requests from client devices (including client device 506) and sending messages to the client devices (e.g., for one or more channels) respectively. - In general, the blocks of data delivered to the MX subscribe
process 508 from theQ node 504 can be stored in thefast buffer 510 in the same order as the blocks were received by thefist buffer 510, although other arrangements or ordering of the data blocks are possible. The blocks are preferably given an offset and a timestamp. This allows the blocks to be located and retrieved from thefast buffer 510 and/or allows the MX subscribeprocess 508 to identify any blocks that are older than their TTL and/or otherwise unavailable for reading. Offsets are generally assigned in ascending order, with blocks added one-by-one to a back portion of a single queue or linked list (e.g., a forward list) in thefast buffer 510. - In certain examples, when accessing or reviewing blocks in the
first buffer 510, the MX subscribeprocess 508 may categorize each block as being, for example, active, inaccessible, or expired, although other designations or categorizations are possible. In general, active blocks are “young” blocks whose TTLs have not expired. Active blocks reside in a visible or otherwise accessible area of thefast buffer 510 and may be accessed with new read requests. Inaccessible blocks are “middle-aged” blocks whose TTLs are about to expire. Inaccessible blocks reside in an invisible or otherwise restricted or no access area of thefast buffer 510 and may be accessed only by read processes that have an established connection (e.g., a previously granted read request). If the TTL of an inaccessible block expires during a read process, the block becomes expired, but the read process may be given additional time to finish reading data from the block. Expired blocks reside in the invisible area of thefast buffer 510 and are deleted periodically from thefast buffer 510. - For example, when the MX subscribe
process 508 determines that one or more blocks are expired, the MX subscribeprocess 508 may shrink the queue in thefast buffer 510. For example, the MX subscribeprocess 508 may shrink a head, a tail, or other suitable portion of thePast buffer 510, so that the expired blocks are removed from the queue and are no longer accessible. In some implementations, however, expired blocks may remain available for reading for additional time to accommodate certain processes that were unable to read messages from the blocks within their TTLs. For example, client devices (and associated MX send processes) having a current or pre-existing connection to the MX node 502 (e.g., a pre-existing read request) may be allowed to read data from inaccessible and/or expired blocks. On the other hand, client devices (and associated MX send processes) having a new connection to the MX node 502 (e.g., with no pre-existing read request) may not be allowed to read data from inaccessible and/or expired blocks. Such client devices with new connections may be limited to reading only active blocks that are presently in a visible or accessible portion of the queue. This approach gives preference to pre-existing connections or pre-existing read requests and allows the corresponding client devices to avoid any gaps in received message data. Such gaps may occur, for example, due to slow connection speeds and/or temporary lapses in a connection. On the other hand, new connections generally do not need to access the older blocks to fill in any gaps, because the new connections have not yet read any message data and can start by reading the unexpired blocks. In various examples, a “pre-existing connection” refers to a connection with which an MX send process has requested and read messages from thefast buffer 510. In various examples, a “new connection” refers to a connection with which an MX send process has not yet read any messages from thefast buffer 510. - Advantageously, this approach of categorizing blocks and periodically deleting expired blocks provides improved control over memory usage. The amount of memory used by channel data can be dynamically allocated and/or can change over time according to demand. For example, as message activity for a channel decreases, the amount of memory used for the channel can automatically decrease, as expired blocks are deleted and fewer new blocks are created. Also, expired blocks can be deleted after existing read processes have had a chance to read the message data and the message data in the blocks is no longer needed. The approach also provides flexibility to add one or more processing steps (e.g., encryption, compression, etc.), and ensures that such steps are not processing or attempting to access memory that has been freed.
- In preferred implementations, the M X subscribe
process 508 can be the only process having authority to modify the structure of thefast buffer 510. For example, the MX subscribeprocess 508 may be configured and allowed to add blocks to thefast buffer 510, remove blocks from thefast buffer 510, and/or modify blocks on thefast buffer 510, as needed. The MX subscribeprocess 508 may also move pointers that identify a head portion and a tail portion of a queue or linked list in thefast buffer 510. Other processes or MIX node components are preferably not allowed to perform such modifications. By giving such authority only to the MX subscribeprocess 508, the need for any locks in the modification operations is eliminated. - In certain examples, the MX send
process 518 can check the front (i.e., head) and back (i.e., tail) of thefast buffer 510 at the beginning of every read operation. When the MX sendprocess 518 is requesting to read a target block, and the target block resides in an accessible or visible region of the fast buffer 510 (e.g., because the target block is an active block), the MX sendprocess 518 may read the block from thefast buffer 510 and forward the message data in the block to theclient device 506. When the MX sendprocess 518 is requesting to read data from more than one block in the accessible region of thefast buffer 510, the MX sendprocess 518 may read an amount of data (e.g., up to 64 Kb) from the fast buffer 510 (e.g., starting from a block having a requested offset or location in the fast buffer 510) and forward the message data to theclient device 506. In certain instances, when the MX sendprocess 518 detects that a target data block is not accessible in the fast buffer 510 (e.g., because the target block has expired), the MX sendprocess 518 may apply a separate logic that involves error reporting and/or waiting. The separate logic used by the MX sendprocess 518 may depend on where the target block resides in comparison to accessible regions of thefast buffer 510. - For example, in some instances, when a client connection is too slow to remain in sync with fist moving channel data, the client may be unsubscribed from the channel and a message may be sent to inform that client that it is “out of sync” with the channel's data. Alternatively, rather than sending the “out of sync” message to the client, the client may be fast forwarded to the oldest data available (e.g., the oldest active block) or to the newest data available (e.g., the newest active block). In certain instances, the systems and methods may identify one or more key active blocks (e.g., as indicated by one or more flags), forward the data from the key active blocks to the client, and then fast forward the client to the newest active block for further reading. The client may have certain preferences that define how error reporting and/or waiting are to be performed. For example, the client may have a preference that causes its state to be reloaded completely (e.g., when the client loses a connection and is able to reconnect).
- In various examples, the MX send process 51.8 can read the accessible blocks in the
fast buffer 510 and then can wait to be notified when additional blocks are added to thefast buffer 510 and available for reading. For example, when the MX sendprocess 518 reaches a tail of thefast buffer 510, the MX sendprocess 518 may turn on or activate anotifier 528 and/or yield to other processes. Thenotifier 528 can be configured to wake up the MX sendprocess 518 upon the arrival of new data into thefast buffer 510. To activate thenotifier 528, the MX sendprocess 518 may send its process identifier (PID) or other suitable identifier to awaiting list 530 along a connection 532. The MX sendprocess 518 may then perform a second check of thefast buffer 510 for new data and, if no new data is present, the MX sendprocess 518 may go to sleep and wait for a message from thenotifier 528. Checking thefast buffer 510 again after providing the PID to thewaiting list 530 may allow the MX sendprocess 518 to identify the presence of any new data that arrived just before or at around the same time as the PID was added to thewaiting list 530. In general, thewaiting list 530 includes a list of PIDs for one or more MX send processes that are waiting to be notified when new data arrives in thefast buffer 510. Thewaiting list 530 may provide PIDs to thenotifier 528 along aconnection 534, and thenotifier 528 may wake up or notify the MX send processes in thewaiting list 530 when new data arrives in thefast buffer 510 and is available for reading. For example, thenotifier 528 may wake up the MX sendprocess 518 by sending a notification along aconnection 536. Like the MX subscribeprocess 508 and thefast buffer 510, thewaiting list 530 and thenotifier 528 are preferably configured to perform operations for a single channel. TheMX node 502 may include additional waiting lists and notifiers to perform operations for additional channels, in alternative implementations, thewaiting list 530 and thenotifier 528 can be configured to perform operations for multiple channels. For example, theMX node 502 may include onewaiting list 530 and onenotifier 528 for receiving PIDs of MX send processes and waking up the corresponding MX send processes, respectively, when new data arrives in thefast buffer 510 for the multiple channels. - In preferred examples, the
notifier 528 can work together and/or can be synchronized with the MX subscribeprocess 508. When a new block of data is added to thefast butler 510, thenotifier 528 may receive a list of PIDs from thewaiting list 530 for the one or more MX send processes waiting to be notified. The notifier 52$ then delivers a notification to the one or more MX send processes. The notification preferably includes the offset or location of the new block in thefast buffer 510. After sending the notifications, the PIDs for the one or more MX send processes may be removed from thewaiting list 530. - In certain implementations, to keep the latency low, the
notifier 528 can use a two-level model of synchronization with the M X subscribeprocess 508. In aninner loop 538, thenotifier 528 may contact the MX subscribeprocess 508 at short time intervals (e.g., every 10 ms) to determine if any new blocks have been added to thefast buffer 510. Thenotifier 528 may, for example, read from a dictionary or log of the MX subscribeprocess 508 to determine the offset of the block most recently added to thefast buffer 510. If the most recent block has not changed for more than a threshold amount of time (e.g., more than 200 ms), thenotifier 528 may switch to anouter loop 540 where thenotifier 528 goes to sleep and/or waits for a notification from the MX subscribeprocess 508. In theouter loop 540, the MX subscribeprocess 508 preferably checks the state of thenotifier 528 and wakes up thenotifier 528 by sending a notification when thenotifier 528 is sleeping and new data has been sent to thefast buffer 510. In general, since checking the state of thenotifier 528 is a significantly faster operation than delivering the notification, this approach allows the MX subscribeprocess 508 to look for new data on the Q node more frequently and to add any such new data to thefast buffer 510. When the MX subscribeprocess 508 informs thenotifier 528 that a new block has been added to thefast buffer 510, thenotifier 528 may switch back to theinner loop 538. The two-loop approach allows theMX node 502 to conserve resources and operate more efficiently. Thenotifier 528 actively looks for new message data only when message data is being added to the fast buffer frequently. Otherwise, thenotifier 528 is able to sleep and is woken up by the MX subscribeprocess 508 when new data arrives. A primary benefit of the two-loop approach is that resources are used only when there is work to do or when work is expected in the near future. If new channel data is not expected soon, the system switches from theinner loop 538 to the use of push notifications in theouter loop 540. The two-loop approach achieves a good balance between latency and central processing unit (CPU) overhead. Theouter loop 540 allows the system to maintain millions of channels that are rarely updated and not waste a single CPU cycle on checking the status of such channels. At the same time, theinner loop 538 allows the system to maintain with minimal latency other channels that are updated more frequently. - In certain implementations, when the
notifier 528 becomes aware of new data added to thefast buffer 510, thenotifier 528 sends a notification to any MX send processes waiting for the new data. For example, thenotifier 528 may obtain from thewaiting list 530 the PIDs of MX send processes that are sleeping and waiting for new data and may send notifications to those MX send processes. The notifications preferably include the offset or other location information for any new blocks of data in thefast butler 510. - In various examples, the systems and methods allow publishers to be prioritized (or deprioritized) with respect to subscribers. For example, the
fast buffer 510 can be considered to (i) receive data from publishers in a publishing loop and (ii) provide data to subscribers in a subscribing loop. The publishing loop can include, for example, a step of obtaining new message data from one or more publishers and a step of writing the new message data to thefast buffer 510. The subscribing loop can include, for example, a step of checking thefast buffer 510 for new data and a step of providing the new data to one or more subscribers (e.g., using the MX send process 518). In preferred examples, the publishing loop and the subscribing loop are loosely coupled or not coupled, such that publishing and subscribing speeds can be independently controlled and/or prioritized. This allows a publishing speed to be different from a subscribing speed, if desired, and/or for each speed to be chosen or prioritized, for example, according to system load. To promote publishing activity, for example, the steps of the publishing loop can be performed taster, so that message data can be written more frequently to thefast buffer 510. Alternatively or additionally, to promote subscribing activity, the steps of the subscribing loop can be performed faster, so that message data can be read more frequently from thefast buffer 510. In general, when the steps of one loop are performed faster, the steps of the other loop can be performed slower, for example, to maintain overall system load at or near a constant value and/or to avoid excessive system load. Such prioritization can be achieved, for example, by creating two worker pools (e.g., one pool for publishers and one pool for subscribers) and using an operating system kernel scheduler prioritization mechanism to set different priorities to threads in the two pools. Other approaches for prioritizing or deprioritzing publishers and subscribers are contemplated. -
FIG. 6 shows example data blocks of thefast butler 510 at different instances in time. At afirst time 650, thefast buffer 510 includes a sequence of data blocks 601, 602, . . . , 608, that were retrieved from one or more streamlets in theQ node 504. The sequence of message data in the data blocks is preferably consistent with or the same as the sequence of message data from the one or more streamlets in theQ node 504. - Each data block 601, 602, . . . . , 608 in the
fast butler 510 is given a respective TTL, in preferred implementations, the TTLs for the blocks are sufficient in duration to give read processes enough time to read data from the blocks. In some instances, the TTLs for one or more blocks are longer than the TTLs for one or more streamlets in theQ node 504. In general, the TTLs for the blocks are chosen to give the MX send processes adequate time to read from thefast buffer 510. In one example, the TTL for the blocks in thefast buffer 510 is about 30 seconds or about 60 seconds, while the TTL for the streamlets in theQ node 504 is about 10 seconds, although other TTLs for the blocks and the streamlets are possible. This way, the fast buffer 510 (on the MX node 502) can hold more messages (e.g., messages for the past 30 seconds) than the streamlets can hold (e.g., messages for the past 10 seconds) for the particular channel. TheMX node 502 can thus provide a longer history of messages to client devices of subscribers (e.g., client device 506). This is beneficial, for example, when a client device has a slower connection and/or otherwise takes longer (e.g., exceeding to seconds) to read messages that may no longer be accessible or stored in streamlets of theQ node 504. - In various examples, there can be three global always-accessible pointers inside the last buffer 510: a
tail pointer 642 identifying a rightmost block of a visible area 662 (i.e., block 608 at time 650), ahead pointer 640 identifying a leftmost block of the visible area 662 (i.e., block 603 at time 650), and aprevious head pointer 644 identifying a leftmost block of an invisible area 660 (i.e., block 601 at time 650). All blocks referenced by pointers represent a single linked list (e.g., a forward list) so all such pointers can be reached by iterating from a first block in the list (e.g., block 601 at time 650) using the block's next pointer. In certain examples, thevisible area 662 includes blocks that are active and theinvisible area 660 includes blocks that are inaccessible and/or expired. - In the sequence of data blocks of the
fast buffer 510, data blocks closer to thehead pointer 640 of thefast buffer 510 have been in thefast buffer 510 longer than blocks closer to thetail pointer 642 of thefast buffer 510. Accordingly, blocks near thehead pointer 640 will generally reach their respective TTLs and expire before blocks near thetail pointer 642 reach their TTLs and expire. For instance, at thefirst time 650, block 603 will expire beforeblock 604, which will expire beforeblock 605, and so on. In general, messages in thefast buffer 510 are arranged in the same order as the messages were arranged in theQ node 504, although other arrangements or ordering of the messages are possible. Accordingly, the messages may expire in thefast buffer 510 in the same order that the messages expire in theQ node 504. - In preferred implementations, the sequence of data blocks of the
fast buffer 510 can be arranged as a linkedlist 620. In the linkedlist 620, each data block has a pointer pointing to the next data block in the sequence of the list. For instance, data block 604 has a pointer pointing to data block 605, which has a pointer pointing to data block 606.Data block 608, located at a tail end of the linkedlist 620, may have a pointer (e.g., NUL) indicating that it is at the tail end of the linkedlist 620. In one implementation, the linkedlist 620 includes thehead pointer 640 identifying the block at the head of the linkedlist 620. Attime 650, for example, the linkedlist 620 may have ahead pointer 640 identifying data block 603, which has not reached its TTL but will be the next block in the linkedlist 620 to expire. - In certain instances, when the MX subscribe
process 508 adds data to thefast buffer 510, the MX subscribeprocess 508 can do two things. First, the MX subscribeprocess 508 can determine the age of the oldest block(s) in the visible area 662 (i.e., block 603 at time 650), to determine whether the oldest block(s) should be moved to theinvisible area 660. If one or more blocks in thevisible area 662 are inaccessible (i.e., because their TTLs will expire soon) or have expired, the MX subscribeprocess 508 may move the one or more blocks to theinvisible area 660 or delete the one or more blocks from thefast buffer 510. In one example, blocks that are inaccessible are moved to theinvisible area 660 and blocks that have expired are deleted. The process of deleting or removing one or more blocks from thefast buffer 510 may be referred to herein as “shrinking”. Next, the MX subscribeprocess 508 can add one or more blocks of the data to a tail of thefast buffer 510. - In preferred examples, the shrinking and adding processes are done without the use of locks or atomic operations, and resulting changes are preferably made in a conflict free order. In general, locks or atomic operations are a significant source of poor performance on symmetric multiprocessing (i.e., SMP or multi core) systems. Every such operation must ensure that core or non-uniform memory access (NUMA) node caches do not have a cached value that can be read right after a change was made. Locks and/or atomic operations are expensive on modern CPUs and can cost hundreds of CPU cycles. By relying on time, however, the
fast buffer 510 performs truly in parallel, such that multiple read processes can read from the same memory simultaneously. With the fast buffer approach described herein, preferably nothing other than boundary pointers requires a synchronization during the shrinking and adding processes. - Referring to
times FIG. 6 , shrinking is preferably done with the following technique. First, expired blacks in theinvisible area 660 are deleted. In some examples, expired or inaccessible blocks in the invisible area 560 that are younger than some threshold time (e.g., TTL*1.5) are retained in theinvisible area 660. Theprevious head pointer 644 is then moved to the current location of thehead pointer 640, and thehead pointer 640 is moved toward thetail pointer 642 until a block is found that is active e.g., block 605 at time 652). Thehead pointer 640 is placed at the front edge of that first active block. If no active blocks are found, thehead pointer 640 and thetail pointer 642 may mark the same location, indicating that thefast buffer 510 includes no active blocks. - The shrinking process is further illustrated at
times fast buffer 510 inFIG. 6 . Attime 652, thehead pointer 640 andprevious head pointer 644 have been moved, such that expiredblocks invisible area 660, and theinvisible area 660 now includesblocks block 611. Likewise, attime 654, thehead painter 640 and theprevious head pointer 644 have been moved, such that expiredblocks invisible area 660. At this time, theinvisible area 660 includes no blocks and thehead pointer 640 and theprevious head pointer 644 mark the same location (i.e., block 605 at time 654). Each block in thefast buffer 510 at time 554 is active and resides in thevisible area 662. Further, attime 658, thehead pointer 640 and theprevious head pointer 644 have been moved, such that expiredblocks invisible area 660 at that time includesblocks previous head pointer 644 has been moved to identify the leftmost block of the invisible area 660 (i.e., block 607 at time 658). Thetail pointer 642 attime 658 has been moved to define a new tail atblock 615, while thehead pointer 640 attime 658 has been moved to define a new head atblock 609. - In general, because there are typically many MX send processes reading from the
fast buffer 510 in parallel, shrinking thefast buffer 510 and repositioning the pointers preferably occurs while certain MX send processes continue to have access to blocks in the newinvisible area 660. The invisible area 660 (e.g., includingblocks visible area 662 to theinvisible area 660. - On the other hand, MX send processes that are new and/or have new read requests may be given access only to the active blocks in the
visible area 662 of thefast buffer 510. In preferred examples, each MX send process (e.g., MX send process 518) having a new read request checks the boundaries of thevisible area 662 and if the process requests access outside of the visible area 662 (e.g., to any blocks between the previous head and the head), the request fails as if there were no data available for the request. In general, MX send processes having new requests are not allowed to access any inaccessible or expired blocks in theinvisible area 660. In preferred examples, theinvisible area 660 can be accessed only by MX send processes having old requests, and the MX send processes preferably finish the read operation from theinvisible area 660 as soon as possible. Typical reading times take a few microseconds or milliseconds, because there are preferably no locks in the reading code. - In general, blocks reside in the
visible area 662 for approximately one TTL and are then moved to theinvisible area 660 where the blocks reside for some fraction (e.g., 0.25, 0.5, or 0.75) of the TTL. For example, after spending, approximately one TTL in thevisible area 662, a block may reside in theinvisible area 660 for an additional TTL*0.5. This time in theinvisible area 660 is generally many orders of magnitude longer than a typical access time and provides pre-existing MX send processes with additional time to obtain the requested message data. As discussed previously, a block resides in thevisible area 662 when the block is active and is moved to theinvisible area 660 when the block is inaccessible or expired. - In one example, to add blocks to the
fast buffer 510, new blocks are combined into a separate single linkedlist 620 that is linked to the block identified by thecurrent tail pointer 642. Thetail pointer 642 is then moved to the last added block. This block addition process is illustrated inFIG. 6 attimes fast buffer 510, compared to the previous time, and thetail pointer 642 has been moved to identify the last added block. For example, attime 652,new blocks tail pointer 642 points to the right-hand edge ofblock 611, indicating thatblock 611 is the last added block. Likewise, attime 656,new blocks tail pointer 642 indicatesblock 613 is the last added block. Further, attime 658, blocks 614 and 615 have been added, and thetail pointer 642 indicatesblock 615 is the last added block. - In certain instances, a data block contains zero messages and may serve as a gap block or a left boundary marker. Gap blocks may he used when there is a gap in block positions in the fast buffer 510 (e.g., when a previous block position plus a previous block length does not equal a next block position). Such gaps can be due to, for example, switching from one streamlet to another streamlet streamlets that are 16m in size can be aligned by a 16 m boundary) or missing data, in which case clients reaching the gap in the buffer may need to be informed about anther provided with the missing data. A left boundary marker may be a zero message block or marker that indicates the leftmost available location in the
fast buffer 510 where data was first written or intended to be written. The zero message block may contain the leftmost offset even when no data is available yet. In such instances, the left boundary matter can help theMX node 502 make decisions about certain read requests and whether it will be possible to provide a client device with the requested data. Having the left boundary defined for every read request may make it possible to determine whether the read request can be satisfied in the future (e.g., when data is received) or not. - In general, the fast buffers described herein are internal components inside a distributed PubSub system. While fast buffers can be included in or used by MX nodes, as shown in
FIG. 5 , fast buffers can also be included in or used by other system components, including Q nodes. A fast buffer publisher is, in general, a component that writes data to a fast buffer, and a fast buffer subscriber is a component that reads data from a fast buffer. As shown inFIG. 5 , for example, the MX subscribeprocess 508 is a fast buffer publisher and the MX sendprocess 518 is a fast buffer subscriber, for thefast buffer 510 inMX node 502. Similar fast buffer publishers and fast buffer subscribers can be used for Q nodes. In one example, a Q node fast buffer publisher is referred to as a Q streamlet process and a Q node fast buffer subscriber is referred to as a Q sender process. Fast butlers incorporated into or used by Q nodes can use the notifier and/or waiting list approach described herein with respect toMX node 502. For example, a Q sender process waiting to retrieve new message data from a Q node fast buffer can be put to sleep when no data is available and/or woken up when new message data arrives. -
FIG. 7 is a flowchart ofexample method 700 for providing messages to a plurality of subscribers. The method includes receiving (step 702) from a plurality of publishers a plurality of messages. Each of the messages is assigned (step 704) to one of a plurality of distinct channels wherein each channel includes an ordered plurality of messages. The messages of each of the channels are stored (step 706) in one or more respective storage butlers according to the order. Each storage buffer has a respective time-to-live of a first time duration. For a particular channel, messages of the particular channel are retrieved (step 708) from respective storage buffers (e.g., those having times-to-live that have not expired and according to the order). The retrieved messages of the particular channel are stored (step 710) in a delivery buffer according to the order. The delivery buffer includes one or more data blocks, and each data block has a respective time-to-live of a second time duration (e.g., that is longer than the first time duration). The messages stored in the delivery buffer are provided (step 712) to a plurality of subscribers of the particular channel. - Examples of the systems and methods described are able to send a stream of data from one source (e.g., an ERLANG process) to multiple destinations (e.g., ERLANG processes). The stream of data is delivered at a high rate (e.g., in messages per second) and with low latency (e.g., 99% of messages are delivered with a latency less than 50 ms.
- A first alternative approach involves sending messages sequentially from a sender process to all the destination processes, which may be represented as a list or other data structure. This process is generally slow and measured to be a bottleneck, for example, because locking on multicore processing can create a contention point around sending a message to processes running on other cores. The process also requires a lot of memory because every sending operation creates a copy of data to be sent (or a copy of reference to the data). The copy typically resides in the destination process message queue until the destination process handles the data. Using multiple senders does not solve the problem.
- A second alternative approach is to use a table storage (e.g., an ERLANG ETS) as an intermediary single-copy storage of the stream data and let every destination process read the data from various places at the same time. This approach is also slow, however, because access to the table storage presents a bottleneck. For example, ERLANG ETS tables utilize a set of locks to protect ranges of keys from multiple accesses and order these accesses. There is also a step of determining a location for data in the ITS table, given that every call to the ETS table is done with an ETS table identifier. Both of these locks are expensive and make it difficult to exceed 10 million updates per second to a table, no matter how many keys and cores (on modern hardware) are used. If locks are eliminated, the number of accesses or processing tasks can be reduced by several orders of magnitude.
- A series of tests were performed to compare these first and second alternative approaches with the fast buffer approach described herein. The tests involved using all three approaches to send 100 byte messages from one publisher to multiple receivers on a 32-core server. The number of receivers was selected from 100, 1,000, or 10,000. The results of the tests are presented in Table 1, below, which includes the resulting number of messages per second each approach was able to send.
-
TABLE 1 Comparison of Fast Buffer approach with two alternative approaches. 1 Publisher to 1 Publisher to 1 Publisher to 100 Subscribers 1,000 Subscribers 10,000 Subscribers Published Received Published Received Published Received Rate Rate Rate Rate Rate Rate Approach (msgs/sec) (msgs/sec) (msgs/sec) (msgs/sec) (msgs/sec) (msgs/sec) First Alternative 400k 400k 380k 380k 350k 350k Approach - Publisher sends messages to subscribers Second Alternative 5000 0.5m 2200 2.2m 1280 5m Approach - Publisher writes messages to ETS, subscribers read messages from ETS Fast Buffer (FB) 160k 16m 160k 160m 83k 830m Approach - Publisher writes messages to FB, subscribers read messages from FB - In various implementations, rather than sending message data to multiple processes from a single process, the fast buffer approach described herein tasks the receivers or subscribers themselves to grab data from a single shared structure (i.e., a fast buffer). To avoid having locking issues around the shared structure, the process utilizes a “lock-free” or non-blocking approach. Receivers are able to grab many pieces of data without blocking or figuring out where to read from and up to which point. As the results in Table 1 indicate, the rates of message transfer are improved considerably when compared to the alternative approaches.
- Embodiments of the subject matter and the operations described in this specification can be implemented in digital electronic circuitry, or in computer software, firmware, or hardware, including the structures disclosed in this specification and their structural equivalents, or in combinations of one or more of them. Embodiments of the subject matter described in this specification can be implemented as one or more computer programs, i.e., one or more modules of computer program instructions, encoded on computer storage medium for execution by, or to control the operation of, data processing apparatus. Alternatively or in addition, the program instructions can be encoded on an artificially-generated propagated signal, e.g., a machine-generated electrical, optical, or electromagnetic signal, that is generated to encode information for transmission to suitable receiver apparatus for execution by a data processing apparatus. A computer storage medium can be, or be included in, a computer-readable storage device, a computer-readable storage substrate, a random or serial access memory array or device, or a combination of one or more of them. Moreover, while a computer storage medium is not a propagated signal, a computer storage medium can be a source or destination of computer program instructions encoded in an artificially-generated propagated signal. The computer storage medium can also be, or be included in, one or more separate physical components or media (e.g., multiple CDs, disks, or other storage devices).
- The operations described in this specification can be implemented as operations performed by a data processing apparatus on data stored on one or more computer-readable storage devices or received from other sources.
- The term “data processing apparatus” encompasses all kinds of apparatus, devices, and machines for processing data, including by way of example a programmable processor, a computer, a system on a chip, or multiple ones, or combinations, of the foregoing. The apparatus can include special purpose logic circuitry, e.g., an FPGA (field programmable gate array) or an ASIC (application-specific integrated circuit). The apparatus can also include, in addition to hardware, code that creates an execution environment for the computer program in question, e.g., code that constitutes processor firmware, a protocol stack, a database management system, an operating system, a cross-platform runtime environment, a virtual machine, or a combination of one or more of them. The apparatus and execution environment can realize various different computing model infrastructures, such as web services, distributed computing and grid computing infrastructures.
- A computer program (also known as a program, software, software application, script, or code) can be written in any form of programming language, including compiled or interpreted languages, declarative, procedural, or functional languages, and it can be deployed in any form, including as a stand-alone program or as a module, component, subroutine, object, or other unit suitable for use in a computing environment. A computer program may, but need not, correspond to a tile in a file system. A program can be stored in a portion of a file that holds other programs or data (e.g., one or more scripts stored in a markup language resource), in a single file dedicated to the program in question, or in multiple coordinated tiles (e.g., files that store one or more modules, sub-programs, or portions of code). A computer program can be deployed to be executed on one computer or on multiple computers that are located at one site or distributed across multiple sites and interconnected by a communication network.
- The processes and logic flows described in this specification can be performed by one or more programmable processors executing one or more computer programs to perform actions by operating on input data and generating output. The processes and logic flows can also be performed by, and apparatus can also be implemented as, special purpose logic circuitry, e.g., an FPGA (field programmable gate array) or an ASIC (application-specific integrated circuit).
- Processors suitable for the execution of a computer program include, by way of example, both general and special purpose microprocessors, and any one or more processors of any kind of digital computer. Generally, a processor will receive instructions and data from a read-only memory or a random access memory or both. The essential elements of a computer are a processor for performing actions in accordance with instructions and one or more memory devices for storing instructions and data. Generally, a computer will also include, or be operatively coupled to receive data from or transfer data to, or both, one or more mass storage devices fear storing data, e.g., magnetic disks, magneto-optical disks, optical disks, or solid state drives. However, a computer need not have such devices. Moreover, a computer can be embedded in another device, e.g., a smart phone, a mobile audio or video player, a game console, a Global Positioning System (GPS) receiver, or a portable storage device (e.g., a universal serial bus (USB) flash drive), to name just a few. Devices suitable for storing computer program instructions and data include all forms of non-volatile memory, media and memory devices, including, by way of example, semiconductor memory devices, e.g., EPROM, EEPROM, and flash memory devices; magnetic disks, e.g., internal hard disks or removable disks; magneto-optical disks; and CD-ROM and DVD-ROM disks. The processor and the memory can be supplemented by, or incorporated in, special purpose logic circuitry.
- To provide for interaction with a user, embodiments of the subject matter described in this specification can be implemented on a computer having a display device, e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor, for displaying information to the user and a keyboard and a pointing device, e.g., a mouse, a trackball, a touchpad, or a stylus, by which the user can provide input to the computer. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback, e.g., visual feedback, auditory feedback, or tactile feedback; and input from the user can be received in any form, including acoustic, speech, or tactile input. In addition, a computer can interact with a user by sending resources to and receiving resources from a device that is used by the user; for example, by sending web pages to a web browser on a user's client device in response to requests received from the web browser.
- Embodiments of the subject matter described in this specification can be implemented in a computing system that includes a back-end component, e.g., as a data server, or that includes a middleware component, e.g., an application server, or that includes a front-end component, e.g., a client computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the subject matter described in this specification, or any combination of one or more such back-end, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication, e.g., a communication network. Examples of communication networks include a local area network (“LAN”) and a wide area, network (“WAN”), an inter-network (e.g., the Internet), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks).
- The computing system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. In some embodiments, a server transmits data (e.g., an HTML page) to a client device (e.g., for purposes of displaying data to and receiving user input from a user interacting with the client device). Data generated at the client device (e.g., a result of the user interaction) can be received from the client device at the server.
- A system of one or more computers can be configured to perform particular operations or actions by virtue of having software, firmware, hardware, or a combination of them installed on the system that in operation causes or cause the system to perform the actions. One or more computer programs can be configured to perform particular operations or actions by virtue of including instructions that, when executed by data processing apparatus, cause the apparatus to perform the actions.
- While this specification contains many specific implementation details, these should not be construed as limitations on the scope of any inventions or of what may be claimed, but rather as descriptions of features specific to particular embodiments of particular inventions. Certain features that are described in this specification in the context of separate embodiments can also be implemented in combination in a single embodiment. Conversely, various features that are described in the context of a single embodiment can also be implemented in multiple embodiments separately or in any suitable subcombination. Moreover, although features may be described above as acting in certain combinations and even initially claimed as such, one or more features from a claimed combination can in some cases be excised from the combination, and the claimed combination may be directed to a subcombination or variation of a subcombination.
- Similarly, while operations are depicted in the drawings in a particular order, this should not be understood as requiring that such operations be preformed in the particular order shown or in sequential order, or that all illustrated operations he performed, to achieve desirable results. In certain circumstances, multitasking and parallel processing may be advantageous. Moreover, the separation of various system components in the embodiments described above should not be understood as requiring such separation in all embodiments, and it should be understood that the described program components and systems can generally be integrated together in a single software product or packaged into multiple software products.
- Thus, particular embodiments of the subject matter have been described. Other embodiments are within the scope of the following claims. In some cases, the actions recited in the claims can be performed in a different order and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In certain implementations, multitasking and parallel processing may be advantageous.
Claims (20)
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US16/506,890 US20190335010A1 (en) | 2016-09-23 | 2019-07-09 | Systems and methods for providing messages to multiple subscribers |
Applications Claiming Priority (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US15/274,281 US9667681B1 (en) | 2016-09-23 | 2016-09-23 | Systems and methods for providing messages to multiple subscribers |
US15/586,773 US9942339B1 (en) | 2016-09-23 | 2017-05-04 | Systems and methods for providing messages to multiple subscribers |
US15/909,710 US10382574B2 (en) | 2016-09-23 | 2018-03-01 | Systems and methods for providing messages to multiple subscribers |
US16/506,890 US20190335010A1 (en) | 2016-09-23 | 2019-07-09 | Systems and methods for providing messages to multiple subscribers |
Related Parent Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US15/909,710 Continuation US10382574B2 (en) | 2016-09-23 | 2018-03-01 | Systems and methods for providing messages to multiple subscribers |
Publications (1)
Publication Number | Publication Date |
---|---|
US20190335010A1 true US20190335010A1 (en) | 2019-10-31 |
Family
ID=58738071
Family Applications (4)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US15/274,281 Expired - Fee Related US9667681B1 (en) | 2016-09-23 | 2016-09-23 | Systems and methods for providing messages to multiple subscribers |
US15/586,773 Expired - Fee Related US9942339B1 (en) | 2016-09-23 | 2017-05-04 | Systems and methods for providing messages to multiple subscribers |
US15/909,710 Expired - Fee Related US10382574B2 (en) | 2016-09-23 | 2018-03-01 | Systems and methods for providing messages to multiple subscribers |
US16/506,890 Abandoned US20190335010A1 (en) | 2016-09-23 | 2019-07-09 | Systems and methods for providing messages to multiple subscribers |
Family Applications Before (3)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
US15/274,281 Expired - Fee Related US9667681B1 (en) | 2016-09-23 | 2016-09-23 | Systems and methods for providing messages to multiple subscribers |
US15/586,773 Expired - Fee Related US9942339B1 (en) | 2016-09-23 | 2017-05-04 | Systems and methods for providing messages to multiple subscribers |
US15/909,710 Expired - Fee Related US10382574B2 (en) | 2016-09-23 | 2018-03-01 | Systems and methods for providing messages to multiple subscribers |
Country Status (6)
Country | Link |
---|---|
US (4) | US9667681B1 (en) |
EP (1) | EP3516830A1 (en) |
JP (1) | JP2019535072A (en) |
CN (1) | CN110121863A (en) |
AU (1) | AU2017331066A1 (en) |
WO (1) | WO2018057066A1 (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US11429290B2 (en) | 2020-11-13 | 2022-08-30 | Argo AI, LLC | Methods and systems for providing a lockless access to a shared memory region in a publish and subscribe system |
US11709620B2 (en) * | 2020-11-13 | 2023-07-25 | Ford Global Technologies, Llc | Methods and systems for memory management in a publish and subscribe system |
Families Citing this family (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
GB2532490B (en) * | 2014-11-21 | 2017-02-22 | Ibm | Publish/subscribe messaging using message structure |
US9407585B1 (en) | 2015-08-07 | 2016-08-02 | Machine Zone, Inc. | Scalable, real-time messaging system |
US9319365B1 (en) * | 2015-10-09 | 2016-04-19 | Machine Zone, Inc. | Systems and methods for storing and transferring message data |
GB2560275B (en) * | 2015-10-15 | 2021-09-01 | Push Tech Limited | Event driven subscription matching |
US9602450B1 (en) | 2016-05-16 | 2017-03-21 | Machine Zone, Inc. | Maintaining persistence of a messaging system |
US9608928B1 (en) | 2016-07-06 | 2017-03-28 | Machine Zone, Inc. | Multiple-speed message channel of messaging system |
US9667681B1 (en) | 2016-09-23 | 2017-05-30 | Machine Zone, Inc. | Systems and methods for providing messages to multiple subscribers |
US10833881B1 (en) * | 2017-11-06 | 2020-11-10 | Amazon Technologies, Inc. | Distributing publication messages to devices |
TWI678087B (en) | 2018-11-22 | 2019-11-21 | 財團法人工業技術研究院 | Method of message synchronization in message queue publish and subscriotion and system thereof |
US10884832B2 (en) | 2019-03-26 | 2021-01-05 | International Business Machines Corporation | Aggregation messaging within an integration environment |
CN110662085B (en) * | 2019-10-16 | 2021-10-01 | 北京字节跳动网络技术有限公司 | Message sending method, device, readable medium and electronic equipment |
CN111314217B (en) * | 2019-12-09 | 2021-12-17 | 上海金融期货信息技术有限公司 | Multilink transmission system |
CN111858099A (en) * | 2020-07-24 | 2020-10-30 | 海尔优家智能科技(北京)有限公司 | Message subscription method and device |
CN114297109A (en) * | 2021-12-28 | 2022-04-08 | 中汽创智科技有限公司 | Data processing method and device based on subscription and publishing mode, electronic equipment and storage medium |
US11941296B2 (en) | 2022-01-03 | 2024-03-26 | International Business Machines Corporation | Intelligent storage of messages in page sets |
Family Cites Families (153)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US4264924A (en) | 1978-03-03 | 1981-04-28 | Freeman Michael J | Dedicated channel interactive cable television system |
US5404488A (en) | 1990-09-26 | 1995-04-04 | Lotus Development Corporation | Realtime data feed engine for updating an application with the most currently received data from multiple data feeds |
US5706331A (en) | 1995-08-22 | 1998-01-06 | Motorola, Inc. | System and method for selecting a subsystem for message traffic in an integrated communication network |
US5878228A (en) | 1996-11-15 | 1999-03-02 | Northern Telecom Limited | Data transfer server with time slots scheduling base on transfer rate and predetermined data |
US6704785B1 (en) | 1997-03-17 | 2004-03-09 | Vitria Technology, Inc. | Event driven communication system |
US5982436A (en) | 1997-03-28 | 1999-11-09 | Philips Electronics North America Corp. | Method for seamless splicing in a video encoder |
WO1999004589A1 (en) | 1997-07-17 | 1999-01-28 | Northern Telecom Limited | Method and system for solving cellular communications frequency planning problem |
US8074055B1 (en) | 1999-01-28 | 2011-12-06 | Ati Technologies Ulc | Altering data storage conventions of a processor when execution flows from first architecture code to second architecture code |
US7065633B1 (en) | 1999-01-28 | 2006-06-20 | Ati International Srl | System for delivering exception raised in first architecture to operating system coded in second architecture in dual architecture CPU |
US8121828B2 (en) | 1999-01-28 | 2012-02-21 | Ati Technologies Ulc | Detecting conditions for transfer of execution from one computer instruction stream to another and executing transfer on satisfaction of the conditions |
ATE414942T1 (en) | 1999-02-26 | 2008-12-15 | Henry Haugland | MASS PRODUCTION OF INDIVIDUAL VIRTUAL SERVERS, VIRTUAL WEBSITES AND VIRTUAL WEB OBJECTS |
US6549959B1 (en) | 1999-08-30 | 2003-04-15 | Ati International Srl | Detecting modification to computer memory by a DMA device |
TW454120B (en) | 1999-11-11 | 2001-09-11 | Miralink Corp | Flexible remote data mirroring |
KR100392879B1 (en) | 2000-06-02 | 2003-08-06 | 주식회사 인터넷엑스퍼트시스템 | E-mail security audit system for corporation security & virus spread by e-mail |
US8086672B2 (en) | 2000-06-17 | 2011-12-27 | Microsoft Corporation | When-free messaging |
US20020010765A1 (en) | 2000-07-21 | 2002-01-24 | John Border | Method and system for prioritizing traffic in a network |
US20020138659A1 (en) | 2000-11-01 | 2002-09-26 | Zissis Trabaris | Method and system for application development and a data processing architecture utilizing destinationless messaging |
US7929562B2 (en) | 2000-11-08 | 2011-04-19 | Genesis Telecommunications Laboratories, Inc. | Method and apparatus for optimizing response time to events in queue |
US6718327B1 (en) | 2001-08-31 | 2004-04-06 | Openwave Systems Inc. | Fault-tolerant queue with autonomous client operation |
US7406537B2 (en) | 2002-11-26 | 2008-07-29 | Progress Software Corporation | Dynamic subscription and message routing on a topic between publishing nodes and subscribing nodes |
US8122118B2 (en) | 2001-12-14 | 2012-02-21 | International Business Machines Corporation | Selection of communication protocol for message transfer based on quality of service requirements |
US20030135556A1 (en) | 2001-12-14 | 2003-07-17 | International Business Machines Corporation | Selection of communication strategies for message brokers or publish/subscribe communications |
US7668908B2 (en) | 2001-12-14 | 2010-02-23 | Microsoft Corporation | System and method for generalized and distributed scalable eventing system |
US7376092B2 (en) | 2002-04-03 | 2008-05-20 | Precache Inc. | Method and apparatus for implementing persistent and reliable message delivery |
US7162628B2 (en) | 2002-07-23 | 2007-01-09 | Cisco Technology, Inc. | Method, system, apparatus and program product for temporary personalization of a computer terminal |
US7720910B2 (en) | 2002-07-26 | 2010-05-18 | International Business Machines Corporation | Interactive filtering electronic messages received from a publication/subscription service |
FR2843210B1 (en) | 2002-08-02 | 2005-10-14 | Meiosys | METHOD FOR MIGRATION OF CONNECTIONS IN A MULTI-COMPUTER ARCHITECTURE, METHOD FOR PERFORMING OPERATING CONTINUITY USING THE METHOD OF MIGRATION, AND MULTI-COMPUTER SYSTEM THUS EQUIPPED |
US8176154B2 (en) * | 2002-09-30 | 2012-05-08 | Avaya Inc. | Instantaneous user initiation voice quality feedback |
US8606859B2 (en) | 2002-10-17 | 2013-12-10 | Tibco Software Inc. | Method and system to communicate messages in a computer network |
US7657596B2 (en) | 2002-10-24 | 2010-02-02 | Sun Microsystems, Inc. | Distributed data sharing methods and systems |
US9077991B2 (en) | 2002-12-10 | 2015-07-07 | Sony Computer Entertainment America Llc | System and method for utilizing forward error correction with video compression |
US8711923B2 (en) | 2002-12-10 | 2014-04-29 | Ol2, Inc. | System and method for selecting a video encoding format based on feedback data |
US7616208B2 (en) | 2002-12-18 | 2009-11-10 | Genesys Conferencing Ltd. | Method and system for application broadcast |
US7676580B2 (en) | 2003-03-27 | 2010-03-09 | Microsoft Corporation | Message delivery with configurable assurances and features between two endpoints |
GB0308035D0 (en) | 2003-04-08 | 2003-05-14 | Ibm | Liveness monitoring in a publish/subscribe messaging system |
US20050047396A1 (en) * | 2003-08-29 | 2005-03-03 | Helm David P. | System and method for selecting the size of dynamic voice jitter buffer for use in a packet switched communications system |
US7885901B2 (en) | 2004-01-29 | 2011-02-08 | Yahoo! Inc. | Method and system for seeding online social network contacts |
US7490083B2 (en) | 2004-02-27 | 2009-02-10 | International Business Machines Corporation | Parallel apply processing in data replication with preservation of transaction integrity and source ordering of dependent updates |
US20050262205A1 (en) | 2004-04-30 | 2005-11-24 | Nikolov Radoslav I | Delivering messages in an enterprise messaging system using message selector hierarchy |
US7640357B2 (en) | 2004-04-30 | 2009-12-29 | Sap Ag | Transmitting enterprise messages based on buffer sizes |
US7649937B2 (en) | 2004-06-22 | 2010-01-19 | Auction Management Solutions, Inc. | Real-time and bandwidth efficient capture and delivery of live video to multiple destinations |
US7613813B2 (en) | 2004-09-10 | 2009-11-03 | Cavium Networks, Inc. | Method and apparatus for reducing host overhead in a socket server implementation |
US8051140B2 (en) | 2004-10-08 | 2011-11-01 | Sharp Laboratories Of America, Inc. | Methods and systems for imaging device control |
US8065384B2 (en) | 2004-10-08 | 2011-11-22 | Sharp Laboratories Of America, Inc. | Methods and systems for imaging device event notification subscription |
US7774720B1 (en) | 2004-10-15 | 2010-08-10 | Oracle America, Inc. | Connectivity map editor |
US20060106840A1 (en) | 2004-11-04 | 2006-05-18 | International Business Machines Corporation | System and method for tracking notifications in a publish subscribe system |
US7779418B2 (en) | 2004-12-30 | 2010-08-17 | Oracle International Corporation | Publisher flow control and bounded guaranteed delivery for message queues |
AU2005322970A1 (en) | 2005-01-06 | 2006-07-13 | Tervela, Inc. | Hardware-based messaging appliance |
EP1851646A2 (en) | 2005-01-06 | 2007-11-07 | Tervela Inc. | Intelligent messaging application programming interface |
US20060174431A1 (en) | 2005-02-09 | 2006-08-10 | Dr. Fresh, Inc. | Electric toothbrush |
WO2007003984A1 (en) | 2005-06-30 | 2007-01-11 | Freescale Semiconductor, Inc. | Device and method for arbitrating between direct memory access task requests |
US8788565B2 (en) | 2005-07-18 | 2014-07-22 | Wayne Bevan | Dynamic and distributed queueing and processing system |
US8042140B2 (en) | 2005-07-22 | 2011-10-18 | Kangaroo Media, Inc. | Buffering content on a handheld electronic device |
US7941448B2 (en) | 2005-08-26 | 2011-05-10 | At&T Intellectual Property Ii, Lp | System and method for event driven publish-subscribe communications |
US7917124B2 (en) | 2005-09-20 | 2011-03-29 | Accenture Global Services Limited | Third party access gateway for telecommunications services |
JP4658142B2 (en) * | 2005-11-30 | 2011-03-23 | 富士通株式会社 | Communication apparatus and frame control method |
US8965409B2 (en) | 2006-03-17 | 2015-02-24 | Fatdoor, Inc. | User-generated community publication in an online neighborhood social network |
US8161168B2 (en) | 2006-03-18 | 2012-04-17 | Metafluent, Llc | JMS provider with plug-able business logic |
US8266321B2 (en) | 2006-06-12 | 2012-09-11 | Cloudsoft Corporation Limited | Self-managed distributed mediation networks |
US8429702B2 (en) | 2006-09-11 | 2013-04-23 | At&T Intellectual Property I, L.P. | Methods and apparatus for selecting and pushing customized electronic media content |
US9270944B2 (en) | 2007-02-14 | 2016-02-23 | Time Warner Cable Enterprises Llc | Methods and apparatus for content delivery notification and management |
US8396959B2 (en) | 2007-03-21 | 2013-03-12 | Inetco Systems Limited | Method and system for monitoring messages passed over a network |
US8068901B2 (en) | 2007-05-01 | 2011-11-29 | Medtronic, Inc. | Method and apparatus for adjusting a sensing parameter |
US8774203B2 (en) | 2007-06-07 | 2014-07-08 | Intel Corporation | One-way message notification with out-of-order packet delivery |
US8145704B2 (en) | 2007-06-13 | 2012-03-27 | Ntech Properties, Inc. | Method and system for providing media programming |
US8799213B2 (en) | 2007-07-31 | 2014-08-05 | Oracle International Corporation | Combining capture and apply in a distributed information sharing system |
US9224150B2 (en) | 2007-12-18 | 2015-12-29 | Napo Enterprises, Llc | Identifying highly valued recommendations of users in a media recommendation network |
US8234193B2 (en) | 2008-03-03 | 2012-07-31 | Wildfire Interactive, Inc. | Method and system for providing online promotions through a social network-based platform |
US20090259756A1 (en) | 2008-04-11 | 2009-10-15 | Mobitv, Inc. | Transmitting media stream bursts |
US8452833B2 (en) | 2008-05-13 | 2013-05-28 | International Business Machines Corporation | Cached message distribution via HTTP redirects |
US8898293B2 (en) | 2009-01-28 | 2014-11-25 | Headwater Partners I Llc | Service offer set publishing to device agent with on-device service selection |
US7913024B2 (en) | 2008-12-09 | 2011-03-22 | International Business Machines Corporation | Differentiating traffic types in a multi-root PCI express environment |
JP5146548B2 (en) | 2009-02-06 | 2013-02-20 | 富士通株式会社 | Packet buffer device and packet discarding method |
US8539359B2 (en) | 2009-02-11 | 2013-09-17 | Jeffrey A. Rapaport | Social network driven indexing system for instantly clustering people with concurrent focus on same topic into on-topic chat rooms and/or for generating on-topic search results tailored to user preferences regarding topic |
US8301687B2 (en) | 2009-03-31 | 2012-10-30 | Software Ag | Systems and/or methods for standards-based messaging |
US20130254314A1 (en) | 2009-06-09 | 2013-09-26 | Edmond K. Chow | Digital content delivery |
US8392555B2 (en) | 2009-06-01 | 2013-03-05 | International Business Machines Corporation | Push-back mechanism for pub/sub brokers |
US20110060812A1 (en) | 2009-09-10 | 2011-03-10 | Level 3 Communications, Llc | Cache server with extensible programming framework |
US8375095B2 (en) | 2009-12-22 | 2013-02-12 | Microsoft Corporation | Out of order durable message processing |
US8346935B2 (en) | 2010-01-15 | 2013-01-01 | Joyent, Inc. | Managing hardware resources by sending messages amongst servers in a data center |
JP5340186B2 (en) | 2010-01-21 | 2013-11-13 | アラクサラネットワークス株式会社 | Packet relay apparatus and packet relay method |
US8850015B2 (en) | 2010-01-26 | 2014-09-30 | Cisco Technology, Inc. | Network-network interface |
US8489674B2 (en) | 2010-03-11 | 2013-07-16 | Yahoo! Inc. | Distributed publish/subscribe system |
US8996563B2 (en) | 2010-04-06 | 2015-03-31 | Tokutek, Inc. | High-performance streaming dictionary |
WO2011146742A2 (en) | 2010-05-19 | 2011-11-24 | Akamai Technologies Inc. | Edge server http post message processing |
US9338064B2 (en) | 2010-06-23 | 2016-05-10 | Twilio, Inc. | System and method for managing a computing cluster |
US8626778B2 (en) | 2010-07-23 | 2014-01-07 | Oracle International Corporation | System and method for conversion of JMS message data into database transactions for application to multiple heterogeneous databases |
US8441965B2 (en) | 2010-08-05 | 2013-05-14 | Apple Inc. | Methods and apparatus for reducing data transmission overhead |
US8589732B2 (en) | 2010-10-25 | 2013-11-19 | Microsoft Corporation | Consistent messaging with replication |
CN102469032B (en) | 2010-10-29 | 2015-03-25 | 国际商业机器公司 | Method and system for issuing-ordering message transmission |
CN103416017B (en) | 2010-11-12 | 2016-11-16 | 交互数字专利控股公司 | For performing channel aggregation and the method and apparatus of media access control re-transmission |
WO2012078971A2 (en) | 2010-12-10 | 2012-06-14 | Rockefeller Consulting Technology Integration, Inc. | Systems and methods for integrating personal social networks within an organization |
US20120150960A1 (en) | 2010-12-13 | 2012-06-14 | Gargi Nalawade | Social Networking |
US9740762B2 (en) | 2011-04-01 | 2017-08-22 | Mongodb, Inc. | System and method for optimizing data migration in a partitioned database |
US9805108B2 (en) | 2010-12-23 | 2017-10-31 | Mongodb, Inc. | Large distributed database clustering systems and methods |
US8908814B2 (en) | 2011-01-07 | 2014-12-09 | Marvell World Trade Ltd. | Low latency SIMD architecture for iterative decoders |
US8909719B2 (en) | 2011-03-24 | 2014-12-09 | Yammer, Inc. | Method of managing feeds based on classifications |
US20120278728A1 (en) | 2011-04-29 | 2012-11-01 | Sling Media Inc. | Download monitoring in a media distribution system |
US8719876B2 (en) | 2011-05-06 | 2014-05-06 | Verizon Patent And Licensing Inc. | Video on demand architecture |
US9760414B2 (en) | 2011-05-18 | 2017-09-12 | International Business Machines Corporation | Preserving event data for lazily-loaded macro components in a publish/subscribe system |
US8886731B2 (en) | 2011-06-01 | 2014-11-11 | Microsoft Corporation | Decentralized relaying algorithm for mobile devices |
US20130031177A1 (en) | 2011-07-29 | 2013-01-31 | Myxer, Inc. | Systems and methods for dynamic media selection |
US9177336B2 (en) | 2011-09-08 | 2015-11-03 | Telefonaktiebolaget L M Ericsson (Publ) | Apparatuses and methods for recommending a path through an information space |
US20130066967A1 (en) | 2011-09-12 | 2013-03-14 | Vizibility Inc. | System and method for discovering and interacting with common connections |
CN103077179B (en) | 2011-09-12 | 2016-08-31 | 克利特股份有限公司 | For showing the computer implemented method of the personal time line of the user of social networks, computer system and computer-readable medium thereof |
US8607249B2 (en) | 2011-09-22 | 2013-12-10 | Oracle International Corporation | System and method for efficient concurrent queue implementation |
JP2013093735A (en) | 2011-10-25 | 2013-05-16 | Fujitsu Ltd | Data sampling device, method, and program |
KR102024182B1 (en) | 2011-11-18 | 2019-09-24 | 톰슨 라이센싱 | System comprising a publish/subscribe broker for a remote management of end-user devices, and respective end-user device |
EP3249546B1 (en) | 2011-12-14 | 2022-02-09 | Level 3 Communications, LLC | Content delivery network |
DE112011106016T5 (en) | 2011-12-23 | 2014-09-11 | Intel Corporation | Common send queue |
US20130290449A1 (en) | 2012-04-25 | 2013-10-31 | Origami Labs, Inc. | Privacy-based social content broadcast systems and methods |
EP3059902B1 (en) * | 2012-06-20 | 2017-08-30 | Huawei Technologies Co., Ltd. | Method, node, mobile terminal, and system for identifying network tethering behavior |
BR112014004057A2 (en) | 2012-07-02 | 2017-03-07 | Sony Corp | transmission device and method, and network device |
US8850263B1 (en) | 2012-09-14 | 2014-09-30 | Amazon Technologies, Inc. | Streaming and sampling in real-time log analysis |
US20140114738A1 (en) | 2012-10-24 | 2014-04-24 | Erick Tseng | Automatic Check-In Using Social-Networking Information |
US9141681B2 (en) | 2012-11-29 | 2015-09-22 | Red Hat, Inc. | Creating a column family in a database |
US9037556B2 (en) | 2012-12-03 | 2015-05-19 | Vmware, Inc. | Distributed, transactional key-value store |
US9258059B2 (en) * | 2012-12-31 | 2016-02-09 | Broadcom Corporation | Non-semaphore linked list management |
US20140237057A1 (en) | 2013-02-21 | 2014-08-21 | Genesys Telecommunications Laboratories, Inc. | System and method for processing private messages in a contact center |
US9043822B1 (en) | 2013-03-14 | 2015-05-26 | Cox Communications, Inc. | Remote time shift buffering with shadow tuning |
US20140279944A1 (en) | 2013-03-15 | 2014-09-18 | University Of Southern California | Sql query to trigger translation for maintaining consistency of cache augmented sql systems |
US8850490B1 (en) | 2013-05-07 | 2014-09-30 | Google Inc. | Consuming paid media in an internet-based content platform |
US9767138B2 (en) | 2013-06-12 | 2017-09-19 | Oracle International Corporation | In-database sharded queue for a shared-disk database |
US8824664B1 (en) | 2013-06-13 | 2014-09-02 | Genesys Telecommunications Laboratories, Inc. | System and method for controlling lifespan of interaction requests |
US9342554B2 (en) | 2013-07-05 | 2016-05-17 | Facebook, Inc. | Techniques to generate mass push notifications |
US9813515B2 (en) | 2013-10-04 | 2017-11-07 | Akamai Technologies, Inc. | Systems and methods for caching content with notification-based invalidation with extension to clients |
US9729653B2 (en) | 2014-01-23 | 2017-08-08 | Software Ag | Systems and/or methods for automatically tuning a delivery system for transmission of large, volatile data |
US20150262151A1 (en) | 2014-03-11 | 2015-09-17 | Nibl, Inc. | Access Control System for Online Content |
US9450895B2 (en) | 2014-03-28 | 2016-09-20 | Futurewei Technologies, Inc. | Context-aware dynamic policy selection for messaging behavior |
US20150317676A1 (en) | 2014-05-01 | 2015-11-05 | Vibe-It LLC | Systems and methods for monitoring and responding to customer feedback |
CN103995786B (en) * | 2014-06-04 | 2017-05-17 | 龙芯中科技术有限公司 | Transmission method and device of high-speed cache consistency information |
US20150379160A1 (en) | 2014-06-29 | 2015-12-31 | Yaniv AVRAHAM | Apparatus and methods for transferring focus control and related return content within a graphical user interface |
US10033784B2 (en) | 2014-07-14 | 2018-07-24 | International Business Machines Corporation | Predictive management of offline storage content for mobile applications and optimized network usage for mobile devices |
US9703675B2 (en) | 2014-08-11 | 2017-07-11 | Microsoft Technology Licensing, Llc | Structured logging and instrumentation framework |
US20160179855A1 (en) | 2014-12-23 | 2016-06-23 | Yahoo! Inc. | Ubiquitous content access and management |
US20160219089A1 (en) | 2015-01-23 | 2016-07-28 | Ebay Inc. | Systems and methods for messaging and processing high volume data over networks |
US9742651B2 (en) | 2015-03-05 | 2017-08-22 | International Business Machines Corporation | Client-side fault tolerance in a publish-subscribe system |
US10432741B2 (en) | 2015-03-25 | 2019-10-01 | Open Access Technology International, Inc. | Systems and methods for guaranteeing delivery of pushed data to remote clients |
WO2016183550A1 (en) | 2015-05-14 | 2016-11-17 | Walleye Software, LLC | Dynamic table index mapping |
US9407585B1 (en) | 2015-08-07 | 2016-08-02 | Machine Zone, Inc. | Scalable, real-time messaging system |
US9319363B1 (en) | 2015-08-07 | 2016-04-19 | Machine Zone, Inc. | Scalable, real-time messaging system |
US9602455B2 (en) | 2015-08-07 | 2017-03-21 | Machine Zone, Inc. | Scalable, real-time messaging system |
US9319365B1 (en) | 2015-10-09 | 2016-04-19 | Machine Zone, Inc. | Systems and methods for storing and transferring message data |
US9385976B1 (en) * | 2015-10-09 | 2016-07-05 | Machine Zone, Inc. | Systems and methods for storing message data |
US9450944B1 (en) | 2015-10-14 | 2016-09-20 | FullArmor Corporation | System and method for pass-through authentication |
US9509684B1 (en) | 2015-10-14 | 2016-11-29 | FullArmor Corporation | System and method for resource access with identity impersonation |
US9762563B2 (en) | 2015-10-14 | 2017-09-12 | FullArmor Corporation | Resource access system and method |
US9397973B1 (en) | 2015-10-16 | 2016-07-19 | Machine Zone, Inc. | Systems and methods for transferring message data |
US9577911B1 (en) | 2016-03-21 | 2017-02-21 | Black Cloud Analytics, Inc. | Distributed computation system incorporating agent network, paths and associated probes |
US9602450B1 (en) | 2016-05-16 | 2017-03-21 | Machine Zone, Inc. | Maintaining persistence of a messaging system |
CN105893635B (en) * | 2016-06-23 | 2019-04-30 | 河海大学 | Netted multiple index matching process towards distribution subscription |
US9608928B1 (en) | 2016-07-06 | 2017-03-28 | Machine Zone, Inc. | Multiple-speed message channel of messaging system |
US9667681B1 (en) | 2016-09-23 | 2017-05-30 | Machine Zone, Inc. | Systems and methods for providing messages to multiple subscribers |
-
2016
- 2016-09-23 US US15/274,281 patent/US9667681B1/en not_active Expired - Fee Related
-
2017
- 2017-05-04 EP EP17724677.4A patent/EP3516830A1/en not_active Withdrawn
- 2017-05-04 AU AU2017331066A patent/AU2017331066A1/en not_active Abandoned
- 2017-05-04 JP JP2019515864A patent/JP2019535072A/en active Pending
- 2017-05-04 US US15/586,773 patent/US9942339B1/en not_active Expired - Fee Related
- 2017-05-04 WO PCT/US2017/031041 patent/WO2018057066A1/en unknown
- 2017-05-04 CN CN201780072371.2A patent/CN110121863A/en active Pending
-
2018
- 2018-03-01 US US15/909,710 patent/US10382574B2/en not_active Expired - Fee Related
-
2019
- 2019-07-09 US US16/506,890 patent/US20190335010A1/en not_active Abandoned
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US11429290B2 (en) | 2020-11-13 | 2022-08-30 | Argo AI, LLC | Methods and systems for providing a lockless access to a shared memory region in a publish and subscribe system |
US11709620B2 (en) * | 2020-11-13 | 2023-07-25 | Ford Global Technologies, Llc | Methods and systems for memory management in a publish and subscribe system |
Also Published As
Publication number | Publication date |
---|---|
EP3516830A1 (en) | 2019-07-31 |
US9942339B1 (en) | 2018-04-10 |
AU2017331066A1 (en) | 2019-04-04 |
US9667681B1 (en) | 2017-05-30 |
CN110121863A (en) | 2019-08-13 |
US20180091612A1 (en) | 2018-03-29 |
WO2018057066A1 (en) | 2018-03-29 |
US10382574B2 (en) | 2019-08-13 |
JP2019535072A (en) | 2019-12-05 |
US20180191854A1 (en) | 2018-07-05 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US10382574B2 (en) | Systems and methods for providing messages to multiple subscribers | |
US10541945B2 (en) | Systems and methods for storing message data | |
US10637947B2 (en) | Scalable, real-time messaging system | |
US10038661B2 (en) | Systems and methods for storing and transferring message data | |
US10630785B2 (en) | Scalable, real-time messaging system | |
US9319363B1 (en) | Scalable, real-time messaging system | |
US9871750B2 (en) | Maintaining persistence of a messaging system | |
US20190268293A1 (en) | Scalable, real-time messaging system | |
US20170339086A1 (en) | Efficient message exchange system |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AS | Assignment |
Owner name: MACHINE ZONE, INC., CALIFORNIA Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:MILYAKOV, IGOR;REEL/FRAME:049713/0246 Effective date: 20161003 |
|
AS | Assignment |
Owner name: SATORI WORLDWIDE, LLC, CALIFORNIA Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNOR:MACHINE ZONE, INC.;REEL/FRAME:049777/0075 Effective date: 20171109 |
|
STPP | Information on status: patent application and granting procedure in general |
Free format text: DOCKETED NEW CASE - READY FOR EXAMINATION |
|
STCB | Information on status: application discontinuation |
Free format text: ABANDONED -- FAILURE TO RESPOND TO AN OFFICE ACTION |