JP2005332082A - Distributed processing system and distributed processing program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a distributed processing program which realizes efficient distributed processing and task sharing in many computers connected via a network. <P>SOLUTION: The distributed processing program uses a plurality of agent programs movable among a plurality of computers through a network communication channel so as to perform the distributed processing of tasks including multiple kinds of processing to be chronologically executed. Each agent program associates and stores, based on a movement table, a processing program for executing the tasks, processing data for executing the tasks, communication channel information indicating a communication channel to which the program belongs, the movement table and evaluation values in the storage device of a computer to which the program belongs after it moves (Step S110). Each agent program is subjected to disruption processing (Step S112) in accordance with an evaluation value, and processing for separating the communication channels is executed in case that the number of the agent programs belonging to the same communication channel is over a prescribed number (Step S116). <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a distributed processing system and a distributed processing program for performing distributed processing of one task using a large number of different computer resources in a system composed of a large number of computers connected by a network.

  With the spread of computers and networks, networks connected in various forms, consisting of many computers of various types, are increasing.

  Under such circumstances, in recent years, ubiquitous computing including a large number of computers has attracted attention (see, for example, Non-Patent Document 1). In ubiquitous computing, it is assumed that the network environment such as the distribution of computer resources is unknown or changes dynamically.

  In the optimization of distributed processing in such an environment, adaptive acquisition of the network structure and processing procedure is expected.

  On the other hand, as research on multi-agent systems (hereinafter referred to as MAS), for example, the gain, reward, and evaluation, which are information within the range perceived by agents, are defined as individual optimality. Dynamically Separating Genetic Algorithms (below) that can simulate the improvement of system optimization by agent learning when each agent defines the total gain gained from each action as system optimization. , DS-GA) ”(for example, see Non-Patent Document 2) and“ Dynamically Separating Learning Algorithms (hereinafter, referred to as DS-LA) ”(for example, see Non-Patent Document 3) ) Has been proposed.

This DS-GA is a genetic algorithm (hereinafter referred to as GA) that dynamically separates individuals into groups called “colony” according to the number of individuals. Other genetic algorithms for separating individuals include the island model GA, but the DS-GA differs in that the separation state changes dynamically according to the number of individuals.
Mark Weiser, “The Computer for the 21st Century,” Scientific American, Vol. 265, No. 3, pp. 99-104, 1991. Koichi Nakayama, Hirokazu Matsui, Yujihiko Nomura: Proposal of Dynamically Isolated GA (DS-GA), Transactions of Information Processing Society of Japan: Mathematical Modeling and Application, Vol.43, No.SIG10 (TOM7), pp .95-109, Nov.2002. K. Nakayama, H. Matsui, K. Shimohara, O. Katai, “Proposal of Dynamically Separating Learning Algorithm for Interactive Computers (DS-LA): Optimization of collective performance in networked computers,” The journal of three dimensional images, Vol. 18, No. 1, pp. 87-94, 2004.

  However, conventionally, when a large number of computers exist on the computer network as described above, a method for efficiently performing distributed processing of one task using a large number of different computer resources by using MAS is as follows. There was a problem that was not always obvious.

  The present invention has been made to solve the above problems, and an object of the present invention is to provide a distributed processing system that realizes efficient distributed processing and task sharing in a large number of computers connected via a network. It is to provide a distributed processing program.

  In order to achieve such an object, a distributed processing system of the present invention is a distributed processing system for distributed processing of a predetermined task including a plurality of processes executed in time series, and is mutually connected via a network. A plurality of information processing devices capable of communicating with each other, and each information processing device performs communication processing for communicating with a network on a communication channel instructed to select among the plurality of channels, and corresponding processing among the plurality of processing. Information processing means for executing and storage means for storing information for processing, wherein the information processing means is a plurality of agent arithmetic processing means movable between a plurality of information processing devices via a communication channel Each of the agent calculation processing means is based on a movement table that associates an information processing apparatus that currently belongs to a task and a processing apparatus that is a movement destination. The processing means for executing the task, the processing data for executing the task, the communication channel information indicating the belonging communication channel, the movement table, the evaluation, Task processing for executing processing corresponding to the information processing device to which the processing belongs among a plurality of processing of the task based on information stored in the storage means of the information processing device to which the information belongs and moving means for storing the values in association with each other The same communication with the means, the evaluation value updating means for updating the evaluation value in response to completion of the task processing, and the division processing means for performing the division processing of the agent arithmetic processing means in accordance with the value of the evaluation value When the number of agent calculation processing means belonging to a channel exceeds a predetermined number, a plurality of agent operations belonging to the same communication channel are processed. And a communication channel separating means for performing processing for separating the communication channels corresponding to the processing means.

  Preferably, each agent computation processing means further includes an erasure instruction means for instructing the agent computation processing means belonging to the selected communication channel among the communication channels to perform erasure processing.

  Preferably, the division processing means generates an agent calculation processing means that takes over the processing program associated with the agent calculation processing means before division, the processing data, the communication channel information, and the movement table.

  Preferably, the division processing means includes a mutation processing means for updating the movement table with a predetermined mutation probability.

  Preferably, the moving unit further includes task designation information for designating a task to be processed by the corresponding agent calculation processing unit among a plurality of tasks, and a task processing procedure table in which a plurality of processes corresponding to each of the plurality of tasks are associated with each other. Is stored in the storage means of the information processing equipment to which the task belongs after the movement, and the task processing means executes processing corresponding to the information processing equipment to which the task specified by the task specification information belongs.

  According to another aspect of the present invention, a plurality of agents that can move between a plurality of information processing devices via a communication channel of a network in order to perform distributed processing of a predetermined task including a plurality of processes executed in time series A distributed processing program that uses a program, and each agent program instructs selection of a plurality of channels based on a movement table that corresponds to a task and associates an information processing device that currently belongs to a processing device that is a movement destination. A communication program that communicates with a network through a communication channel, and stores a processing program for executing a task, processing data for executing the task, and communication channel information indicating the communication channel to which the task belongs. , Movement step to store the movement table and evaluation value in association with each other, and the information to which it belongs Based on the information stored in the storage means of the physical device, out of a plurality of tasks processing, a task processing step for executing processing corresponding to the information processing device to which it belongs, and in response to completion of the task processing, An evaluation value update step for updating the evaluation value, a split processing step for splitting the agent arithmetic processing means according to the value of the evaluation value, and a predetermined number of a plurality of agent arithmetic processing means belonging to the same communication channel In response, the information processing device is caused to execute a communication channel step for performing processing for separating communication channels corresponding to a plurality of agent calculation processing means belonging to the same communication channel.

  Preferably, each agent program further includes an erasure instruction step for instructing an erasure process to an agent program belonging to the selected communication channel among the communication channels.

  Preferably, the division processing step includes a step of generating an agent program that takes over the processing program associated with the agent calculation processing means before the division, the processing data, the communication channel information, and the movement table.

  Preferably, the division processing step includes a mutation processing step of updating the movement table with a predetermined mutation probability.

  Preferably, the moving step further includes task designation information for designating a task to be processed by the corresponding agent calculation processing means among a plurality of tasks, a task processing procedure table in which a plurality of processes corresponding to each of the plurality of tasks are associated with each other. Is stored in the storage means of the information processing device to which the task belongs after the movement, and the task processing step executes processing corresponding to the information processing device to which the task designated by the task designation information belongs.

  Embodiments of the present invention will be described below with reference to the drawings.

(Outline of the present invention)
In the present invention, a large number of computers connected by a network or the like use a large number of autonomously executing programs called agents, and efficiently distribute one task using a large number of different computer resources. .

  As will be described below, in the present invention, as a method for optimizing the distributed processing in the entire computer network, a multi-agent system (MAS) learning method, in particular, DS-- which is one of the above-described MAS learning methods. Use LA.

  In DS-LA, agents are dynamically separated into groups called colonies. Each agent belongs to a specific group. In DS-LA, the interaction of agents is restricted within the group in principle. Also imitate the group of other agents who get high performance. As a result, the composition of the group changes dynamically.

  More specifically, in the present invention, as an example of this “group”, the entire network is virtually divided into several networks called channels that have no mutual interference. When the number of agents belonging to each channel increases, agents belonging to that channel are dynamically separated.

  In the present invention, in order to achieve the purpose of distributed processing of one task using a large number of different computer resources, a large number of agents belonging to the computer use the destination computer resources while moving between the computers. Process the task. Each agent has information such as processing data and its processing program (processing program code), a destination for each condition, a communication channel used for movement, and a self-evaluation value. The computer provides a part of the computer resources to the agent to which it belongs. Unlike computer viruses, agents process tasks only with the provided computer resources. The agent is moved through the communication channel to which each agent belongs.

  In learning by DS-LA, agents are separated for each channel, and those whose self-evaluation values updated based on only information that can be perceived by themselves on the network exceed a specified value are split. Channels in which the number of agents exceeds the specified value are further separated dynamically.

  The efficiency of distributed processing is achieved by learning of this agent, but the processing order (serial distribution) is also optimized by learning of agent at the time of efficiency improvement. Accordingly, the network structure itself, in other words, the computer group structure as described above can be optimized by agent learning.

[Embodiment of the Invention]
FIG. 1 is a conceptual diagram showing an example of a distributed processing system 1000 according to the present invention.

  Referring to FIG. 1, system 1000 includes computers 10.1 to 10. that can communicate with each other via a wireless network. N and computers 10.1-10. And a wireless access point 20 that can connect N to each other by a wireless network in a so-called “infrastructure mode”.

  Here, the wireless access point 20 and the computers 10.1 to 10. N is assumed to perform communication through one of channels obtained by further dividing a predetermined communication frequency band into a plurality of frequency bands. It is assumed that a computer that communicates on a certain channel and a computer that communicates on another channel are separated as a network. Further, as will be described later, each computer 10.1-10. N can change the channel to be communicated based on the learning process according to the agent program belonging to the user. Therefore, the wireless access point 20 has a configuration capable of simultaneously performing communication on a plurality of channels with one unit. As will be described later, communication between different channels is normally not permitted, but communication between different channels is also permitted in predetermined broadcast communication.

  Further, when considering an example of such a network, the configuration of the network is not limited to the wireless network as shown in FIG. 1, and may be a wired network as long as each computer can select a channel with which the computer communicates. .

  FIG. 1 is only one specific example for explaining the present invention, and more generally, the present invention can be applied to a computer network system that satisfies the following conditions.

(Conditions of the system to which the present invention is applied)
i) There are many computers connected by a network.

  ii) The hardware configuration and processing capability of the computer may be inhomogeneous.

  iii) There are many programs called autonomous and mobile agents having "data to be processed" and "processing systems (programs, etc.)", and tasks assigned to these agents may be inhomogeneous.

iv) The computer provides computer resources to the agent. Agents process autonomously in computers using their task processing system [Hardware configuration]
FIG. 2 is a diagram illustrating an example of the appearance of the computer 10.1 illustrated in FIG.

  Referring to FIG. 2, this computer 10.1 stores information in a disk drive 108 for reading information on a disk 118 such as a CD-ROM (Compact Disc Read-Only Memory) and a flexible disk (FD) 116. A computer main body 102 having an FD drive 106 for reading and writing, a display 104 as a display device connected to the computer main body 102, a keyboard 110 and a mouse 112 as input devices also connected to the computer main body 102, A wireless communication device 114 connected to the computer main body 102.

  FIG. 3 is a block diagram showing the configuration of the computer 10.1.

  As shown in FIG. 3, in addition to the disk drive 108 and the FD drive 106, the computer main body 102 constituting the computer 10.1 includes a CPU (Central Processing Unit) 120 and a ROM (ROM) connected to the bus BS, respectively. It includes a memory 122 including a read only memory (RAM) and a random access memory (RAM), a direct access memory device, for example, a hard disk 124, and a communication interface 128 for exchanging data with the wireless communication device 114. A disk 118 is loaded in the disk drive 108. An FD 116 is attached to the FD drive 106.

  The disc 118 is a medium other than a CD-ROM, such as a DVD-ROM (Digital Versatile Disc) or a memory card, as long as it can record information such as a program installed in the computer main body. In this case, the computer main body 102 is provided with a drive device that can read these media.

  The main part of the distributed processing system of the present invention is composed of computer hardware and software executed by the CPU 120. Generally, such software is stored and distributed in a storage medium such as the disk 118 or the FD 116, read from the storage medium by the disk drive 108 or the FD drive 106, and temporarily stored in the hard disk 124. Alternatively, when the device is connected to the network, it is temporarily copied from the server on the network to the hard disk 124. Then, the data is further read from the hard disk 124 to the RAM in the memory 122 and executed by the CPU 120. In the case of network connection, the program may be directly loaded into the RAM and executed without being stored in the hard disk 124.

  The computer hardware itself and its operating principle shown in FIGS. 2 and 3 are general. Therefore, the most essential part of the present invention is software stored in a storage medium such as the FD 116, the disk 118, and the hard disk 124.

  As a general tendency, various program modules are prepared as a part of a computer operating system, and an application program generally calls a module in a predetermined arrangement and advances the processing when necessary. In such a case, the software itself for realizing the distributed processing system does not include such a module, and the distributed processing system is realized only when the computer cooperates with the operating system. However, as long as a general platform is used, it is not necessary to distribute software including such modules, and the software itself not including these modules and the recording medium storing the software (and the software distributes on the network). Data signal) can be considered to constitute the embodiment.

  Other computers 10.2 to 10. N also has basically the same configuration as the computer 10.1.

[Basic Principle of Distributed Processing System According to the Present Invention]
[Overview of DS-LA]
FIG. 4 is a diagram illustrating the concept of DS-LA.

  DS-LA is an algorithm having an overwriting of an action strategy by an updated action strategy and a dynamic separation by an affiliation change. As an overwriting of the action strategy described below, the table indicating the agent's destination is updated, and agents whose evaluation value associated with them exceeds a predetermined value are split. The table (action strategy) shown is taken over. Further, agents belonging to channels whose number of agents belonging to a channel exceeds a predetermined number are separated into two channels as shown in FIG. In such dynamic separation, each agent belongs to one of a virtual group called a plurality of channels, and in principle does not interact with agents belonging to other channels. If the number of computers belonging to a channel increases beyond a certain limit, for example, half of the agents belonging to that channel change their affiliation to a new channel.

[DS-LA algorithm used for simulation experiments]
FIG. 5 is a flowchart for explaining the DS-LA algorithm.

  Hereinafter, the DS-LA algorithm will be described in more detail with reference to FIG.

  Note that the processing performed by each agent below is executed by the CPU 120 of the computer to which the agent belongs at that time based on the processing program of the corresponding task that moves accompanying the agent.

(1) Initialization First, the system 1000 includes N computers (N: a natural number of 2 or more) to which DS-LA is applied, for example, five computers {A, B, C, D, E}, a predetermined number. For example, assume that there are 10,000 communication channels and a predetermined number, for example, 100,000 agents. That is, in the following description, the computers 10.1 to 10. Assume that computers A to E exist in the system 1000 as N.

  In each of the computers A to E, first, the following initial setting process is performed (step 100).

  Each agent belongs to a randomly selected computer and channel, and has a table indicating its own evaluation value, the type of task to be processed, task data to be processed, and the computer to be moved.

  FIG. 6 is a diagram showing an example of a data table stored in the hard disk 124 of the computer A when a certain agent belongs to the computer A.

  That is, this agent stores, for example, in a predetermined storage area of the hard disk 124, a variable BLpc indicating a computer to which the agent belongs, a task type Tk to be processed, data D of a task to be processed, an assigned channel number Nchan (i), The evaluation value U (i, t) and the number Nag of agents belonging to the same channel as the self acquired as described later are stored. Further, a variable t indicating the passage of unit time of processing is set to 1.

  In this example, “variable BLpc” takes any value of the computers {A, B, C, D, E}.

  “Affiliation channel number Nchan (i)” is a number indicating a channel to which the i-th agent belongs, and here, Nchan (i) ∈ {1, 2,..., Nchansum (= 10000)}. To do.

  “Evaluation value U (i, t)” is a learning index of each computer based on its own cumulative evaluation, and initial value U (i, 0) is, for example, “10”.

  Further, FIG. 7 is a diagram illustrating an example of a task processing procedure table and a migration table indicating a migration destination computer that are stored in the hard disk 124 of the computer A when a certain agent belongs to the computer A.

  As shown in FIG. 7A, when there are, for example, two tasks X and Y as the types of tasks, the “task processing procedure table” includes processing as the first to fifth processing in task X. a, process b, process c, process d, and process e are performed in this order, and in task Y, process a, process c, process e, process b, and process d are performed in this order as the first to fifth processes. Stores information that is performed in

  On the other hand, as shown in FIG. 7B, in the “movement table”, an agent performing a task I (I = X, Y) has a computer J (J = A, B, C, D) at a certain time step. , E), information T (J, I) (where T (J, I) ∈ {A, B, C, D, E) indicating to which computer to move to in the next time step. }) Is stored.

  The values of the “movement table”, the assigned channel number Nchan (i), and the evaluation value U (i, t) change as a result of learning.

  As an initial “movement table”, T (J, X) and T (J, Y) have values randomly selected from {A, B, C, D, E}, and the initial channel It shall have the channel number chosen at random as a number. Each agent inquires about another agent existing in the same channel as the agent, and the number Nag obtained as a result of the inquiry is recorded in the hard disk 124.

  Referring to FIG. 5 again, following the initial setting, the following processing is performed.

  (2) Task acquisition: An agent that does not hold a task or completes a task obtains a randomly selected task X or Y (step S102). In other words, according to a random number generated by the agent, one of the tasks is selected for processing.

  (3) Task processing: Each agent performs a processing procedure according to FIG. 7A according to the task selected in step S102 (step S104). Here, the agents belonging to the computers A, B, C, D, and E can perform the process a, the process b, the process c, the process d, and the process e, respectively. At this time, if the computer to which the agent belongs does not correspond to the processing to be performed at that time, the processing for the unit time is not performed for the processing of the unit time, and the processing proceeds to the next step. .

  Each computer can determine which processing is possible by storing each computer as information, and the agent inquiring the computer. Alternatively, each agent may move while holding a table of what processing is possible on each computer, or each agent diagnoses hardware information of the computer to which it currently belongs and It may be determined each time whether the processing to be performed in step S3 can be executed by the computer to which the user currently belongs.

  (4) Task completion determination: Each agent determines whether the task has been completed (step S106), and the agent that has completed the processing of the task updates the evaluation value, for example, according to the following equation (step S108).

U (i, t) = U (i, t−1) + (50−t × 2)
In other words, the evaluation value increases as the agent completes the task in a shorter time, and the evaluation value of the agent that took a predetermined time or longer to complete the task falls below the initial value.

  For example, if the agent that has finished the task notifies the predetermined computer of the processing result, the user can view the processing result of the task.

  On the other hand, if the task has not been completed, the process proceeds to step S110.

  (5) Migration: Each agent has a migration table value (T (J, I) ∈ {A, B, C, D, E}) for each of 10 state spaces (task type × computer type to which it belongs). And moves with the task held by itself (step S110).

  That is, each agent moves with the information of the data table shown in FIG. 6 and the information of the “task processing procedure table” and “movement table” shown in FIG. In other words, the source CPU 120 sends the agent processing program itself, the information in the data table shown in FIG. 6 and the information in the “task processing procedure table” and “movement table” shown in FIG. The data is transmitted via the corresponding communication channel and stored in the hard disk 124 of the destination computer. The destination CPU 120 executes processing based on the processing program of the agent that has moved and the accompanying information.

  (6) Agent splitting: Subsequently, each agent executes splitting processing (step S112).

  FIG. 8 is a flowchart showing the procedure of such division processing. Referring to FIG. 8, when the evaluation value of the agent exceeds a predetermined value, for example, twice the initial evaluation value (step S200), the agent is divided (step S202). At this time, evaluation values are also distributed (step S202). For example, the evaluation value of each agent after division is half of the evaluation value before division, and the movement table is taken over.

  That is, in the splitting of the agent, the agent generates on the hard disk 124 of the computer to which the agent belongs the agent that succeeded the processing program, processing data, communication channel information, and movement table associated with the agent.

  Subsequently, the values in the table are changed according to the mutation probability Pmut (for example, Pmut = 0.05). At this time, when mutation is to be performed according to the mutation probability, for example, both T (J, X) and T (J, Y) are randomly selected from {A, B, C, D, E}. A value shall be selected.

  (7) Channel dynamic separation: Again referring to FIG. 5, each agent determines whether or not the number of agents to which it belongs has exceeded the limit number of individuals (for example, 10) (step S114). If the limit number is exceeded, the agent belonging to the channel is separated into two channels (step S116). At this time, since agents belonging to the same channel can communicate with each other, for example, the wireless access point 20 is inquired about a channel that is currently idle, and then exchanges information with each other so that the previous channel is obtained. Determine the remaining agents and agents to move to a new channel (empty channel). At this time, a condition may be added so that the two groups have substantially the same number.

  Further, for example, among agents that have separated channels, agents remaining in the original channel randomly select other channels by mutual communication, and by broadcast communication to all agents belonging to the selected channel. Instructs erasure. The agent belonging to the channel instructed to delete deletes itself (step S118).

  Subsequently, the variable t is incremented, and the process returns to step S102.

  A period during which the processing from step S102 to step S118 is performed is referred to as “unit time”.

The operation of the agent as described above, each agent learns the optimum one moving table in this case 5 ¹⁰ ways.

(Results of simulation experiment)
FIG. 9 is a diagram showing the change over time in the individual number ratio for each state space in the movement table as one of the experimental results.

  For example, for an agent that executes task X on computer A, the next destination T (A, X) converges on computer B over time. Other computers B to E converge to a specific destination.

  Similarly, for the agent that executes the task Y on the computer A, the next destination T (A, Y) converges on the computer C as time passes. Also in the task Y, the movement destinations in the other computers B to E converge to a specific computer.

  FIG. 10 is a diagram showing a network of task processes generated when the movement destination converges in this way.

  As shown in FIG. 10, for task X, a network of computer A → computer B → computer C → computer D → computer E → computer A is created. This is a network corresponding to the processing order of task X.

  Similarly, for task Y, a network of computer A → computer C → computer E → computer B → computer D → computer A can be created. This is also a network corresponding to the processing order of task Y.

  FIG. 11 is a diagram showing temporal changes in the average evaluation value of agents. As time elapses, the average value of the evaluation values of the agents converges to a constant value as time elapses.

  From these results shown in FIG. 9 to FIG. 11, it can be seen that each agent autonomously selects an optimum destination according to the type of task and the computer to which the agent belongs and obtains a high evaluation value. .

  Numerous agents whose task processing procedures and network status are unknown learn by DS-LA, and it is possible to generate emergent networks from experimental results of moving tasks between computers and processing tasks. I know that there is.

  The embodiment disclosed this time should be considered as illustrative in all points and not restrictive. The scope of the present invention is defined by the terms of the claims, rather than the description above, and is intended to include any modifications within the scope and meaning equivalent to the terms of the claims.

It is a conceptual diagram which shows an example of the distributed processing system 1000 of this invention. It is a figure which shows the external appearance of the computer 10.1 shown in FIG. 1 as an example. It is a figure which shows the structure of the computer 10.1 in block diagram format. It is a figure which shows the concept of DS-LA. It is a flowchart for demonstrating the algorithm of DS-LA. 4 is a diagram illustrating an example of a data table stored in a hard disk 124 of a computer A when an agent belongs to the computer A. FIG. It is a figure which shows the example of the movement procedure table which shows the process sequence table of a task, and a movement destination computer. It is a flowchart which shows the procedure of a division | segmentation process. It is a figure which shows the time change of the individual number ratio for every state space in a movement table. It is a figure which shows the network of the task process produced | generated when a movement destination converges. It is a figure which shows the time change of the average evaluation value of an agent.

Explanation of symbols

  10.1-10. N computer, 20 wireless access point, 102 computer main body, 104 display, 106 FD drive, 108 disk drive, 110 keyboard 110, 112 mouse, 114 wireless communication device, 118 CD-ROM, 120 CPU, 122 memory, 124 hard disk, 128 Communication interface, 1000 system.

Claims (10)

A distributed processing system for distributed processing of a predetermined task including a plurality of processes executed in time series,
It has multiple information processing devices that can communicate with each other via a network.
Each of the information processing devices
A communication means for communicating with the network in a communication channel instructing selection from a plurality of channels;
Information processing means for executing a corresponding process among the plurality of processes;
Storage means for storing information for performing the processing,
The information processing means includes a plurality of agent arithmetic processing means capable of moving between the plurality of information processing devices via the communication channel,
Each of the agent calculation processing means
To communicate based on a movement table that associates the information processing device that currently corresponds to the task and the processing device that is the movement destination, and executes the task in the storage unit of the information processing device that belongs after the movement The processing program, the processing data for executing the task, the communication channel information indicating the communication channel to which the task belongs, the moving table, and a moving means for storing the evaluation value in association with each other,
Based on information stored in the storage means of the information processing device to which the task belongs, out of the plurality of processes of the task, task processing means for executing processing corresponding to the information processing device to which the task belongs,
An evaluation value updating means for updating the evaluation value in response to completion of processing of the task;
In accordance with the value of the evaluation value, splitting processing means for splitting the agent arithmetic processing means,
The communication channels corresponding to the plurality of agent calculation processing units belonging to the same communication channel are separated in accordance with the number of the plurality of agent calculation processing units belonging to the same communication channel exceeding a predetermined number. A distributed processing system including communication channel separation means for performing processing.   2. The distributed processing system according to claim 1, wherein each of the agent computation processing units further includes an erasure instruction unit that instructs the agent computation processing unit belonging to a selected communication channel of the communication channels to perform erasure processing. The division processing means is
2. The distributed processing system according to claim 1, wherein an agent arithmetic processing unit that inherits the processing program associated with the agent arithmetic processing unit before division, the processing data, the communication channel information, and the movement table is generated. 3. . The division processing means is
The distributed processing system according to claim 3, further comprising a mutation processing unit that updates the movement table with a predetermined mutation probability. The moving means further includes a task processing procedure table in which task designation information for designating a task to be processed by the corresponding agent calculation processing means among a plurality of tasks is associated with the plurality of processes corresponding to the plurality of tasks. Are stored in the storage means of the information processing equipment to which the user belongs after the movement,
The distributed processing system according to claim 1, wherein the task processing unit executes a process corresponding to the information processing device to which the task specified by the task specifying information belongs. A distributed processing program using a plurality of agent programs that can move between a plurality of information processing devices via a network communication channel in order to perform a distributed processing of a predetermined task including a plurality of processes executed in time series. And
Each said agent program
Based on a movement table that associates the information processing apparatus that currently belongs to the task and the processing apparatus that is the movement destination, communicates with the network through a communication channel instructed to be selected from a plurality of channels, and belongs after movement In the storage unit of the information processing device, a processing program for executing the task, processing data for executing the task, communication channel information indicating a communication channel to which the task belongs, the movement table, an evaluation value, A moving step for storing and associating,
Based on information stored in the storage means of the information processing device to which the task belongs, a task processing step of executing processing corresponding to the information processing device to which the task belongs among the plurality of processes of the task;
An evaluation value update step of updating the evaluation value in response to completion of processing of the task;
In accordance with the value of the evaluation value, a split processing step for performing split processing of the agent arithmetic processing means,
The communication channels corresponding to the plurality of agent calculation processing units belonging to the same communication channel are separated in accordance with the number of the plurality of agent calculation processing units belonging to the same communication channel exceeding a predetermined number. A distributed processing program for causing the information processing device to execute a communication channel step for processing.   The distributed processing program according to claim 6, wherein each of the agent programs further includes an erasure instruction step for instructing the agent program belonging to a selected communication channel among the communication channels to perform erasure processing. The splitting step includes
The distributed processing according to claim 6, comprising a step of generating an agent program that takes over the processing program associated with the agent operation processing means before division, the processing data, the communication channel information, and the movement table. program. The splitting step includes
The distributed processing program according to claim 8, further comprising a mutation processing step of updating the movement table with a predetermined mutation probability. The movement step further includes a task processing procedure table in which task designation information for designating a task to be processed by the corresponding agent arithmetic processing means among a plurality of tasks is associated with the plurality of processes corresponding to the plurality of tasks. And storing in the storage means of the information processing equipment that belongs to after the movement,
The distributed processing program according to claim 6, wherein the task processing step executes processing corresponding to the information processing device to which the task specified by the task specification information belongs.

Images