JPH05233596A - Resource uniform distribution system - Google Patents

Abstract

PURPOSE:To provide the resource uniform distribution system which unifomizes use rates of respective resources as much as possible without concentrating information of the whole of the system. CONSTITUTION:With respect to the resource distribution system in the multi- agent system where many autonomously operated computers called agents are connected by a network, this resource uniform distribution system is characterized by determining a resource management agent for each of plural kinds of resource and determining the target of an actually assigned capacity (use ratio) out of the overall capacity of the resource, which is managed by the resource management agent itself, by the resource management agent and setting and adjusting the price of the resource in accordance with the deviation of the actual use ratio from the target use rate and requesting assignment of each resource from agents, which will use resources, based on the price.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a resource equalization system for equalizing resource usage rates in resource allocation in a multi-agent system in which a large number of autonomous computers called agents are connected by a network.

[0002]

2. Description of the Related Art In a multi-agent system in which a large number of autonomously acting computers called agents as described above are connected by a network, resources such as a network capacity and a data storage area of the network connecting the agents exist in a distributed manner. , It becomes necessary to distribute these resources among agents. In particular, in order to respond to a sudden demand for resources, there is a demand to make the ratio of resources actually used as equal as possible. As a conventional resource distribution system, the following has been generally proposed. (1) Method using a broker: A method in which all the information about resources in the system is collected in one special agent called a broker, and this broker centrally determines the distribution of resources. In this method, when deciding the distribution of resources, it suffices to make the usage rates of the resources as uniform as possible. (2) Method based on contract net protocol: One agent for one resource serves as a manager and broadcasts the existence of that resource to other agents. Upon receiving the resource announcement, the agent sends a bid message to the manager when the agent determines that the resource is necessary. Upon receiving the bid message, the manager selects the one with the best conditions from among them and distributes the resources to the agent. A method in which resources are distributed by repeating this process.

[0003]

DISCLOSURE OF THE INVENTION Problem to be Solved by the Invention
The (broker-based method) has a drawback in that when a broker fails, the entire resource distribution stops. Further, it is necessary to collect all the information in the broker, and when the number of agents is large, there is a problem that the burden on the broker increases excessively and the broker becomes an overall bottleneck. Also, in the latter of the above-mentioned conventional techniques (method based on contract net protocol), it is not necessary to concentrate the information of the entire system on one computer, but the resource is distributed as soon as possible, so the entire system is There is a problem that it is difficult to equalize the use of resources of The present invention has been made in view of the above circumstances, and an object of the present invention is to solve the above-described problems in the conventional technique and to make the usage rate of each resource as uniform as possible without concentrating the information of the entire system. It is to provide an even distribution method of resources that can be realized.

[0004]

The above objects of the present invention are as follows:
In a resource allocation method in a multi-agent system in which a large number of autonomously operating computers called agents are connected by a network, an agent that manages the resource is defined for each of a plurality of types of resources, and the resource management agent manages itself. Set the target of the amount (usage rate) actually allocated to the total amount of resources to be used, set and adjust the price of the resource according to the difference between the target usage rate and the actual usage rate, and use the resource This is achieved by the resource equalization distribution method, characterized in that the agent makes an allocation request for each resource based on the price.

[0005]

In the uniform resource distribution system according to the present invention, as shown in FIG. 1, in addition to general agents that use resources (hereinafter referred to as "user agents"), resources are managed in a multi-agent system. Agent for
(Hereinafter, referred to as "resource management agent") is provided, and this resource management agent sets a target of the resource actually managed by the resource management agent, that is, a resource usage target, By autonomously setting and adjusting the price of resources according to the difference between the target usage rate and the actual usage rate, the resource usage rates of the user agents are equalized throughout the system. Needless to say, the above-mentioned “price” is used as a parameter for simulating the market economy with the object of the present invention of equalizing the resource usage rates in the entire system. It has the same meaning as that used in other research reports.

[0006]

Embodiments of the present invention will now be described in detail with reference to the drawings. As described above, FIG. 1 is a diagram showing an outline of a multi-agent system according to an embodiment of the present invention. The aforementioned user agent and resource management agent are connected by a network. FIG. 2 is a configuration diagram of a computer that constitutes the above resource management agent. The resource management agent is, in addition to the CPU 11, the main memory 12 and the I / O interface 13 which are the components of a normal computer, information about the resource to be managed, that is, the total capacity and the user agent that may use the resource. , A target resource usage rate, a current price, a price history, a current allocation request, and a resource management memory 14 for recording the allocation request history. These are connected by the system bus,
It is connected to the network via the I / O interface 13.

FIG. 3 is a block diagram of a computer constituting the above-mentioned user agent. The resource management agent
In addition to the CPU 21, the main memory 22, and the I / O interface 23, which are the components of a normal computer, information on resource use, that is, the total amount of resources to be secured, and which resources can be used in securing resources List, a resource price included in the options, a unit price for each option, an allocation amount for each option, and a resource use information memory 24 for recording an allocation request for each resource. These are connected by a system bus, and are connected to the network via the I / O interface 23. In the resource management agent, regarding the resource of the resource to be managed, it is assumed that the total capacity and the list of user agents that may use the resource are known in advance, and those data are stored in the resource management memory 14 in advance. It has been written. Further, in the user agent, it is assumed that the total amount of resources to be secured and a list of options for securing resources are known in advance, and those data are written in the resource use information memory 24 in advance.

The operation of this embodiment constructed as described above is
In the following, first, description will be made based on the processing flow chart of the resource management agent shown in FIG. Step 31:

[Outer 1] Step 32: Notify the concerned agent of the above-mentioned price. Here, the related agent is a user agent that may request allocation of the resource, and can be known from the data previously written in the resource management memory 14.

Step 33: Wait for a resource allocation request from the user agent. Step 34:

[Outside 2] After this, the process returns to step 32. Repeat this cycle.

Next, FIG. 5 shows a processing flow in the user agent. Step 41: Check the price of the resource that may be used by the user and write it in the resource use information memory 24. Step 42: Calculate a unit price for each option based on the resource price checked in step 41, and write it in the resource use information memory 24. Step 43: The amount of resources to be secured as a whole is distributed to each option by the ratio of the reciprocal of the unit price of the option calculated in step 42, the allocation amount for each option is calculated, and the result is stored in the resource use information memory 24. Write. Step 44: Calculate an allocation request for each resource from the allocation amount for each option obtained in step 43, and write the value in the resource use information memory 24. Step 45: Notify the resource management agent managing each resource of the allocation request for each resource. After this,
Return to step 41. Repeat this cycle.

As a concrete example, the result of the simulation performed based on the method shown in the above example will be described below. In this simulation, there are 12 resource management agents R0 to R11. The total capacity of each resource is R0 = 10, R1 = 20, R2
= 30, R3 = 40, R4 = 50, R5 = 60, R6 =
10, R7 = 20, R8 = 30, R9 = 40, R10 =
50 and R11 = 60. There are six user agents U0 to U5. Each user agent needs to reserve a total of five resources. Further, each user agent may combine a certain three types of resources and secure the required amount of resources in total of the three types of resources. Specifically, U0 is (R0, R1, R2), U1 is (R3, R4,
R5), U2 are (R6, R7, R8), U3 is (R9, R10,
R11) and U4 are (R0, R2, R4), and U5 is (R6, R8,
R10) can be combined. For example, U0 is R0
By securing 1 for R1, 2 for R2, and 2 for R2, it means that a total of 5 resources are secured as U0. Here, since there are combinations of three types of resources, the user agent has three options. here,
The unit price for each option is nothing but the price of the resource for each option.

Under the above-mentioned settings, the target resource utilization rate is set to 1.0, 0.5, and 0.1, and the resource utilization rate of the resource when the simulation is performed by the above-mentioned method for each case The change in dispersion is shown in FIG. Here, the number of repetitions on the horizontal axis indicates the number of price revisions. As is clear from FIG. 6, according to the present embodiment, although the individual agents do not collect the entire information, but determine the behavior simply based on the usage rate and the price of the individual resources, As a result, it can be seen that the variance of the resource usage rate is small, that is, equalization can be realized. In particular, it can be seen that the larger the utilization target, the faster the convergence. It is needless to say that the above embodiment shows one example of the present invention, and the present invention should not be limited to this. For example, the number of repetitions may be appropriately limited depending on the application.

[0013]

As described above in detail, according to the present invention, it is possible to realize a resource equalization distribution system capable of equalizing the usage rates of respective resources as much as possible without concentrating the information of the entire system. It has a remarkable effect.

[0014]

[Brief description of drawings]

FIG. 1 is a configuration diagram showing an outline of a multi-stage system according to an embodiment of the present invention.

FIG. 2 is a configuration diagram of a computer that constitutes a resource management agent according to the embodiment.

FIG. 3 is a configuration diagram of a computer that constitutes a user agent according to the embodiment.

FIG. 4 is a process flow diagram of a resource management agent according to the embodiment.

FIG. 5 is a process flow diagram of the user agent of the embodiment.

FIG. 6 is a graph showing changes in the variance of the resource usage rate in the simulation when the resource usage target is set to 1.0, 0.5, and 0.1.

[Explanation of symbols]

11: CPU, 12: main memory, 13: I / O interface, 14: resource management memory, 21: CPU, 2
2: Main memory, 23: I / O interface, 2
4: Resource usage information memory.

Claims (1)

[Claims] 1. In a resource allocation method in a multi-agent system in which a large number of autonomously acting computers called agents are connected by a network, an agent for managing the resources is defined for each of a plurality of types of resources, and the resource management agent is defined. Is the amount that is actually allocated to the total amount of resources that it manages.
The target of (usage rate) is set, the price of the resource is set and adjusted according to the difference between the target usage rate and the actual usage rate, and the agent who uses the resource determines the allocation request for each resource based on the price. A uniform resource distribution method characterized by performing.