JP2011170751A - Bus system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a bus system capable of efficiently distributing loads to be required for each server. <P>SOLUTION: This invention relates to the bus system which collects load conditions of a plurality of servers which provide service according to a request from a client to attain load distribution based on collection results, and includes: a collection means for collecting load conditions of each server; a prediction means for predicting a load amount of a server concerning provision of service according to the request based on processing characteristics in the past of the service provided by each server; a selection means for selecting a server being a request destination based on the collected load conditions of each server, and the predicted load amount of the server; and a transmitting means for transmitting the request from the client to the selected server, wherein the selection means selects the server which provides the service according to the request and has the highest priority to be set according to the load conditions of each server, as the server being the request destination. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to a bus system such as an ESB (Enterprise Service Bus).

  As a technique used to service an existing system in SOA (Service-Oriented Architecture), there is a method of using a bus system called ESB. The ESB concept is a concept in software design in which an SOA system that accesses services (applications and components) and cooperates and cooperates with a plurality of services is configured based on a logical software bus.

FIG. 9 is a diagram illustrating a configuration example of a conventional load distribution system 50.
Services A and B such as applications installed in the servers 10a and 10b and client programs CP1 to CP3 installed in the clients 20a to 20c are connected via the ESB system 30. All communication between the services A and B and the programs CP1 to CP3 is performed via the ESB system 30. Communication between each service A, B and the ESB system 30 is performed by general-purpose HTTP, SOAP, JMS, or the like.

  Here, the ESB system 30 accesses a desired service (for example, service A) in response to a request from the client application CP. However, the processing capability of the server (for example, the server 10a) that implements the service A that has received the request is low. If the service A has decreased, the service A may not be able to process the request.

  In order to solve such a problem, it is now common to prevent a load from being concentrated on a single server by mounting a load balancer that realizes load balancing in the ESB system (for example, Patent Documents). 1).

JP 2005-182641 A

10 and 11 are diagrams for explaining problems of the ESB system 30 in which the load balancer 31 is mounted.
As shown in FIG. 10, it is assumed that the client program CP2 makes a request for the message size M1 to the service B while the client program CP1 is using the service A. In this case, since the load balancer 31 of the ESB system 30 uses the service A of the server 10a by the client program CP1, the resource is not used, not the service B of the server 10a where the resource is already used. The request is sent to the service B of the server 10b. As a result, the server 10a uses 50% of the resources by using the service 10a, and the server 10b uses 40% of the resources by using the service 10b (see FIG. 10). .

  Thereafter, as shown in FIG. 11, it is assumed that the client program CP3 makes a request to the service B with a large message size M2 (> M1) that requires use of 60% of the resource. Since the service B is already used in the server 10b, the load balancer 31 tries to make a request to the service B of the server 10a. However, in the server 10a, since 50% of the resources are used by using the service A, there may arise a problem that the request from the client program CP3 cannot be satisfied.

  The present invention has been made in view of the circumstances described above, and an object thereof is to provide a bus system capable of efficiently distributing the load required for each server.

  A bus system according to the present invention is a bus system that collects load statuses of a plurality of servers that provide services in response to a request from a client, and realizes load distribution based on the collected results, and the load statuses of the servers. Collecting means for collecting the server, prediction means for predicting the load amount of the server related to the provision of the service according to the request based on the past processing characteristics of the service provided by the server, and the collected server A selection unit that selects a server that is a request destination based on the load status and the predicted load amount of the server; and a transmission unit that transmits a request from the client to the selected server. Can provide a service according to the request and is set according to the load status of each server. The highest server, and selects a server to which the request destination.

  According to such a configuration, since the requests are distributed to an appropriate server according to the load status of the server (in other words, the resource usage status), the server can be used efficiently.

  As described above, according to the present invention, it is possible to efficiently distribute the load required for each server.

It is a figure which shows schematic structure of the load distribution system which concerns on this embodiment. It is the figure which illustrated resource information concerning the embodiment. It is the figure which illustrated the status table which concerns on the same embodiment. It is the figure which illustrated the prediction processing performance table which concerns on the same embodiment. It is the figure which illustrated the load balancer table concerning the embodiment. It is a sequence diagram which shows the table update process which concerns on the same embodiment. It is a sequence diagram which shows the request process which concerns on the same embodiment. It is a figure which shows schematic structure of the load distribution system which concerns on a modification. It is a schematic block diagram of the conventional load distribution system. It is a figure for demonstrating the problem of the conventional ESB system carrying a load balancer. It is a figure for demonstrating the problem of the conventional ESB system carrying a load balancer.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

A. Embodiment (1) Configuration of Embodiment FIG. 1 is a diagram illustrating a configuration example of a load distribution system 500 according to the present embodiment.
A server 100 that provides a plurality of services, a client 200 equipped with a client program CP1 that uses each service, and an ESB system 300 that allocates requests from the client program CP to appropriate services of the server 100, and the like. Configured. In FIG. 1, for convenience of explanation, only two servers and one client are shown, but there are originally a large number of servers and clients. Also, it goes without saying that the types of services provided by each server 100 (assuming services A, B, and C in the present embodiment), the number, etc. are not limited to those shown in FIG.

  Each server 100 provides a service in response to a request from the client program CP. Each service implemented in the server 100 implements a listener unit 111, an application execution unit 112, a resource monitor unit 113, and an agent unit 114 by cooperating with hardware resources (CPU, ROM, RAM, etc.) of the server 100. To do.

  The listener unit 111 receives a request from the client program CP via the ESB system 300, while the application execution unit 112 executes processing of the service according to the request received by the listener unit 111. These listener means 111 and application execution means 112 are realized by well-known Web server software (such as Apache HTTP Server) and HTML files.

  For each service provided by the server 100, the resource monitor unit 113 aggregates information (hereinafter referred to as resource information) indicating the usage status of hardware resources (CPU, memory, etc.) at a predetermined timing, and transmits it to the agent unit 114. To do. The agent means 114 of each service 100 sends the listener means 311b of the ESB system 300.

FIG. 2 is a diagram illustrating resource information provided from the services A and B of the server 100 a to the ES system 300.
As shown in FIG. 2, the resource information provided by the services A and B includes a client ID for identifying the client 200 that uses the service, a size of a message (request message size) transmitted from the client 200, and And information (detailed resource information) representing the detailed usage status of hardware resources. For example, regarding the use of the service A, the CPU usage rate “20%”, the memory usage amount “40 KB”, the NIC usage amount “in response to a request message size“ 2 MB ”transmitted from the client α. While 10% "hardware resources are used, the CPU usage rate is" 10% ", the memory usage is" 20KB ", the NIC usage is" in response to a request message size "1MB" transmitted from the client β. Information such as “10%” hardware resource is used is included in the resource information of service A.

  Similarly, regarding the use of the service B, the CPU usage rate “20%”, the memory usage amount “20 KB”, and the NIC usage amount “10%” according to the request message size “2 MB” request transmitted from the client β. While the hardware resource is used, the CPU usage rate is “30%”, the memory usage is “40 KB”, and the NIC usage is “15%” according to the request message size “3 MB” transmitted from the client γ. Information such as hardware resources used is included in the resource information of the service SV2.

  In the above description, the resource information provided to the ESB system 300 from the services A and B implemented on the server 100a is illustrated. However, the resource information is also provided for each service implemented on other servers 100 in the same manner. provide.

  The ESB system 300 realizes listener means 311a, 311b, agent totaling means 312, load balancer 313, and registry updating means 314 by cooperation of software installed in the ESB system 300 with hardware resources.

  The listener unit 311 a receives a request for a specific service (for example, service A) from the client 200 (client program CP), and transmits it to the load balancer 313. The listener unit 311b receives the resource information transmitted from each service and passes it to the agent totaling unit 312.

The agent aggregation means (collection means) 312 creates a status table TA1 by collecting resource information (server load status) received from each service.
FIG. 3 is a diagram illustrating registered contents of the status table TA1.
In the status table TA1, the service type, the name of the server 100 that provides the service, and the resource detailed information of the server are registered in association with each other. For example, the service A is provided by the server 100a and the server 100b. The server 100a uses the service A so that the CPU usage rate is “60%”, the memory usage is “1 GB”, and the NIC usage is “40%”. ”Is used, and the server 100b has a hardware resource with a CPU usage rate of“ 20% ”, a memory usage amount of“ 0.5 GB ”, and a NIC usage amount of“ 20% ”by using the service SB. used.

The registry update unit (prediction unit) 314 creates / updates a prediction processing performance table TA2 for each request based on the resource information received from each service (in other words, the past processing characteristics of the service). Store in the registry 314a.
FIG. 4 is a diagram illustrating an example of the prediction processing performance table TA2.
In the prediction processing performance table TA2, detailed request information and prediction resource information are registered in association with each other. Here, the detailed request information is information including a service name and a request message size (hereinafter simply abbreviated as message size), and the predicted resource information is the predicted hardware resource usage status of the server 100 ( In other words, the server load information is predicted information). For example, when a request message having a message size of 2 MB is transmitted from the client 200 to the service A, the CPU usage rate (predicted value) is “20%”, the memory usage (predicted value) is “40 KB”, NIC It is predicted that the usage amount (predicted value) will be “20%”. Similarly, when a request message having a message size of 1 MB is transmitted from the client 200 to the service A, the CPU usage rate (predicted value) is “10%”, the memory usage amount (predicted value) is “20 KB”, It is predicted that the NIC usage (predicted value) is “10%”.

  The load balancer 313 creates / updates the load balance table TA3 based on the resource information received from each service, and stores this in the load balance registry 313a. Further, the load balancer (selection unit, transmission unit) 313 refers to the load balance table TA3 and the like to select a server as a transmission destination of a request from the client 200, and a service of the selected server 100 (for example, the server 100a). The request is transferred to the service A).

FIG. 5 is a diagram illustrating the load balance table TA3.
In the load balance table TA3, the name of the server 100, the type of service provided by the server 100, the priority related to the use of the server 100 (hereinafter simply referred to as priority), the status (active, nonactive, etc.) Are registered in association with each other.

  Here, the priority means that the higher the numerical value, the higher the probability of assigning a request from the client. Therefore, in the example shown in FIG. 5, since the priority set to the server 100a is the highest (specifically, the priority is “10”), the server 100a is likely to be assigned the next request from the client. It is predicted.

(2) Operation of the embodiment 2-1. Table Update Processing FIG. 6 is a sequence diagram showing table update processing executed in the ESB system 300. In the following description, it is assumed that resource information is transmitted from the service A of the server 100a to the ESB system 300.
The agent means 114 of the service A periodically requests the resource monitor means 113 to transfer resource information (for each request or overall resource information) (step A1). The transfer request may be irregular.

  The resource monitoring unit 113 acquires the resource information of the server 100a on which the service A operates, and returns it to the agent unit 114 (step A2). The resource information acquired by the resource monitoring unit 113 includes, for example, a client ID for identifying the client 200 using the service A as shown in the upper part of FIG. 2 and the size of the message transmitted from the client 200 (request message size). ) And information (resource detailed information) representing the detailed usage status of the hardware resource. It should be noted that the detailed resource information may include various information such as the number of communication connections and the network bandwidth usage rate in addition to the CPU usage rate, the memory usage amount, and the NIC usage amount.

  Upon obtaining the resource information, the agent means 114 transmits it to the listener 311b of the ESB system 300 (step A3). The communication protocol is not particularly limited. When the listener means 311b receives the resource information from the service A of the server 100a, the listener means 311b sends it to the agent counting means 312 (step A4). The agent totaling unit 312 collects various resource information transmitted from each service of each server 100, creates a status table TA1 as shown in FIG. 3, and predicts as shown in FIG. 4 based on the resource information. The processing performance table TA2 is created / updated (step A5).

  Then, the agent totaling unit 312 accesses the load balancer registry 313a and reads the load balancer table TA3 (step A6). As shown in FIG. 5, in the load balance table TA3, the name of the server 100, the type of service provided by the server 100, and the priority (status such as active and nonactive) are registered in association with each other. ing. For example, based on the status table TA1 illustrated in FIG. 3 and the load balancer table TA3 illustrated in FIG. Then, the load balancer table TA3 is updated (step A7 → step A8). This priority is set higher as the load at that time (that is, the resource usage rate) is larger.

  When the above processing is completed, the agent totaling unit 312 requests the registry updating unit 314 to update the prediction processing performance table TA2 registered in the service registry 314a (step A9). The registry update unit 314 receives the update request from the agent totaling unit 312, associates it with the message size for each request, updates the predicted resource information for each message size (step A10), and ends the process. The priority change policy, the priority setting rule, and the like can be set and changed as appropriate according to the design of the load distribution system 500 and the like.

2-2. Request Processing FIG. 7 is a sequence diagram showing request processing executed in the ESB system 300 when a service request is received from a client.
The client (client program CP) 200 sends a request of a predetermined message size (for example, 2 MB) to the ESB system 300 in order to use a specific service (for example, service A) (step B1). When the load balancer 313 of the ESB system 300 receives the request from the listener unit 311a, the load balancer 313 performs protocol conversion according to the protocol of the request destination service (for example, service A) and acquires the load balancer table TA3 from the load balancer table TA3. (Step B2). Further, the load balancer 313 acquires the prediction processing performance table TA2 registered in the service registry 314a (step B3), and uses the message size of the request transmitted from the client 200 and the service of the request destination as a key to update the registry. By retrieving the TA, the server predictive resource information (CPU usage rate, memory usage amount, NIC usage amount, etc.) necessary for providing the service corresponding to the request is grasped.

  Then, the load balancer 313 refers to the load balancer table TA3 to grasp the server (for example, the server 100a) on which the requested service is deployed, and when there are a plurality of such servers (for example, the servers 100a and 100b). ), With reference to the priority registered in association with each server, the server with the highest priority (for example, server 100a) is selected (step B3). Here, the higher the resource usage, the higher the priority. Therefore, when selecting a server, the server 100 that can process the request and that has the least surplus resources is selected. The load balancer 313 transfers the request to the service of the server 100 selected in this way (for example, the service A of the server 100a) (step B4), and ends the process.

As described above, according to the present embodiment, since the request is distributed to an appropriate server according to the usage status of the server resource, the server can be used efficiently.
Further, since the activity of each server is also checked (see the load balancer table TA3), it is possible to prevent an operation such as accessing a server that is down. Furthermore, since resource depletion can be suppressed, TAT (Turn Around Time) after sending a request can be flattened.
In addition, since the activity check is performed, it is possible to dynamically change the resource by intentionally stopping the server.

B. Modified Example In the above-described embodiment, an example in which the load balancer registry 313a in which the load balancer table TA3 is stored is provided in the ESB system 300 is illustrated, but the load balancer table TA3 is stored outside the ESB system 300. A database DB may be provided.

FIG. 8 is a diagram illustrating a configuration of a load distribution system 500 ′ according to a modification, and corresponds to FIG. Note that portions corresponding to those in FIG. 1 are denoted by the same reference numerals, and detailed description thereof is omitted.
The load distribution system 500 ′ includes a plurality of ESB systems 300a and 300b, and a database DB in which a load balancer table TA3 is stored is provided outside the ESB systems 300a and 300b.

  When the database DB is provided outside the ESB systems 300a and 300b, network communication is required when each ESB system 300a and 300b acquires information registered in the load balancer table TA3. There is a merit that 300a and 300b can share information. Furthermore, by using a highly reliable database DB, it is possible to perform loss of data registered in the database DB or rollback when a problem occurs.

  It should be noted that the steps of each process shown in this embodiment and the like can be executed in any order or in parallel as long as the process contents do not contradict each other. Further, in this specification and the like, the term “means” does not simply mean a physical means, but also includes a case where the functions of the means are realized by software. Furthermore, the function of one means may be realized by two or more physical means, or the functions of two or more means may be realized by one physical means. In addition, the software development support program according to the present invention is downloaded through various recording media such as an optical disk such as a CD-ROM and a DVD-ROM, a magnetic disk, and a semiconductor memory, or via a communication network. Can be installed or loaded on a computer.

  In addition, a part or all of the aspects shown in the present embodiment and the modified examples can be described as an appendix, but are not limited thereto.

(Supplementary note 1) A bus system that collects the load status of a plurality of servers that provide services in response to a request from a client and implements load distribution based on the collection result, and collects the load status of each server Means, predicting means for predicting the load amount of the server related to the provision of the service according to the request, based on the past processing characteristics of the service provided by each server, and the collected load status of each server, A selection unit that selects a server that is a request destination based on the predicted load amount of the server; and a transmission unit that transmits a request from the client to the selected server. A server with the highest priority set according to the load status of each server. , Bus system, characterized in that, to select a server to which the request destination.

(Supplementary note 2) The bus system according to supplementary note 1, wherein the collection unit collects information representing a usage status of a resource of the server from each service provided by the server.

(Appendix 3) The past processing characteristics of the service include the type of the service, the request message size of the service, and the load amount of the server required for providing the service. Or the bus system of 2.

(Supplementary note 4) The bus system according to any one of supplementary notes 1 to 3, wherein the priority is set higher as the load at the time is larger by the selection unit.

(Additional remark 5) The said prediction means updates the load amount of the server which concerns on provision of the service according to the said request to estimate based on the load condition of each server collected by the said collection means, Additional remark 1-4 The bus system according to any one of the supplementary notes.

(Supplementary Note 6) A client, a plurality of servers that provide services in response to requests from the client, and a bus system that collects load status of each server and realizes load distribution based on the collection result A load distribution system, wherein the bus system relates to provision of a service according to the request based on a collection unit that collects a load status of each server and a past processing characteristic of the service provided by each server. Prediction means for predicting the load amount of the server, selection means for selecting the server that is the request destination based on the collected load status of each server, and the predicted load amount of the server, and the selected server Transmitting means for transmitting a request from the client, and the selecting means provides a service corresponding to the request Rukoto can, and the load balancing system the highest server priority which is set according to the load status of each server, selects a server to which the request destination, characterized in that.

(Supplementary note 7) A method for collecting load statuses of a plurality of servers that provide services in response to requests from clients and realizing load distribution based on the collection results, the collecting step collecting the load statuses of the respective servers A prediction step of predicting the load amount of the server related to the provision of the service according to the request based on the past processing characteristics of the service provided by each server, the collected load status of each server, and the prediction A selection step of selecting a server that is a request destination based on the load amount of the server, and a transmission step of transmitting a request from the client to the selected server. In the selection step, the request Can be provided according to the service, and is set according to the load status of each server. The highest server degrees, is selected as the server to be the request destination, the load distribution wherein the.

500, 500 '... load distribution system, 100 ... server, 111, 311a, 311b ... listener means, 112 ... application execution means, 113 ... resource monitoring means, 114 ... agent means, 200 ... client, 300 ... ESB system, 312 ... Agent aggregation means, 313 ... load balancer, 313a ... load balancer registry, 314 ... registry update means, 314a ... service registry, DB ... database, TA1 ... status table, TA2 ... prediction processing performance table, TA3 ... load balancer table.

Claims (7)

A bus system that collects the load status of multiple servers that provide services in response to requests from clients and implements load distribution based on the collected results.
Collection means for collecting the load status of each server;
Prediction means for predicting the load amount of the server related to the provision of the service according to the request based on the past processing characteristics of the service provided by each of the servers,
Selection means for selecting a server as a request destination based on the collected load status of each server and the predicted load amount of the server;
Transmission means for transmitting a request from the client to a selected server,
The selection means can select a server that can provide a service according to the request and has the highest priority set according to the load status of each server as the server that is the request destination. A bus system characterized by this. The collecting means includes
The bus system according to claim 1, wherein information representing a usage status of a resource of the server is collected from each service provided by the server.   The past processing characteristics of the service include a type of the service, a request message size of the service, and a load amount of a server required for providing the service. The described bus system.   The bus system according to any one of claims 1 to 3, wherein the priority is set higher as the load at the time is larger by the selection means.   5. The server according to claim 1, wherein the prediction unit updates a load amount of a server related to provision of a service according to the request to be predicted based on a load state of each server collected by the collection unit. The bus system according to claim 1. With the client,
A plurality of servers that provide services in response to requests from the clients;
A load distribution system comprising: a bus system that collects a load situation of each server and realizes load distribution based on a collection result;
The bus system is
Collection means for collecting the load status of each server;
Prediction means for predicting the load amount of the server related to the provision of the service according to the request based on the past processing characteristics of the service provided by each of the servers,
Selection means for selecting a server as a request destination based on the collected load status of each server and the predicted load amount of the server;
Transmission means for transmitting a request from the client to a selected server,
The selection means can select a server that can provide a service according to the request and has the highest priority set according to the load status of each server as the server that is the request destination. A load balancing system characterized by that. A method for collecting load statuses of a plurality of servers that provide services in response to a request from a client and realizing load distribution based on the collected results,
A collecting step of collecting a load status of each of the servers;
A predicting step of predicting a load amount of a server related to provision of a service according to the request based on a past processing characteristic of a service provided by each of the servers;
A selection step of selecting a server as a request destination based on the collected load status of each server and the predicted load amount of the server;
Sending a request from the client to a selected server,
In the selecting step, a server that can provide a service according to the request and has the highest priority set according to the load status of each server is selected as the server that is the request destination. A load balancing method characterized by that.

Images