Distributed processing system, distributed processing method, and recording medium

1. A system for distributed processing of tasks comprising:

an analysis node which analyzes input information; and

a management node which is connected to the analysis node via a network, and holds performance information representing performance of the analysis node, wherein

the management node includes performance information integration unit-integrating the performance information pertaining to the analysis node connected to the management node, and

the analysis node includes task determination unit determining,

based on a first cost, being calculated based on the performance information pertaining to the analysis node, of executing a task in the analysis node, and

a second cost being calculated based on the performance information pertaining to another of the analysis nodes different from the analysis node integrated by the performance information integration unit, and representing a cost of load distribution of distributing the task to the another analysis node and then executing the task,

whether to execute the task in the analysis node, or distribute the task to the another analysis node and then execute the task,

wherein

the task determination unit

calculates the first cost by use of a communication delay time and a processing delay time, which are included in the performance information, when the task is executed, and

calculates the second cost by use of a communication delay time and a processing delay time, which are included in the integrated performance information, when the task is distributed to the another analysis node and then executed.

2. The system for distributed processing of tasks according to claim 1 , wherein

the task determination unit in the analysis node

determines, when the task includes primary processing being processing of detecting a target from the input information, and secondary processing being processing of analyzing the target, whether to execute the secondary processing in the analysis node, or distribute the secondary processing to another of the analysis nodes and then execute the secondary processing, after executing the primary processing.

3. The system for distributed processing of tasks according to claim 1 , wherein

the management node further includes task saving unit saving the task load-distributed by the analysis node, and

the analysis node further includes task execution unit executing the task saved in the task saving unit when the analysis node has a surplus in a resource thereof.

4. The system for distributed processing of tasks according to claim 1 , wherein

the performance information integration unit integrates the performance information by statistically calculating the performance information.

5. The system for distributed processing of tasks according to claim 3 , wherein

the performance information integration unit integrates

the performance information of the analysis node, and

a performance information of at least one of other analysis node connected to the management node.

6. The system for distributed processing of tasks according to claim 1 , wherein

the task determination unit

calculates the first cost at a certain point of time by use of a history of the first cost representing a shift of the first cost during a certain period before a certain point of time, and

calculates the second cost at a certain point of time by use of a history of the second cost representing a shift of the second cost during a certain period before a certain point of time.

7. The system for distributed processing of tasks according to claim 1 , wherein

the performance information includes throughput of the analysis node, and a network delay.

8. A distributed processing method comprising:

by an analysis node,

analyzing input information;

calculating a first cost of executing a task by the analysis node itself, based on performance information representing performance of the analysis node itself;

calculating a second cost representing a cost of load distribution of distributing the task to another of the analysis nodes and then executing the task, based on performance information which is integrated by a management node holding performance information pertaining to one or more of the analysis nodes connected to the management node via a network, and which pertains to the another analysis node; and

determining, based on the calculated first cost and second cost, whether to execute the task by the analysis node itself, or execute load distribution of the task to the another analysis node,

wherein

when determining the task

calculating the first cost by use of a communication delay time and a processing delay time, which are included in the performance information, when the task is executed, and

calculating the second cost by use of a communication delay time and a processing delay time, which are included in the integrated performance information, when the task is distributed to the another analysis node and then executed.

9. A non-transitory computer readable recording medium recording a program which causes a computer functioning as an analysis node to execute:

processing of analyzing input information;

processing of calculating a first cost of executing a task by the computer itself, based on performance information representing performance of the computer itself;

processing of calculating a second cost representing a cost of load distribution of distributing the task to another of the analysis nodes and then executing the task, based on performance information which is integrated by a management node holding performance information pertaining to one or more of the analysis nodes connected to the management node via a network, and which pertains to the another analysis node; and

processing of determining, based on the calculated first cost and second cost, whether to execute the task by the computer itself, or execute load distribution of the task to the another analysis node,

wherein

when processing of determining the task

calculating the first cost by use of a communication delay time and a processing delay time, which are included in the performance information, when the task is executed, and

calculating the second cost by use of a communication delay time and a processing delay time, which are included in the integrated performance information, when the task is distributed to the another analysis node and then executed.