JP2007172531A - Load distribution device and method, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To effectively distribute a load of a server device performing processing on the basis of data from a plurality of client devices. <P>SOLUTION: This load distribution device 2 in one mode determines correspondence relation between the plurality of client devices C1-C5 and a plurality of server devices S1-S3. The load distribution device 2 has: an evaluation value calculation means 6 calculating a present load distribution evaluation value about the present correspondence relation between the client devices C1-C5 and the server devices S1-S3; a selection means 8 selecting the new correspondence relation between the client devices C1-C5 and the server devices S1-S3, different from the presence correspondence relation; an evaluation value recalculation means 9 calculating a new load distribution evaluation value related to the selected new correspondence relation; and a change means 10 changing the correspondence relation between the client devices C1-C5 and the server devices S1-S3 into the new correspondence relation selected by the selection means 8 when the new load distribution evaluation value expresses a load-distributed state better than the present load distribution evaluation value. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to a load distribution apparatus, method, and program for performing load distribution among a plurality of server apparatuses.

  In Patent Document 1 (Japanese Patent Application Laid-Open No. 10-240697), a new session is used for measuring a load of a plurality of processing devices by a relay device and allocating a transaction so that the load due to transaction processing is equalized to each processing device. A load balancing method based on rate is described.

  In Patent Document 1, a transaction request is transmitted to a processing device having the lowest session usage rate.

JP-A-10-240697

However, in the load distribution method based on the session usage rate of Patent Document 1, no consideration is given to reducing the load on a processing device with a large load.
Further, in the invention of Patent Document 1, it is difficult to perform optimal transaction allocation when the load on the processing apparatus changes dynamically.

  Further, in Patent Document 1, even when a plurality of terminals handle related data, the plurality of terminals are not always associated with the same processing device. For this reason, it is necessary to transmit / receive related data between a plurality of processing devices, and there is an overhead of communication between processing devices.

  The present invention has been made in view of the above circumstances, and provides a load distribution device, method, and program for effectively distributing the load of a server device that performs processing based on data from a plurality of client devices. For the purpose.

  The load distribution apparatus according to the first example of the present invention determines the correspondence between a plurality of client apparatuses and a plurality of server apparatuses. The load distribution apparatus according to the first example is an evaluation value calculation means for calculating a current load distribution evaluation value representing the degree of current load distribution with respect to the current correspondence relationship between a plurality of client apparatuses and a plurality of server apparatuses. And a selection means for selecting a new correspondence relationship between a plurality of client devices and a plurality of server devices different from the current correspondence relation, and a new indicating the degree of load distribution regarding the new correspondence relation selected by the selection means Evaluation value recalculation means for calculating the load distribution evaluation value of the client, and when the new load distribution evaluation value represents a state in which the load distribution evaluation value is distributed more than the current load distribution evaluation value, a plurality of client devices and a plurality of servers And changing means for changing the correspondence relationship with the apparatus to a new correspondence relationship selected by the selection means.

As a result, a client device associated with a server device with a large load can be associated with another client device, and load distribution can be performed.
A load distribution apparatus according to a second example of the present invention is the load distribution apparatus according to the first example, wherein the load distribution apparatus stores a load index set for each of a plurality of client devices and is set for each of a plurality of server apparatuses. A processing capability index storage device for storing the processed processing capability index, and the evaluation value calculation means calculates the load factors of the plurality of server devices (load indexes of the associated client devices) for the current correspondence relationship. The maximum load factor among the load factors of the plurality of server devices is set as the current load distribution evaluation value, and the evaluation value recalculation unit is configured to calculate the new correspondence relationship with the plurality of servers. The load factor of the device is calculated from (the sum of the load indexes of the associated client devices / processing capability index), and the maximum load factor among the load factors of the plurality of server devices is calculated as the new load component. If the new load distribution evaluation value is smaller than the current load distribution evaluation value, the new load distribution evaluation value is more load distributed than the current load distribution evaluation value. It is assumed that it is in a state.

Thereby, when the load of the client device associated with the server device is excessive with respect to the processing capability of the server device, the load of the server device can be reduced and assigned to another server device.
The load distribution apparatus according to the third example of the present invention is the load distribution apparatus according to the first example. In the load distribution apparatus according to the first example, the evaluation value calculation unit calculates the maximum value of the resource usage of the plurality of server apparatuses for the current correspondence relationship. The current load distribution evaluation value is set, and the evaluation value recalculation means sets the maximum value of the resource usage of the plurality of server devices as the new load distribution evaluation value for the new correspondence, and the change means sets the new load distribution evaluation value. When the load distribution evaluation value is smaller than the current load distribution evaluation value, it is determined that the new load distribution evaluation value is determined to be in a state where the load distribution evaluation value is distributed more than the current load distribution evaluation value.

Thereby, when the resource usage of a server apparatus is excessive, the load of this server apparatus can be allocated to another server apparatus.
The load distribution apparatus according to a fourth example of the present invention is the load distribution apparatus according to the second or third load distribution apparatus, wherein the selection unit selects a server apparatus having a high load distribution evaluation value from a plurality of servers in the current correspondence relationship. The client device associated with the selected server device with a high load distribution evaluation value is selected, and the selected client device is associated with another server device that is not a server device with a high load distribution evaluation value. It is said.

Accordingly, the client device associated with the server device with a high load can be associated with the other server device, and the correspondence relationship between the plurality of server devices and the plurality of client devices can be changed.
The load distribution apparatus according to a fifth example of the present invention is the load distribution apparatus according to the first to fourth examples, wherein the plurality of client apparatuses belong to at least one client group that handles mutually related data and are selected. The means associates all client devices belonging to the same client group as the client device to be changed in association with the same server device.

Thereby, it is possible to distribute the load of the plurality of server devices while associating the client devices belonging to the same client group with the same server device, and to suppress overhead and data communication between the plurality of server devices.
A load balancer according to a sixth example of the present invention is the load balancer according to any of the first to fourth examples, wherein a plurality of client devices belong to at least one client group that handles mutually related data and are selected. The means selects an association change target client group associated with the association change target server device, and sets a plurality of other server devices that are different from the association change target server device. The number of associations of client devices belonging to the client group is obtained, and the association change target client that is associated with the server device of the association change target is associated with the server device with a large number of associations among a plurality of other server devices. Assume that client devices belonging to a group are associated.

  As a result, when the association of the client device is changed from the server device that is the association change target to another server device, the client device that is the association change target is changed to a client device that belongs to the same client group. Can be associated with the server device, and overhead and data communication between the plurality of server devices can be suppressed.

  The first to sixth examples are expressed as devices. However, the present invention is not limited to this, and the first to sixth examples may be expressed by a method, a program, a computer-readable storage medium, and the like.

  According to the present invention, it is possible to effectively distribute the load of a server device that performs processing based on data from a plurality of client devices.

Hereinafter, embodiments of the present invention will be described with reference to the drawings. In addition, in the following each figure, the part which implement | achieves the same function attaches | subjects the same code | symbol, and abbreviate | omits description.
(First embodiment)
FIG. 1 is a block diagram illustrating an example of a load distribution apparatus according to the present embodiment.

The network system 1 includes a plurality of sensors A1 to A5, a plurality of client devices C1 to C5, a plurality of server devices S1 to S3, and a load distribution device 2.
The sensors A1 to A5 are connected to the client devices C1 to C5, respectively. In the present embodiment, for example, the sensors A1 to A5 are various abnormality detection devices installed in the hallway. The sensors A1 to A5 and the client devices C1 to C5 belong to at least one of the client groups G1 to G4. In the present embodiment, two client devices C1 and C2 adjacent to each other belong to the client group G1, and two client devices C2 and C3 adjacent to each other belong to the client group G2. Two client devices C3 and C4 adjacent to each other belong to the client group G3, and two client devices C4 and C5 adjacent to each other belong to the client group G4.

  As the sensors A1 to A5, for example, a far-infrared human sensor, a fire detector, a seismometer, a camera, or the like can be used. The client devices C1 to C5 transmit the sensor data measured by the sensors A1 to A5, respectively, to the server device associated with itself among the server devices S1 to S3 via the load distribution device 2.

The server devices S1 to S3 issue an alarm when sensor data is received from both of two client devices belonging to the same group.
In this embodiment, the load balancer 2 is described as an independent device. However, the function of the load balancer 2 may be realized by other devices in the network system 1 according to its role. The network system 1 may be distributed among a plurality of devices. The load balancer 2 may be a network relay device such as a router or a gateway.

The load distribution device 2 includes a load index storage device 3, a processing capability index storage device 4, an initial association unit 5, an evaluation value calculation unit 6, storage devices 7a to 7e, a selection unit 8, an evaluation value recalculation unit 9, and a change unit. 10 and an updating unit 11 are provided.
The load index storage device 3 stores, for example, load index data D1 as shown in FIG. In the load index data D1, the load index set for each of the client devices C1 to C5 is managed. In this embodiment, a load index proportional to the size of sensor data of the sensors A1 to A5 connected to the client devices C1 to C5 is defined.

  The processing capacity index storage device 4 stores, for example, processing capacity index data D2 as shown in FIG. In the processing capacity index data D2, the processing capacity index set for each of the server apparatuses S1 to S3 is managed. In the present embodiment, a processing capability index proportional to the bandwidth of the network connected to each of the server apparatuses S1 to S3 is defined.

  As an initial state, the initial association unit 5 associates each client group G1 to G4 at random with the server devices S1 to S3, and stores the current correspondence relationship data D3 representing the current correspondence relationship in the storage device 7a. The initial association unit 5 associates client devices belonging to the same client group with the same server device.

Here, as shown in FIG. 4, initially, client groups G1 and G4 are associated with server apparatus S1, client group G2 is associated with server apparatus S2, and client group G3 is associated with server apparatus S3. Is associated with.
The evaluation value calculation unit 6 includes the current state correspondence data D3 stored in the storage device 7a, the load index data D1 stored in the load index storage device 3, and the processing capability index stored in the processing capability index storage device 4. Based on the data D2 and the evaluation function, the current load distribution evaluation value representing the current load distribution degree is calculated, and the current evaluation data D4 which is the calculation result is stored in the storage device 7b.

  In the present embodiment, for example, for each of the server devices S1 to S3, the evaluation value calculating unit 6 is a ratio of the total load index of the associated client device and the processing capability index of the server device (for example, corresponding The load index of the attached client device / the processing capability index of the server device) is calculated as the load factor, and the maximum value among the load factors of the server devices S1 to S3 is the current load distribution evaluation value, Current state evaluation data D4 including the load factor of each server device S1 to S3 and the current load distribution evaluation value is stored in the storage device 7b.

In the current correspondence relationship shown in FIG. 4 above, the load factor “2” of the server apparatus S1 is the maximum, so the current load distribution evaluation value is 2.
The selection unit 8 selects the server device S1 having a large load factor based on the current state evaluation data D4 of the storage device 7b. For example, the selection unit 8 may select the server device having the maximum load factor as the server device to be associated with, or select one of the server devices having a load factor equal to or greater than a predetermined value as the association modification target. You may select as a server apparatus.

Next, the selection unit 8 randomly selects one client group from the client groups G1 and G4 associated with the selected server device S1. Here, it is assumed that the client group G1 is selected.
Next, the selection unit 8 obtains client devices C1 and C2 belonging to the selected client group G1.

  Next, the selection unit 8 associates the client devices C1 and C2 belonging to the selected client group G1 with respect to the other server devices S2 and S3 different from the selected server device S1 having a large load factor. Find a number. Here, one client device belonging to the client group G1 is associated with the server device S2, and no client device belonging to the client group G1 is associated with the server device S3.

  Next, the selection unit 8 is a server device having a large number of associations of client devices belonging to the selected client group G1 among the other server devices S2 and S3 different from the selected server device S1 having a large load factor. The server device S2 is selected, the client devices C1 and C2 belonging to the selected client group G1 are associated with this server device S2, and new correspondence data D5 indicating the result is stored in the storage device 7c.

  As shown in FIG. 5, the selection unit 8 associates the client groups G1 and G2 with the server device S2, the client group G3 with the server device S3, and the client group G4 with the server device S1. A new corresponding relationship is obtained.

  The evaluation value recalculation unit 9 includes new correspondence relationship data D5 stored in the storage device 7c, load index data D1 stored in the load index storage device 3, and processing capability stored in the processing capability index storage device 4. Based on the index data D2 and the evaluation function, a new load distribution evaluation value indicating the degree of the new load distribution is calculated, and the new evaluation data D6 as the calculation result is stored in the storage device 7d.

  In the present embodiment, the evaluation value recalculation unit 9 obtains a new load distribution evaluation value using the same evaluation function as that of the evaluation value calculation unit 6. In other words, the evaluation value recalculation unit 9 is associated with the ratio of the total load index of the newly associated client device to the processing capability index of the server device (for example, for each server device S1 to S3). The total load index of the client device / the processing capability index of the server device) is calculated as a new load factor, and the maximum value among the load factors of the server devices S1 to S3 is set as a new load distribution evaluation value. New evaluation data D6 indicating new load factors and new load distribution evaluation values of the devices S1 to S3 is stored in the storage device 7d.

In the new correspondence relationship shown in FIG. 5, the load factor “1.5” of the server device S3 is the maximum, so the current load distribution evaluation value is 1.5.
Based on the current evaluation data D4 stored in the storage device 7b and the new evaluation data D6 stored in the storage device 7d, the changing unit 10 uses the new load distribution evaluation value for the current load distribution evaluation. It is determined whether or not a load distribution state is represented rather than a value, and when the new load distribution evaluation value represents a load distribution state rather than the current load distribution evaluation value, it is stored in the storage device 7c. According to the new correspondence relationship data D5, the correspondence relationship between the plurality of client devices C1 to C5 and the plurality of server devices S1 to S3 is changed to the new correspondence relationship selected by the selection unit 8.

In the present embodiment, the smaller the load distribution evaluation value is, the more the load distribution is determined. In the above example, since the current load distribution evaluation value 2> the new load distribution evaluation value 1.5, It is determined that the corresponding relationship can distribute the load.
When the correspondence relationship between the plurality of client devices C1 to C5 and the plurality of server devices S1 to S3 is changed, the changing unit 10 transitions to the hold mode and changes the association (here) Then, the sensor data from the client devices C1, C2) belonging to the client group 1 is stored in the storage device 7e without being transferred to the server device.

  The change unit 10 instructs the change source server apparatus S1 to change the client group G1 to the change destination server apparatus S2. The change source server apparatus S1 transmits data necessary for processing the sensor data of the client group G1 (IP addresses of the sensors A1 and A2, data accumulated so far, etc.) to the change destination server apparatus S2.

Thereafter, the change unit 10 transmits the sensor data stored in the storage device 7e to the server server S2 that is the change destination, and cancels the hold mode. As a result, the load distribution evaluation value decreases from 2 to 1.5, and load leveling is realized.
When the changing unit 10 changes the correspondence between the plurality of client devices C1 to C5 and the plurality of server devices S1 to S3, the updating unit 11 updates the new correspondence relationship data D5 stored in the storage device 7c. The content of the current correspondence data D3 stored in the storage device 7a is updated according to the content.

FIG. 6 is a flowchart showing an example of the operation of the load distribution apparatus 2 according to the present embodiment.
In step F1, the load index storage device 3 stores the load index set for each of the client devices C1 to C5, and the processing capability index storage device 4 sets the processing capability set for each of the server devices S1 to S3. Remember the exponent.

  In step F2, the initial association unit 5 initially associates each of the client groups G1 to G4 (that is, client devices C1 to C5) with the server devices S1 to S3 at random, and indicates the current correspondence relationship. The related data D3 is stored in the storage device 7a.

In step F3, the evaluation value calculation unit 6 calculates a current load distribution evaluation value indicating the degree of the current load distribution, and stores the current evaluation data D4 that is the calculation result in the storage device 7b.
In step F4, the selection unit 8 selects a new correspondence relationship between the client groups G1 to G4 (that is, the client devices C1 to C5) and the server devices S1 to S3 in order to simulate the change in the correspondence relationship. New correspondence data D5 indicating the result is stored in the storage device 7c.

In step F5, the evaluation value recalculation unit 9 calculates a new load distribution evaluation value indicating the degree of the new load distribution for the new correspondence, and stores the new evaluation data D6 as the calculation result in the storage device 7d. Remember.
In step F6, the changing unit 10 compares the current load distribution evaluation value with the new load distribution evaluation value, and the new load distribution evaluation value is more load distributed than the current load distribution evaluation value. Whether or not is represented.

If the new load distribution evaluation value is not in a state where the load distribution evaluation value is distributed more than the current load distribution evaluation value, the process proceeds to step F9.
When the new load distribution evaluation value represents a state in which the load distribution evaluation value is distributed more than the current load distribution evaluation value, in step F7, the changing unit 10 determines that the client groups G1 to G4 (that is, client devices C1 to C5) and The correspondence relationship between the plurality of server devices S1 to S3 is changed to a new correspondence relationship.

In step F8, the update unit 11 updates the current correspondence relationship with a new correspondence relationship.
In step F9, it is determined whether or not to repeat the above process. If so, the process returns to step F3. If not, the process ends.
FIG. 7 is a flowchart showing an example of the operation of the selection unit 8 of the load distribution apparatus 2 according to the present embodiment.

In step I1, the selection unit 8 selects the server device S1 having a high load factor among the server devices S1 to S3 as the server device to be changed in association.
In step I2, the selection unit 8 randomly selects one client group from the client groups G1 and G4 associated with the selected server device S1.

In step I3, the selection unit 8 obtains client devices C1 and C2 belonging to the selected client group G1.
In step I4, the selection unit 8 obtains the number of associations between the client devices C1 and C2 belonging to the selected client group G1 with respect to the other server devices S2 and S3 that are different from the server device S1 whose association is changed.

In step I5, the selection unit 8 selects a server device S2 having a large number of associated client devices belonging to the client group G1 among the other server devices S2 and S3 different from the selected server device S1 having a high load factor. select.
In step I6, the selection unit 8 associates the selected server device S2 with the client devices C1 and C2 belonging to the selected client group G1, and stores new correspondence data D5 indicating the result in the storage device 7c. Remember.

  In the present embodiment described above, with respect to the correspondence relationship between the client devices C1 to C5 and the server devices S1 to S3, the evaluation function is used when the current load distribution evaluation value is calculated by the evaluation function and the correspondence relationship is changed. When a new load distribution evaluation value is calculated, and the new load distribution evaluation value indicates that the load is distributed more than the current load distribution evaluation value, the new correspondence after change is adopted. The

In the present embodiment, the correspondence relationship between the client devices C1 to C5 and the server devices S1 to S3 is changed so that the maximum load factor of the plurality of server devices S1 to S3 is suppressed. For this reason, the load factors of the server apparatuses S1 to S3 can be leveled.
In the present embodiment, the server device with the maximum load factor is selected as the server device for which the correspondence relationship is to be changed, and the client device associated with the selected server device is associated with another server device. It is done.

As described above, by changing the correspondence relationship of the client device associated with the server device having the maximum load factor, it is possible to quickly reach the optimum state in which the load factor is leveled.
Further, in this embodiment, assuming that a set of client devices that handle related data belong to the same client group, load distribution is performed so that client devices belonging to the same client group are associated with the same server device.

  Thereby, communication between server apparatuses can be omitted and processing can be executed at high speed. That is, for example, as shown in FIG. 8, when client devices belonging to the same client group are associated with different server devices, processing such as an inquiry of data received from the client device is required between the server devices. Become. However, in the present embodiment, as shown in FIG. 9, since client devices belonging to the same client group are associated with the same server device, processing such as an inquiry of data received from the client device between server devices. Is unnecessary, the amount of processing can be reduced, and the efficiency of processing can be improved.

In the present embodiment, the evaluation value calculation unit 6 and the calculation value recalculation unit 9 perform the same calculation, and may be integrated to be a common unit.
(Second Embodiment)
In the present embodiment, a modification of the first embodiment will be described.
In the present embodiment, instead of the load factors of the server devices S1 to S3 in the first embodiment, the resource usage amounts (or resource usage rates) of the server devices S1 to S3 may be used. change.

That is, in the present embodiment, the load distribution apparatus 2 measures the current resource usage in the plurality of server apparatuses S1 to S3 for the current correspondence.
Next, the load distribution apparatus 2 sets the maximum value among the current resource usage amounts in the plurality of server apparatuses S1 to S3 as the current load distribution evaluation value.

Next, the load distribution apparatus 2 measures the new current resource usage (or resource usage rate) in the plurality of server apparatuses S1 to S3 for the new correspondence.
Next, the load distribution apparatus 2 sets the maximum value among the new resource usage amounts in the plurality of server apparatuses S1 to S3 as a new load distribution evaluation value.

Then, when the new load distribution evaluation value is smaller than the current load distribution evaluation value, the new load distribution evaluation value is more load distributed than the current load distribution evaluation value. Judge that it is.
As the resource usage, for example, CPU usage rate, traffic, and the like can be used.

In the present embodiment described above, the maximum resource usage for the server apparatuses S1 to S3 can be reduced, and the resource usage can be leveled between the server apparatuses S1 to S3.
In the present embodiment, even when the loads on the server apparatuses S1 to S3 change dynamically (when the load applied to the server apparatus by the client apparatus changes with time), the loads on the server apparatuses S1 to S3 Can be leveled.

  In the above embodiments, various modifications may be made as long as the same function can be realized. Further, for example, the load distribution device 2 according to each of the above embodiments may execute the processing for load distribution when a load distribution start command is input, and the load distribution is performed every predetermined time. The processing for the above may be executed.

The number of repetitions of processing for load distribution in the load distribution apparatus 2 according to each of the above embodiments may be freely set.
The load distribution in the network system 1 according to each of the above embodiments may be realized by a processor controlled by a program, for example.

  The program used for load distribution can be applied to a computer having a processor by writing it in a storage medium such as a magnetic disk (flexible disk, hard disk, etc.), optical disk (CD-ROM, DVD, etc.), semiconductor memory, etc. is there. The program can also be applied to a computer by being transmitted through a communication medium. The computer reads the program and controls the operation by the program, thereby realizing the function as the load distribution device 2. The program may be distributed and arranged in a plurality of computers, and the processing may be executed while the plurality of computers cooperate with each other.

  The present invention is effective in the control field in which loads are distributed among a plurality of servers.

The block diagram which shows an example of the load distribution apparatus which concerns on the 1st Embodiment of this invention. The figure which shows an example of load index data. The figure which shows an example of processing capacity index data. The block diagram which shows an example of the correspondence of an initial client group and a server apparatus. The block diagram which shows an example of the correspondence of a new client group and a server apparatus. The flowchart which shows an example of operation | movement of the load distribution apparatus which concerns on the embodiment. The flowchart which shows an example of operation | movement of the selection part of the load distribution apparatus which concerns on the embodiment. The block diagram which shows an example of the correspondence which requires the inquiry of data between servers. The block diagram which shows an example of the correspondence at the time of matching the client apparatus which belongs to the same client group with the same server.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 ... Network system, A1-A5 ... Sensor, C1-C12 ... Client apparatus, S1-S5 ... Server apparatus, G1-G6 ... Client group, 2 ... Load distribution apparatus, 3 ... Load index storage apparatus, 4 ... Processing capacity index Storage device, D2 ... processing capability index data, 5 ... initial association unit, 6 ... evaluation value calculation unit, 7a to 7e ... storage device, 8 ... selection unit, 9 ... evaluation value recalculation unit, 10 ... change unit, 11 ... update unit, D1 ... load index data, D2 ... processing capability index data, D3 ... current state correspondence data, D4 ... current state evaluation data, D5 ... new correspondence data, D6 ... new evaluation data

Claims (10)

In a load balancer that determines the correspondence between a plurality of client devices and a plurality of server devices,
Evaluation value calculation means for calculating a current load distribution evaluation value representing a current load distribution level for the current correspondence relationship between the plurality of client devices and the plurality of server devices;
Selecting means for selecting a new correspondence between the plurality of client devices and the plurality of server devices different from the current correspondence;
Evaluation value recalculation means for calculating a new load distribution evaluation value representing the degree of load distribution related to the new correspondence selected by the selection means;
When the new load distribution evaluation value represents a state in which load distribution is performed more than the current load distribution evaluation value, the correspondence relationship between the plurality of client devices and the plurality of server devices is selected. A load distribution apparatus comprising: a change unit configured to change the new correspondence selected by the unit. The load distribution apparatus according to claim 1, wherein
A load index storage device for storing a load index set for each of the plurality of client devices;
A processing capability index storage device for storing a processing capability index set for each of the plurality of server devices;
The evaluation value calculation means calculates the load factor of the plurality of server devices from (the total load index of the associated client devices / processing capability index) for the current correspondence relationship, and the plurality of server devices The maximum load factor among the load factors is the current load distribution evaluation value,
The evaluation value recalculation means calculates the load factor of the plurality of server devices with respect to the new correspondence relationship based on (total load index of the associated client device / processing capability index), and the plurality of servers. The maximum load factor among the load factors of the device is the new load distribution evaluation value,
The changing means is configured such that when the new load distribution evaluation value is smaller than the current load distribution evaluation value, the new load distribution evaluation value is more load distributed than the current load distribution evaluation value. It is judged that it is in the state where it is. The load distribution apparatus according to claim 1, wherein
The evaluation value calculation means sets, as the current load distribution evaluation value, the maximum value of the resource usage of the plurality of server devices for the current correspondence relationship,
The evaluation value recalculation means sets, as the new load distribution evaluation value, the maximum value of the resource usage of the plurality of server devices for the new correspondence relationship,
The changing means is configured such that when the new load distribution evaluation value is smaller than the current load distribution evaluation value, the new load distribution evaluation value is more load distributed than the current load distribution evaluation value. It is judged that it is in the state where it is. In the load distribution apparatus according to claim 2 or 3,
The selection means includes
In the current correspondence relationship, a server device having a high load distribution evaluation value is selected from the plurality of servers,
Select the client device associated with the selected server device with a high load balancing evaluation value,
The load distribution apparatus, wherein the selected client apparatus is associated with another server apparatus that is not the server apparatus with the selected load distribution evaluation value. The load distribution apparatus according to any one of claims 1 to 4,
The plurality of client devices belong to at least one client group that handles mutually related data,
The load distribution apparatus, wherein the selection unit associates all client devices belonging to the same client group as the client device to be changed in association with the same server device. The load distribution apparatus according to any one of claims 1 to 4,
The plurality of client devices belong to at least one client group that handles mutually related data,
The selection means includes
Select the client group of the association change target that is associated with the server device of the association change target,
For a plurality of other server devices different from the association change target server device, obtain the number of associations of client devices belonging to the association change target client group,
Of the plurality of other server devices, a server device having a large number of associations is associated with a client device that is associated with the association change target server device and belongs to the association change target client group. A characteristic load balancer. In a load balancing method for determining correspondences between a plurality of client devices and a plurality of server devices,
For the current correspondence relationship between the plurality of client devices and the plurality of server devices, a current load distribution evaluation value representing the current load distribution degree is calculated,
Selecting a new correspondence between the plurality of client devices and the plurality of server devices different from the current correspondence;
Calculating a new load distribution evaluation value representing the degree of load distribution related to the new correspondence;
When the new load distribution evaluation value represents a state in which load distribution is performed more than the current load distribution evaluation value, the correspondence relationship between the plurality of client devices and the plurality of server devices is selected. The load distribution method is characterized by changing to the new correspondence selected by the means. The load balancing method according to claim 7, wherein
Storing a load index set for each of the plurality of client devices, and storing a processing power index set for each of the plurality of server devices,
For the current correspondence relationship, the load factor of the plurality of server devices is calculated from (the sum of the load indexes of the associated client devices / processing capability index), and the maximum load factor of the plurality of server devices is calculated. The load factor is the current load distribution evaluation value,
For the new correspondence, the load factor of the plurality of server devices is calculated from (the total load index of the associated client devices / processing capability index), and the maximum load factor of the plurality of server devices is calculated. The load factor is the new load distribution evaluation value,
When the new load distribution evaluation value is smaller than the current load distribution evaluation value, it is determined that the new load distribution evaluation value is more load distributed than the current load distribution evaluation value. And a load balancing method. The load balancing method according to claim 7, wherein
For the current correspondence, the maximum value of the resource usage of the plurality of server devices is the current load distribution evaluation value,
For the new correspondence, the maximum value of the resource usage of the plurality of server devices is the new load distribution evaluation value,
When the new load distribution evaluation value is smaller than the current load distribution evaluation value, it is determined that the new load distribution evaluation value is more load distributed than the current load distribution evaluation value. And a load balancing method. In a program for realizing load balancing processing for determining the correspondence between a plurality of client devices and a plurality of server devices,
Computer
With respect to the current correspondence relationship between the plurality of client devices and the plurality of server devices, a current load distribution evaluation value representing the degree of current load distribution is calculated, and the current evaluation data as the calculation result is first Evaluation value calculation means stored in the storage device;
Selecting means for selecting a new correspondence relationship between the plurality of client devices and the plurality of server devices different from the current correspondence relationship, and storing new correspondence relationship data as a selection result in a second storage device;
A new load distribution evaluation value indicating the degree of load distribution related to the new correspondence relationship represented by the correspondence relationship data stored in the second storage device is calculated, and the new evaluation data as the calculation result is stored in the third storage. Evaluation value recalculation means stored in the apparatus;
The new load distribution evaluation value represented by the new evaluation data stored in the third storage device is the current load distribution evaluation value represented by the current evaluation data stored in the first storage device. When the load distribution state is expressed, the function as a changing unit that changes the correspondence relationship between the plurality of client devices and the plurality of server devices to the new correspondence relationship selected by the selection unit. Program to let you.

Images