JP2008299527A - Grid computing system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To minimize the alternation of a status quo task system, and to achieve simple grid system construction in introducing a grid system. <P>SOLUTION: In a task system 101, a task program 203 whose load is low and a task program 202 whose load is high coexist. A plug-in 103 is configured to be assembled into the task program 202 whose load is high. This grid computing system includes a means for transmitting/receiving a processing request from the task program 202 whose load is high through a network 205 to a resource pool 104. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a grid computing system, and more particularly to a processing allocation technique related to an application service provided on a computer network, and more specifically, to a grid computing system in which a plurality of information processing apparatuses are connected on a network. Connection technology.

  There is a tendency for the amount of information to be processed by a computer to increase, and there is a case where it cannot be processed by a single computer. In recent years, a technique called grid computing has begun to attract attention as a solution to such problems. The grid constitutes a heterogeneous mixed system composed of a plurality of computers and the like, and aims to distribute the load. As a technique related to such a system, there is the following Patent Document 1.

JP 2005-56201 A

  By the way, in a load distribution system using a grid, the entire system to be distributed is replicated in advance on a plurality of computers, the same program execution environment is prepared, and load distribution is performed. However, even in a heavily loaded system, the load of all internal programs is not always high, and the load is often a part of the entire system. In other words, the current grid system also performs distributed processing for program portions that do not need to be distributed. For this reason, there arises a disadvantage that the user who operates the grid substantially invests in an unnecessary computer.

  When considering the operation of a business system that uses a grid system, when a business system that is operated in a form that does not use a grid system is transferred to a load balancing environment that uses the grid system, This is accompanied by a change in the operating environment. For this reason, the cost of modifying the current business system for shifting to the grid is required, and a risk for an operation failure is also generated.

  When considering distributed processing in a grid system, it is necessary to design and rule in advance with regard to how the load is distributed in the grid system. Alternatively, it is necessary to define computers in the grid system.

  However, such work is very complicated and cumbersome, and it is difficult for anyone other than an engineer with specialized knowledge.

  It is an object of the present invention to minimize the modification of a current business system and to easily construct a grid system when introducing a grid system. Another object is to enable efficient use of a computer system.

  According to an aspect of the present invention, a business system in which a plurality of business programs operate, a resource pool including a load balancer that is network-connected to the business system and balances loads of a plurality of processing nodes, and the resource pool A grid management system having a management console for controlling the system, and a plug-in that is incorporated in the business system in units of the business program and enables load distribution by the grid system A plug-in grid computing system is provided. The plug-in is attached to a part with a high load in the business program, the plug-in and the resource pool are connected to the network, a processing request from the plug-in to the processing node, and a response to the processing request And first communication means for transmitting and receiving data between the plug-in and the resource pool based on the exchange. It is preferable to have a 2nd communication means which transmits / receives data between the said grid system and the said grid management system.

  Further, it is preferable that the management console receives a load report of the resource pool by the second communication unit and determines load distribution to the processing nodes based on a load situation based on the load report.

  The plug-in may include replacement means for requesting processing to the resource pool via the first communication means instead of the processing by the business program by the base software that is the base of the business program. . When the plug-in is determined to be able to perform distributed processing in the resource pool, and when it is determined that distributed processing is possible, to request processing to the resource pool, and when it is determined that distributed processing cannot be performed And a means for processing using the software before the replacement by the replacement means.

  Means for compressing and transmitting data from the plug-in between the plug-in and the resource pool by the first communication means; means for compressing and transmitting data from the resource pool; and the plug Means for receiving and decompressing the compressed data from the in-line, and means for receiving and decompressing the compressed data from the resource pool. It is preferable that the resource pool and the management console have management means for managing resources by virtually arranging the plurality of processing nodes in the resource pool in a network. Further, each of the processing nodes arranged in a mesh shape may have means for transmitting its own load status to the management console. Furthermore, the management means may add processing nodes autonomously according to a report from the processing nodes when the load increases, and autonomously reduce processing nodes according to a report from the processing nodes when the load decreases. According to the system described above, a grid load distribution system can be easily introduced without modifying the business program.

  It should be noted that the components shown in the solution means can take the form of the entire system or the form of individual devices (systems). Further, a program for performing the above-described processing and a computer-readable recording medium on which the program is recorded also fall within the scope of the present invention.

The plug-in grid system of the present invention has the following effects.
(1) Even in a system with a high load, there are a part with a high load and a part with a low load inside, but only a part with a high load can be distributed and the efficiency is increased.
(2) Since the inside of processing with high load is replaced with a plug-in that only receives the request for the grid system that distributes the load and the processing result, the grid without modifying the system of the high load system You can enjoy the benefits.
(3) If the computer resources used in the resource pool are autonomously determined and increased or decreased, the user does not need to define and the processing is simple.

  In this specification, a plug-in refers to a small program for providing an additional function to application software such as business software, and particularly refers to a program that enables bidirectional communication between a business program and a resource pool. .

  Hereinafter, grid computing according to an embodiment of the present invention will be described with reference to the drawings. The grid computing according to the present embodiment is characterized by applying a plug-in.

  FIG. 1 is a functional block diagram showing a configuration example of a grid computing system according to the present embodiment. As shown in FIG. 1, the grid computing according to this embodiment includes a business system 101, a resource pool 104, and a management console 107. A large number of existing business programs 102 are installed in the business system 101, and a plug-in 103 is newly incorporated so that communication (L1) with the resource pool 104 can be performed. The resource pool 104 includes a plurality of processing nodes 105 and a load balancer 106 connected to each of these processing nodes 105. When there is a processing request from the plug-in 103, processing is performed at the processing node 105. The load balancer 106 performs processing allocation to the processing node 105 and communication with the management console 107 (L2, L3). The management console 107 has means for communicating with a resource pool 104 (not shown), and includes a GUI 108 for operating the grid computing system and an output unit (not shown) that outputs an operation / failure report 109. Yes.

  FIG. 2 is a diagram illustrating a detailed configuration example of the business system 101 illustrated in FIG. 1. The business system 101 shown in FIG. 2 includes a business program 203 with a low load and a business program 202 with a high load. The plug-in 103 can be incorporated into the business program 202 with a high load, and means for transmitting / receiving a processing request from the business program 202 with a high load to the resource pool 104 via the network 205. I have.

  FIG. 3 is a diagram illustrating a configuration example of the resource pool 104 illustrated in FIG. The resource pool 104 includes a processing node 105 and a load balancer 106. The processing request L1 from the plug-in 103 is received by the load balancer 106 and transmitted to each processing node including the processing node 105 to perform processing. Further, the processing result from the processing node 105 is returned to the plug-in 103 via the load balancer 106. In addition, the load balancer 106 includes a management unit (not shown) that manages the entire resource pool 104. As a result, it is possible to change the load distribution to the processing nodes in response to a configuration management request from the management console 107 (FIG. 1). Furthermore, a means for transmitting operation / failure information in the resource pool 104 to the management console 107 is provided. Reference numerals L1 (processing request and response), L2 (event), and L3 (management) will be described later.

  FIG. 4 is a diagram showing a configuration example of the management console 107 described in FIG. The management console 107 includes a controller 406 that manages the entire grid computing system, and a GUI 405 that configures an interface for a user to operate the system. The GUI 405 accepts a grid configuration management operation by a user and an operation failure report output operation. The controller 406 requests a configuration management operation from the GUI 405 to the resource pool as a configuration management request L3. Further, when an operation / failure report output request is received from the GUI 405, a report 404 can be created and output based on the operation failure event L2 from the resource pool. Report 404 includes information such as system performance and faults.

  FIG. 5 is a diagram illustrating a detailed configuration example of the actual business system 101. As shown in FIG. 5, the business system 101 is a system based on Java, which is a base software that absorbs differences between OSs, for plug-in incorporation. Hereinafter, a case where a business program for processing XML data is running will be described as an example. First, it is assumed that the business system 101 according to the present embodiment employs Java as the base software 502. The business program 102 performs an XML operation from the reading of the XML to the business data processing by using the XML processing program possessed by the base software 502. The plug-in 103 is configured so as to be incorporated in the business system 101 in such a manner that the XML processing program is overwritten. Therefore, the plug-in 103 for grid connection can be incorporated without modifying the business program 102. That is, the plug-in 103 does not process all the programs internally, but selectively requests a program with a high load from the external resource pool 104 by communication. At this time, the business program 102 need not be modified at all.

  FIG. 6 is a diagram for explaining the processing of the plug-in 103 shown in FIG. 5 as an example in which a grid is applied to the XML parsing processing in the business program 102. As illustrated in FIG. 6, when the plug-in 103 receives a processing request from the business program 102, the plug-in 103 determines whether the request can be distributed by the resource pool 104. At this time, if the request can be distributed, the plug-in communication function 603 is called to transfer the processing request to the resource pool (grid system) 104 and receive the result from the resource pool 104.

  On the other hand, when a request not to be distributed is received from the business program 102, processing is performed using a conventional function (API originally possessed by JAVA) that the plug-in 103 has left unmodified. With this configuration, when viewed from the business program 102, the same processing result can be obtained before and after the plug-in 103 is installed.

  Note that which program has a high load may be preliminarily estimated and specified by the user in the business program (for example, a process in which a high load such as an XML process is predicted) or a program load. May be monitored as needed, and based on the result, a high-load program may be picked up and sent to the resource pool by communication. The latter method has an advantage that load distribution can be performed in real time.

  FIG. 7 is a diagram for explaining communication between the plug-in 103 and the resource pool 104. The business program 102 in the business system 101 makes a processing request to the plug-in 103. At this time, the plug-in 103 performs compression when transmitting data to the resource pool 104, and the compressed data 702 is transmitted toward the resource pool 104. On the resource pool 104 side, processing such as XML is performed after the compressed data 702 is decompressed and restored. Also in the reply from the resource pool 104 to the plug-in 103, the same compression processing is performed, and the compressed data 704 is transmitted to the plug-in 103. Thus, by performing data compression, there is an advantage that the load on the network (communication line) is reduced and more processing requests can be distributed in the resource pool 104. Note that XML processing or the like is performed when the amount of data is reduced when compression is performed, and processing such as jpg is performed without performing compression. As described above, it is preferable to determine whether or not the compression processing is performed according to data.

  Hereinafter, a processing example in the resource pool will be described. FIG. 8 is a diagram showing an initial amount management method for processing nodes in the resource pool 104. In the grid system according to the present embodiment, an initial amount can be assigned to the resource pool 104 for each type of plug-in. In addition, the processing nodes in the resource pool 104 (indicated by 804a and 804b in the figure, and in many cases, actually have more processing nodes) are shown in the figure by the management console 107. In addition, it is virtually managed as each cell defined by a mesh line.

  When an initial amount allocation instruction is given from the management console 107 to the resource pool 104 as the configuration management request 802, the resource pool 104 allocates the instructed processing nodes, for example, 804a and 804b as initial amounts. The allocation of the initial amount is realized when the load balancer 106 described with reference to FIG. 3 passes all processing requests to the processing nodes 804a and 804b.

  According to the above method, each processing node in the resource pool 104 does not need to manage the configuration of the resource pool 104 with consciousness, and can manage the entire grid system in a simple manner using only the management console 107. it can.

  FIG. 9 is a diagram illustrating an example of information transmission to the management console when the load increases in the resource pool in the state of FIG. Each processing node in the resource pool 104 can monitor its own load. When the load increases, the processing node 804b that has not reached the predetermined load does not report the load. Further, the processing node 804a exceeding the determined load transmits a load report 902 to the management console 107 through the load balancer 106 described with reference to FIG. The management console 107 that has received the load report 902 executes processing node addition processing. The processing node addition processing and node reduction processing will be described with reference to FIGS. 10 to 13 described later.

  FIG. 10 is a diagram illustrating an example of a processing node addition / deletion method. A load report 902 is transmitted to the management console 107 in accordance with the method described in FIG. 9 to the processing node 804 a that has become highly loaded in the resource pool 104. The management console 107 that has received the load report 902 assigns the nodes adjacent to the processing node 804a having the high load (four nodes on the top, bottom, left and right in the figure) 1004 as additional processing nodes. As a result, as shown in the resource pool 104a, the network of the resource pool 104 that is virtually managed in the management console 107 includes five nodes, that is, 804a and four adjacent nodes. The management console 107 transmits a node addition request 1007 to the resource pool 104a, and the resource pool 104a that has received the node addition request 1007 newly uses the processing node 1004. Further, when the load decreases, the reverse processing is performed to decrease the number of nodes. There are two types of management related to the increase / decrease of the node, and there are a method of increasing / decreasing preferentially from the adjacent node and a method of instructing from the management console 107.

  Details of the operation example of the resource pool at the time of adding / deleting the processing node will be described below with reference to FIGS. FIG. 11 is a diagram illustrating a detailed processing example of the processing node addition processing described with reference to FIG. 10. As shown in FIG. 11, when the processing node load increases in the first processing node 104 a, a load increase report 1102 is transmitted to the management console 107. The management console 107 determines to newly assign a node adjacent to the first processing node 104a whose load has increased (1105). As a result, the virtual network configuration managed by the management console 107 changes as indicated by reference numeral 104. The management console 107 transmits a processing node change (addition) request 1106 to the load balancer 106 in order to realize this virtual configuration. The load balancer 106 adjusts the load balance in order to add the second processing node 104b, and the balance between the first processing node 104a and the second processing node (additional node) 104b is, for example, 5 to 5. Distribute to achieve an even balance. As a result, the load is equally distributed between the first processing node 104a and the second processing node 104b (1108).

FIG. 12 is a diagram illustrating a detailed example of processing node reduction processing described in FIG. 10. When the processing node load decreases in the processing node 104a, a load decrease report 1202 is transmitted to the management console 107. The management console 106 determines to release the node adjacent to the processing node 104a whose load has decreased (1205). As a result, a virtual network configuration managed by the management console is as indicated by 1204. In order to realize this virtual configuration, the management console 107 transmits a processing node reduction request 1206 to the load balancer 106. The load balancer 106 adjusts the load balance to delete the processing node 104b, and distributes the balance between the processing nodes 104a and 104b to 10-0. In this way, no processing request is received at the processing node 104b, processing is performed exclusively at the processing node 104a, and the load is concentrated on the processing node 104a. it can.

As described above, the plug-in grid system according to the present embodiment has the following effects.
(1) Even in a system with a high load, if there are a part with a high load and a part with a low load, only the part with a high load can be distributed.
(2) Since the inside of processing with high load is replaced with a plug-in that only receives the request for the grid system that distributes the load and the processing result, the grid without modifying the system of the high load system You can enjoy the benefits.
(3) The management of the grid system configuration is easy.
(4) The operating status of the grid system can be easily known.
(5) Although communication with the grid system is performed via a network, since data compression can be performed, the load on the network is minimal.
(6) The computer resources used in the resource pool can be autonomously determined and increased or decreased. In this way, the user does not need to define.

  The present invention is applicable to a grid computing system.

It is a figure which shows the example of 1 structure of the grid computing system by one embodiment of this invention. It is a figure which shows the structural example of the plug-in of the grid computing system by this Embodiment. It is a figure which shows the structural example of the resource pool of the grid computing system by this Embodiment. It is a figure which shows the structural example of the management console by this Embodiment. It is a figure which shows the example of a plug-in incorporation in the business system based on Java. It is a figure which shows the operation example of a plug-in. It is a figure which shows the communication system of a plug-in. It is a figure which shows the example of the management system of the processing node in a resource pool. It is a figure which shows the management system of load. It is a figure which shows the allocation method of a processing node. It is a figure which shows the example in the case of the load distribution system at the time of processing node allocation, and addition. FIG. 10 is a diagram illustrating an example of a load balancing method at the time of processing node allocation and deletion.

Explanation of symbols

DESCRIPTION OF SYMBOLS 101 ... Business system, 102 ... Business program, 103 ... Plug-in, 104 ... Resource pool, 105 ... Processing node, 106 ... Load balancer, 107 ... Management console, 108 ... GUI, 109 ... Operation / failure report, 202 ... Business program (High load), 203 ... business program (low load), 205 ... network, 404 ... operation / failure report, 405 ... GUI, 406 ... controller, 502 ... base software, 603 ... plug-in communication function, 604 ... plug-in 702: Communication from plug-in to resource pool, 704 ... Communication from resource pool to plug-in, 802 ... Configuration management request, 804 ... Assigned processing node, 902 ... Operation failure event, 1004 ... Added processing node, 1102 ... Load increase Report, 1105 ... Decision of management console, 1106 ... Configuration change request, 1108 ... Load flow after adding processing node, 1202 ... Load drop report, 1205 ... Decision of management console, 1206 ... Configuration change request, 1208 ... Processing Load flow after node reduction.

Claims (9)

A business system that operates a plurality of business programs, a resource pool that is network-connected to the business system and includes a load balancer that balances the load of a plurality of processing nodes, and a management console that controls the resource pool A grid management system comprising:
A grid computing system comprising a plug-in that is incorporated in a business system in units of the business program and enables load distribution by the grid system.   The plug-in is attached to a part with a high load in the business program, the plug-in and the resource pool are connected to the network, a processing request from the plug-in to the processing node, and a response to the processing request And a first communication unit for transmitting and receiving data between the plug-in and the resource pool based on the exchange. system.   And a second communication unit that transmits and receives data between the grid system and the grid management system, and the management console receives a load report of the resource pool by the second communication unit. The grid computing system according to claim 1, wherein load distribution to the processing nodes is determined based on a load situation based on the load report.   The plug-in includes replacement means for requesting processing to the resource pool via the first communication means instead of the processing by the business program by the base software serving as the base of the business program. The grid computing system according to any one of claims 1 to 3. The plug-in is
Means for determining whether distributed processing can be performed in the resource pool;
Means for requesting processing to the resource pool when it is determined that distributed processing is possible;
5. The grid computing system according to claim 4, further comprising means for processing using software before replacement by said replacement means when it is determined that distributed processing cannot be performed.   6. The resource pool and the management console further comprising management means for managing resources by virtually arranging the plurality of processing nodes in the resource pool in a network. The grid computing system according to item 1.   The grid computing system according to claim 6, further comprising means for each of the processing nodes arranged in a network form transmitting a load status of the processing node to the management console.   Furthermore, the management means autonomously adds processing nodes according to a report from the processing nodes when the load increases, and autonomously reduces processing nodes according to a report from the processing nodes when the load decreases. The grid computing system according to 6 or 7. A business system that operates a plurality of business programs, a resource pool that is network-connected to the business system and includes a load balancer that balances the load of a plurality of processing nodes, and a management console that controls the resource pool A grid management system, and a business system in a grid system configured to include:
A plug-in that is incorporated in the business program unit and enables load distribution by the grid system,
The plug-in is attached to a part with a high load in the business program, the plug-in and the resource pool are connected to the network, a processing request from the plug-in to the processing node, and a response to the processing request And a first communication means for exchanging data between the plug-in and the resource pool based on the exchange,
A business system comprising: the grid system; and second communication means for transmitting and receiving data between the grid system and the grid management system.

Images