JP2003058520A - Computer arrangement method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a computer arrangement method which can realize the most suitable arrangement of processing processes corresponding to processing capabilities of respective computers by quantitatively setting the arrangement of processing processes in many computers connected to a network in consideration of the variance of process load and the weight of communication load. SOLUTION: With respect to a method for arranging many computers connected to the network, a processing efficiency coefficient E (E=(L/P)max +α.Σ(T/ P)) is calculated and stored where L, P, T, (L/P)max , Σ(T/P), and α are the processing process load of a computer, the processing capability, the communication load required for data transmission and reception between computers, a maximum value of a processing process load capability ratio (L/P), the sum total of communication load capability ratios (T/P) of respective computers, and a weight coefficient respectively, and the computers are arranged over the network so as to minimize the processing efficiency coefficient E.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a plurality of computers connected to a network, the relation between the processing capacity of the computer and the processing load of each computer from the network, and the transmission / reception of data between the computers. The present invention relates to a method of arranging computers, which is arranged according to the required communication load.

[0002]

2. Description of the Related Art FIG. 7 shows an example of a network in which a plurality of computers are connected. In the figure, reference numeral 3 denotes a plurality of computers 1 (CPU # 1, CPU # 2, CPU # 3, ...
PU # n) is connected. Reference numeral 2 is a processing process executed by each of the computers 1. For example, CPU # 1 of the computer 1 is processing processes A and D,
The CPU # 2 is arranged like the processing processes C and E, the CPU # 3 is arranged like the processing process B, ....

Each computer 1 (CPU # 1, CP
U # 2, CPU # 3 ... CPU # n)
Arrangement of the memory card 2 is performed by each computer 1 (CPU # 1, CPU
# 2, CPU # 3 ... CPU # n) processing capacity and each processing
In Process 2 (A, B, C, D, E ... N)
Process load L₁, L_Two, L_Three, L_Four, L₅... L _n
Set according to the operator's knowledge and experience in consideration of the size of
is doing.

[0004]

However, in the conventional computer arranging method as shown in FIG. 7, the processing capacity of each computer 1 and the process load L of each processing process 2 are associated with each other. Since the processing process 2 is arranged in each computer 1 according to the knowledge and experience of the operator, the computer 1 in which the processing process 2 is arranged when the process load L changes.
The processing performance of the computer is not optimal, and the processing performance of the computer system connected to the network 3 is reduced.

Further, since the arrangement of the processing processes 2 in each computer 1 is set based on the knowledge and experience of the operator, the quantitative arrangement of the processing processes 2 is not performed, and from this aspect, the processing of the entire computer system is performed. Inevitable performance degradation. Further, in such a conventional technique, since the processing process 2 is arranged in each computer 1 without considering the communication load required for transmitting and receiving data between the computers, the processing process 2 is optimal. No placement is done. There are problems such as.

In view of the above-mentioned problems of the prior art, the present invention makes it possible to set the arrangement of processing processes to a plurality of computers connected to a network quantitatively and by incorporating the fluctuation of the process load and the size of the communication load, An object of the present invention is to provide a computer arranging method capable of realizing the optimum arrangement of processing processes corresponding to the processing capability of each computer.

[0007]

In order to solve the above problems, the present invention provides, as an invention according to claim 1, a plurality of computers connected to a network, the processing capacity of the computer and each computer from the network. In the method of arranging computers according to the relationship with the processing process load, the processing process load capacity ratio (L), which is the ratio of the processing process load (L) of the computer and the processing capacity (P) of the computer, for each computer. / P) and select the maximum value, and the communication load (T) required to send and receive data between the computers.
The communication load capacity ratio (T / P), which is the ratio of the processing capacity of the computer to the processing capacity of the computer, is calculated for each computer, and the sum thereof is calculated to obtain the maximum value of the processing process load capacity ratio (L / P). A processing efficiency coefficient (E = (L / P) _max + α · Σ (T / T / S) that is the sum of the communication load capacity ratio (T / P)
P) (where α is the weighting factor of the communication load with respect to the computer load)) is arranged in the network so that the plurality of computers are arranged in the network.

The invention according to any one of claims 2 to 6 relates to a method for arranging computers by a genetic algorithm method, and the invention according to claim 2 relates to a plurality of computers connected to a network, from the processing capacity of the computers and the network. In the method of arranging computers according to the relationship with the processing process load on each computer, an arbitrary initial placement plan of the processing processes for the plurality of computers is created, and the initial placement plan and the initial placement plan are used as a base. Regarding the rearrangement arrangement plan that has been rearranged, the processing efficiency coefficient E (E) is calculated according to the processing process load (L) of the computer, the processing capacity (P) of the computer, and the communication load (T) required to send and receive data between the computers. = (L
/ P) _max + α · Σ (T / P): Here (L / P)
_max is the maximum value of the processing process load capacity ratio (L / P),
Σ (T / P) is the sum of the communication load capacity ratios (T / P) of each computer, and α is the weighting factor of the communication load with respect to the computer load), and the accumulated values are calculated. Generations are updated and repeated until E becomes smaller than a predetermined value and a predetermined end condition is satisfied, and when the end condition is satisfied, the allocation plan with the minimum processing efficiency coefficient E is selected. It is characterized in that it is output as an optimum arrangement.

According to a second aspect of the present invention, it is preferable that the ending condition is satisfied when the processing efficiency coefficient E becomes smaller than a certain threshold value. Further, in the second aspect, it is preferable that the method of setting the rearrangement placement plan is configured as in the fourth, fifth and sixth aspects. That is, in claim 4, the rearrangement placement plan is set to any two of the plurality of sets of initial placement plans.
Computers that execute a predetermined processing process among a plurality of processing processes are exchanged between sets of placement plans. In the fifth aspect, the rearrangement placement plan is created by switching the processing processes between the computers in one placement plan at a certain frequency. In the sixth aspect, the rearrangement placement plan is created by moving the processing process executed by a predetermined computer in the placement plan of the plurality of computers to another computer.

The inventions of claims 7 to 8 relate to a method of arranging computers by an annealing method, and the invention of claim 7 relates to a plurality of computers connected to a network,
In a computer arranging method for arranging according to a relation between a processing capacity of the computer and a processing process load from the network to each computer, the computer is arranged with respect to a current allocation plan corresponding to the processing process loads of the plurality of computers. Processing process load (L), processing capacity of the computer (P), communication load (T) required to send and receive data between the computers
And the processing efficiency coefficient E (E = (L / P) _max + α ·
Σ (T / P): Here, (L / P) _max is the maximum value of the processing process load capacity ratio (L / P), and Σ (T / P) is the communication load capacity ratio (T / P) of each computer. Sum, α is the weighting coefficient of communication load to computer load),
The alternative placement plan is set by exchanging the processing processes at any position among the plurality of computers in the initial placement plan, and then the processing efficiency coefficient E1 corresponding to the processing efficiency coefficient E is calculated for the alternative placement plan, When the processing efficiency coefficient E1 in the alternative placement plan is smaller than the processing efficiency coefficient E in the current placement plan (that is, when E-E1> 0)
The alternative placement plan is output as an optimal placement.

According to an eighth aspect of the present invention, there is provided a computer arranging method for arranging a plurality of computers connected to a network according to a relation between a processing capacity of the computer and a processing process load from the network to each computer. Regarding the current arrangement plan corresponding to the processing process load of a plurality of computers, the processing process load (L) of the computer, the processing capacity (P) of the computer, and the communication load (T) required for transmitting and receiving data between the computers. ) And the processing efficiency coefficient E (E =
(L / P) _max + α · Σ (T / P): where (L / P
P) _max is the maximum value of the processing process load capacity ratio (L / P), Σ (T / P) is the sum of the communication load capacity ratio (T / P) of each computer, and α is the weighting coefficient of the communication load with respect to the computer load. ) Is calculated, the processing process is switched at an arbitrary position among a plurality of computers in the current placement plan to set an alternative placement plan, and then the processing efficiency corresponding to the processing efficiency coefficient E for the alternative placement plan. Coefficient E1
When the processing efficiency coefficient E1 in the alternative placement plan is larger than the processing efficiency coefficient E in the current placement plan (that is, E−E1 ≦ 0), the application condition of the alternative placement plan is set. The operation of determining whether or not the alternative placement plan is applied to the current placement plan using the threshold value (S) is repeated while decreasing the threshold value (S) to obtain the current placement of the alternative placement plan. It is characterized in that the alternative placement plan is output as the optimal placement when it is no longer applied to the plan.

According to this invention, the maximum value ((L / P) _max ) of the ratio of the processing process load to the processing capacity of the computer connected to the network and the ratio of the total communication load to the computer processing capacity (Σ ( T / P)), that is, the processing efficiency coefficient (E) is introduced, and the processing efficiency coefficient (E)
Since the processing process is allocated to each computer so that the processing efficiency is minimized, the allocation of the processing process to the computer connected to the network is performed so that the processing efficiency coefficient (E) is minimized. It can be automatically arranged to be uniform.

Further, when there is a change in the process load, the change in the process load is the maximum value ((L / P)) of the ratio between the processing process load and the processing capacity of the computer.
by being incorporated into _{ma x),} can be properly form the arrangement of the treatment process of automatically correspond to variations in the process load on each computer.

Further, since the communication loads of all the computers connected to the network are included in the processing efficiency coefficient (E), the processing processes can be optimally arranged in the computers in consideration of the communication loads. it can.
As described above, effective processing is performed as the entire computer system connected to the network, and the processing efficiency of the computer system is improved.

If the processing process to the computer is arranged by the genetic algorithm method as in the invention according to claims 2 to 6, the arrangement plan is replaced and the generation is updated until the arrangement plan satisfies a predetermined end condition. By repeating it, the initial placement plan is sequentially shifted to the preferred placement plan, and the optimum placement plan is converged. Therefore, the optimum placement can be obtained with high accuracy.

Further, when the processing process is arranged in the computer by the annealing method as in the invention according to claims 7 to 8, the threshold value (S), which is the application condition of the alternative placement plan, is reduced and the alternative placement plan is reduced. By adopting the placement plan when the application of No. is not performed as the optimum placement, even if there is a change in the process load on the way, the optimum placement including the load change can be obtained.

[0017]

BEST MODE FOR CARRYING OUT THE INVENTION The present invention will be described in detail below with reference to the embodiments shown in the drawings. However, the dimensions, materials, shapes, relative positions, etc. of the components described in this embodiment are not intended to limit the scope of the present invention thereto, unless there is a specific description, and are merely illustrative examples. Nothing more.

FIG. 1 is a control flow chart showing a first embodiment of the computer placement method according to the present invention, and FIG. 2 is a control flow chart showing a second embodiment of the computer placement method. FIG. 3 is a procedure (part 1) of arranging processing processes on computers in the first and second embodiments, and FIG. 4 is a procedure (part 2) of arranging processing processes on computers in the first and second embodiments. Shows the procedure (3) of arranging the processing processes on the computer in the first and second embodiments. FIG. 6 is a block diagram showing the configuration of the computer placement system according to the present invention.

Of the computer placement system according to the present invention
In FIG. 6 showing the configuration, 1 is connected to the network 3.
CPUs # 1 and C
PU # 2, CPU # 3 ... CPU # n is the computer number
No. 2 (A, B, C, D ... N) is for each of the above
Processing processes placed (loaded) on the computer,
L ₁, L_Two, L_Three, L_Four...... L_nIs each processing process
2 (A, B, C, D ... N) process load, T₁₂,
T_Thirteen, T₁₄, T₁₅...... T_1nIs the computer
Communication with a processing process running on a device other than (CPU)
Show each load.

An optimum placement device 5 places the processing process 2 set in the computer 1 by the placement method described later. 6 is each computer 1 (CPU
# 1, CPU # 2, CPU # 3 ... CPU # n) is a computer processing capacity setting table in which processing capacities P ₁ , P ₂ , P ₃ , P ₄ ... _Pn are set. 7 is a processing process and communication load setting table, and the processing process 2
Process load L ₁ in (A, B, C, D ... N),
Communication loads T ₁₂ , T ₁₃ , T ₁₃ between L ₂ , L ₃ , L ₄ ... L _n and the processing process 2 operating on a computer other than the computer (CPU) 1 on which the processing process 2 is processed.
₁₄ , T ₁₅ ... T _{1n and the} like are set. The setting data in the computer processing capacity setting table 6 and the setting data in the processing process and communication load setting table 7 are input to the optimum placement apparatus 5. Also,
The load processed by each computer 1 is fed back to the optimum placement device 5.

Next, based on FIGS. 1 to 5, a processing process 2 in each computer 1 in the optimum placement apparatus 5 will be described.
The arrangement method of will be described. FIG. 1 is a control flowchart of a computer arranging method by the genetic algorithm method according to the first embodiment of the present invention. In FIG. 1, first, in an initial state (t = 0), a processing process 2 is performed using an arbitrary method. Of the computer 1 (CPUs # 1, # 2, # 3, ... #n) that processes (A, B, C, D ... N),
Initial arrangement plan between processing process 2 and computer 1 (Fig. 3
A plurality of placement plans 1 and 2) are created and stored in the placement plan pool (step (1) in FIG. 1).

Next, each of the above arrangement plans is evaluated by the following method. The processing efficiency coefficient E is an index indicating the optimum arrangement for uniform processing of each computer 1, that is, the arrangement that minimizes the processing cost of each computer 1.
To introduce. The processing efficiency coefficient (E) is expressed by the following equation (1). E = (L / P) _max + α · Σ (T / P) (1) where P: computer processing capacity L: computer processing process load T: communication load α required to send and receive data between computers: Communication load weighting factor for computer load

As shown in the equation (1), the processing efficiency coefficient E is a processing process load capacity ratio which is a ratio of the processing process load L of the computer 1 to the processing capacity P of the computer 1 for each computer 1. (L / P)
The maximum value (L / P) _max of the computer and the sum of the computer of the communication load capacity ratio (T / P) which is the ratio of the communication load T obtained for each computer to the processing capacity P of the computer 1 and the weighting factor α. It is defined by the sum of the value .alpha..SIGMA. (T / P) multiplied by .multidot., And in this embodiment, as will be described later, the plurality of computers 1-process 2 are arranged so that the processing efficiency coefficient E is minimized. Set the arrangement between.

As described above, the processing efficiency coefficient E is calculated and evaluated for the plurality of initial placement plans, that is, the initial t generation placement plans (step (2)). In the evaluation of the processing efficiency coefficient E, it is determined whether or not there is an optimal allocation plan that minimizes the processing efficiency coefficient E in the initial allocation plan and whether to move to the next generation (t + 1). (Step (3)),
When the optimum placement plan is obtained, the process is ended, and the placement between the processing process 2 and the computer 1 is set in the optimum placement plan (step (4)). A plurality of processing efficiency coefficients E are selected (step (5)).

Any two placement plans are selected from the plurality of selected placement plans, and as shown in FIG. 3, any place in the placement plan is replaced by an arbitrary number (for example, the placement plan). 1 and placement plan 2 are replaced), and a plurality of new placement plans (for example, placement plan 1A and placement plan 2A) are created and stored in the placement plan pool (step (6)). As shown in FIG. 4, as a mutation of the placement plan, the new placement plan accumulated in the placement plan pool is replaced with a processing process at an arbitrary position at an arbitrary frequency in the new placement plan. That is,
A processing process 2 (for example, processing process B) executed by a predetermined computer 1 (for example, CPU # 1) is moved to another computer 1 (for example, CPU # 3) and executed by the other computer 1 (for example, CPU # 3). The processing process 2 (for example, the processing process E) to be performed is moved to the predetermined computer 1 (for example, the CPU # 1) (step (7)).

As a second case of the mutation of the placement plan, as shown in FIG. 5, a predetermined computer (eg, CPU # 1) is executed in the placement plan (eg, placement plan 1) of a plurality of computers 1. The processing process 2 (for example, the processing process B) that executes
The operation of moving to # 3) and creating a new placement plan (eg, placement plan 1A) is performed at an arbitrary frequency (step (7)).
Then, the operation of creating a plurality of new placement plans shown in FIG. 3 (step (6)) and the operation in two cases of mutations shown in FIGS. 4 and 5 (step (7)) are added to the placement plan pool. Repeat until N new (recombinant) placement plans are accumulated (step (8)).

From the operation (step (2)) of calculating and evaluating the processing efficiency coefficient E for the t-th generation placement plan, the operation of accumulating N new (recombined) placement plans in the placement plan pool (step (8) )) As one generation (t generation), the generations are sequentially updated (t = t + 1), and repeated until the search end condition is satisfied (step (9)), and when the end condition is satisfied, the data is accumulated. Among the placement plans, the placement plan having the smallest processing efficiency coefficient E is output as the optimum placement. As the ending condition, a case where the processing efficiency coefficient E becomes smaller than a certain threshold value, a case where the search time exceeds a predetermined allowable time, or the like is used.

According to the first embodiment (genetic algorithm method), the placement plan is replaced and the generation is updated with a predetermined termination condition, that is, the processing efficiency coefficient E becomes smaller than a certain threshold value. In this case, by repeating until the search time exceeds a predetermined permissible time, etc., the initial placement plan is sequentially shifted to a preferred placement plan and converged to the optimal placement plan, so that the optimal placement can be obtained with high accuracy. be able to.

Next, FIG. 2 is a control flow chart of the computer placement method by the annealing method according to the second embodiment of the present invention. In FIG. 2, first, as in the case of the first embodiment, the initial state is set. Processing process 2 (A, B, C, D ... N) using an arbitrary method at (t = 0)
Computer 1 (CPUs # 1, # 2, # 3) for processing
.. #n), a plurality of initial placement plans between the processing process 2 and the computer 1 (placement plans 1 and 2 in FIG. 3) are created, and the processing efficiency coefficient (E ) Is calculated and the initial placement plan is stored in the placement plan pool. (Step (1) in FIG. 2).

As shown in FIG. 4, the initial placement plan accumulated in the placement plan pool, that is, the current placement plan, is replaced with the processing process at any position in the current placement plan at an arbitrary frequency. That is, a predetermined computer 1 (for example, a CPU
Process process 2 (for example, process B) executed by # 1) is moved to another computer 1 (for example, CPU # 3), and process 2 (for example, process # 3) executed by the other computer 1 (for example, CPU # 3) is executed. The process E) is moved to the predetermined computer 1 (eg CPU # 1) to create a new alternative placement plan (eg placement plan 1A) (see step (7) in FIG. 1).

Further, as shown in FIG. 5, the processing process 2 (for example, the processing process B) executed by a predetermined computer (for example, CPU # 1) in the current allocation plan (for example, the allocation plan 1) is replaced with another computer (for example, the processing process B). For example CPU #
Go to 3) and choose a new alternative placement plan (eg placement plan 1A)
Is performed at an arbitrary frequency (see step (7) in FIG. 1).

Next, the processing efficiency coefficient E1 is calculated from the above equation (1) for the plurality of alternative placement plans. Then, the processing efficiency coefficient E in the current placement plan and the processing efficiency coefficient E1 in the alternative placement plan are compared to determine whether or not the alternative placement plan should be applied (step (3)). That is, when E-E1> 0, the alternative placement plan is output as the optimal placement.

On the other hand, when the processing efficiency coefficient E1 in the alternative placement plan is larger than the processing efficiency coefficient E in the current placement plan, that is, E-E1≤0, the optimum placement is obtained as follows. That is, using the threshold value S which is the application condition of the alternative placement plan, the alternative placement plan is applied with the probability exp ((E-E1) / S) (2). The initial value of the threshold value S is a value sufficiently larger than E-E1.

When the alternative placement plan is applied as described above, the current placement plan is set as the alternative placement plan, and the probability exp ((EE) is generated from the creation of the alternative placement plan (step (2)).
The operation up to the application of the alternative placement plan by 1) / S) is repeated any number of times (step (4)). Iteratively reduces the threshold S at an arbitrary rate (steps (5) and (6)) to optimally place the alternative placement plan when the alternative placement plan is no longer applied to the current placement plan. (Steps (7) and (8)).

According to the second embodiment (annealing method), the threshold value (S) which is the application condition of the alternative placement plan is reduced to optimize the placement plan when the alternative placement plan is no longer applied. Since the arrangement is adopted, even if the process load fluctuates in the middle, the optimum arrangement including the load fluctuation can be obtained.

[0036]

As described above, according to the present invention, the maximum value of the ratio of the processing process load and the processing capacity of the computer connected to the network and the processing which is the sum of the ratio of the total communication load and the computer processing capacity. Since the efficiency coefficient (E) is introduced and the processing process is allocated to each computer so that the processing efficiency coefficient (E) is minimized, the allocation of the processing process to the computer connected to the network is performed by each computer. Can be automatically arranged to be uniform.

Further, when there is a change in the process load, the change amount of the process load is automatically included in the change in the process load by incorporating it into the maximum value of the ratio of the processing process load and the processing capacity of the computer. It is possible to appropriately arrange the processing processes in each computer. Further, since the communication loads of all the computers connected to the network are included in the processing efficiency coefficient (E), the processing processes can be optimally arranged in the computers in consideration of the communication loads. As described above, effective processing is performed as the entire computer system connected to the network, and the processing efficiency of the computer system is improved.

When the processing process to the computer is arranged by the genetic algorithm method as in the inventions according to claims 2 to 6, the arrangement plan is replaced and the generation is updated until the arrangement plan satisfies a predetermined end condition. By repeating it, the initial placement plan is sequentially shifted to the preferred placement plan and the optimal placement plan is converged, so that the optimal placement can be obtained with high accuracy.

Further, when the processing processes for the computer are arranged by the annealing method as in the inventions according to claims 7 to 8, the threshold value (S) which is the application condition of the alternative arrangement plan is decreased to reduce the alternative arrangement plan. By adopting the placement plan when the application of No. is not performed as the optimum placement, even if there is a change in the process load on the way, the optimum placement including the load change can be obtained.

[Brief description of drawings]

FIG. 1 is a control flowchart showing a first embodiment of a computer placement method according to the present invention.

FIG. 2 is a control flowchart showing a second embodiment of a computer placement method.

FIG. 3 is a procedure (No. 1) of arranging processing processes on a computer in the first and second embodiments.

FIG. 4 is a procedure (No. 2) of arranging processing processes on a computer in the first and second embodiments.

FIG. 5 is a procedure (No. 3) of arranging processing processes on a computer in the first and second embodiments.

FIG. 6 is a block diagram showing a configuration of a computer placement system according to the present invention.

FIG. 7 is a diagram corresponding to FIG. 6 showing a conventional technique.

[Explanation of symbols]

1 (CPU # 1, CPU # 2, CPU # 3 ... CPU #
n) Computer 2 (A, B, C, D ... N) Processing process 3 Network 5 Optimal placement device 6 Computer processing capacity setting table P ₁ , P ₂ , P ₃ , P ₄ ... P _n Computer processing capacity L ₁ , L ₂ , L ₃ , L ₄ ... L _n process load T ₁₂ , T ₁₃ , T ₁₄ , T ₁₅ ... T _1n communication load 7 processing process and communication load setting table

Claims (8)

[Claims] 1. A method of arranging a plurality of computers connected to a network according to the relationship between the processing capacity of the computer and the processing process load from the network to each computer, wherein the computer is provided for each computer. Of the processing process load (L) and the processing capacity (P) of the computer, the processing process load capacity ratio (L / P) is calculated, the maximum value is selected, and data is transmitted and received between the computers. The communication load capacity ratio (T / P), which is the ratio of the communication load (T) required by the computer to the processing capacity (P) of the computer, is calculated for each computer, and the total sum is calculated. / P) maximum value and the sum of the communication load capacity ratio (T / P), which is the processing efficiency coefficient (E = (L
/ P) _max + α · Σ (T / P) (α is a weighting coefficient of the communication load with respect to the computer load)) is arranged in the network so that the plurality of computers are arranged in the network. Arrangement method. 2. A computer arranging method for arranging a plurality of computers connected to a network according to a relation between a processing capacity of the computer and a processing process load from the network to each computer. An arbitrary initial placement plan of the processing process is created, and the initial placement plan and the rearrangement placement plan rearranged based on the initial placement plan are the processing process load (L) of the computer and the processing capacity (P) of the computer. , A processing efficiency coefficient E (E = (L
/ P) _max + α · Σ (T / P): Here (L / P)
_max is the maximum value of the processing process load capacity ratio (L / P),
Σ (T / P) is the sum of the communication load capacity ratios (T / P) of each computer, and α is the weighting factor of the communication load with respect to the computer load), and the accumulated values are calculated. Generations are updated and repeated until E becomes smaller than a predetermined value and a predetermined end condition is satisfied, and when the end condition is satisfied, the allocation plan with the minimum processing efficiency coefficient E is selected. A method for arranging computers, which outputs as an optimum arrangement. 3. The computer arranging method according to claim 2, wherein the termination condition is set when the processing efficiency coefficient E becomes smaller than a predetermined threshold value. 4. The rearrangement placement plan is created by replacing a computer that executes a predetermined processing process among a plurality of processing processes between arbitrary two sets of placement plans in a plurality of sets of initial placement plans. The method for arranging a computer according to claim 2, wherein: 5. The computer arranging method according to claim 2, wherein the rearrangement arrangement plan is created by exchanging processing processes at a constant frequency among the computers in one arrangement proposal. 6. The computer according to claim 2, wherein the rearrangement placement plan is created by moving a processing process executed by a predetermined computer in the placement plan of the plurality of computers to another computer. How to place. 7. A computer arranging method for arranging a plurality of computers connected to a network according to a relation between a processing capacity of the computer and a processing process load from the network to each computer, Regarding the current placement plan corresponding to the processing process load, the processing efficiency is determined by the processing process load (L) of the computer, the processing capacity (P) of the computer, and the communication load (T) required to send and receive data between the computers. Coefficient E (E = (L / P) _max
+ Α · Σ (T / P): Here, (L / P) _max is the maximum value of the processing process load capacity ratio (L / P), and Σ (T / P) is each of the communication load capacity ratio (T / P). Computer sum, α
Calculates a weighting factor of the communication load with respect to the computer load), replaces the processing process at any position among the plurality of computers in the initial placement plan to set the alternative placement plan, and then performs the above-mentioned processing for the alternative placement plan. The processing efficiency coefficient E1 corresponding to the efficiency coefficient E is calculated, and when the processing efficiency coefficient E1 in the alternative placement plan is smaller than the processing efficiency coefficient E in the current placement plan (that is, E-E1> 0), the alternative placement A method for arranging computers, characterized in that the plan is output as an optimum arrangement. 8. A computer arranging method for arranging a plurality of computers connected to a network according to a relation between a processing capacity of the computer and a processing process load from the network to each computer, wherein Regarding the current placement plan corresponding to the processing process load, the processing efficiency is determined by the processing process load (L) of the computer, the processing capacity (P) of the computer, and the communication load (T) required to send and receive data between the computers. Coefficient E (E = (L / P) _max
+ Α · Σ (T / P); where (L / P) _max is the maximum value of the processing process load capacity ratio (L / P), and Σ (T / P) is each of the communication load capacity ratio (T / P). Computer sum, α
Calculates a weighting factor of communication load with respect to computer load), sets a substitute placement plan by exchanging processing processes at arbitrary positions among a plurality of computers in the current placement plan, and then, regarding the substitute placement plan, A processing efficiency coefficient E1 corresponding to the processing efficiency coefficient E is calculated, and when the processing efficiency coefficient E1 in the alternative placement plan is larger than the processing efficiency coefficient E in the current placement plan (that is, E−E1 ≦ 0).
In this case), the threshold value (S) that is the application condition of the alternative placement plan is used to determine whether or not the alternative placement plan should be applied to the current placement plan. ) Is repeatedly reduced while outputting the alternative placement plan as an optimal placement when the alternative placement plan is no longer applied to the current placement plan.