Dynamic load-distributed computer system using estimated expansion ratios and load-distributing method therefor

Abstract

In a method for distributing a transaction in a computer system including a plurality of computers each formed by processors, a plurality of estimated loads are calculated for the computers. Then, a plurality of estimated process time expansion ratios are calculated for the computers, in accordance with the estimated loads and the numbers of processors in the computers. Then, a plurality of load indexes one for each of the computers are calculated in accordance with the estimated process time expansion ratios. Finally, the transaction is distributed in accordance with the load indexes.

Description

BACKGROUND OF THE INVENTION

[0001] 1. Field of the Invention

[0002] The present invention relates to a dynamic load-distributed computer system including tightly-coupled computers which is called a cluster structure.

[0003] 2. Description of the Related Art

[0004] In a load-distributed computer system including tightly-coupled computers, each computer formed by a plurality of symmetrical type multi-processors (SMPS), i.e., central processing units (CPUs), when a large number of messages requesting transaction processes arrive at the computer system, the messages are optimally distributed to the computers to enhance the performance of the computer system. Generally, since each transaction process is a small job, after one of the computers is selected to process this transaction process, the transaction process is completely carried out by the selected computer. Also, since transaction processes are generally interactively processed, the target of the load distribution is to minimize the average and deviation values of response time. When selecting one of the computers, one computer having the minimum load is selected in accordance with a load index such as the CPU utilization rate, the number of executed transactions and the past response time individually or in combination.

[0005] In a first prior art load-distributed computer system (see; JP-A-10-312365), a load such as a CPU utilization rate for each of the computers is measured and stored at predetermined time periods. Then, when a message requesting a transaction process arrives from a terminal unit, one of the computers having the minimum load is selected, so that this message is allocated to the selected computer. On the other hand, the terminal unit always determines whether or not a response time is larger than a threshold value. Only when the response time is larger than the threshold value, does the terminal unit request the computer system to change the selection of the computers.

[0006] In the above-described first prior art load-distributed computer system, however, since the CPU utilization rate as the first load index is calculated by the past average value which does not reflect the accurate current load, the reliability is not high. Particularly, under a dynamic control, when the same load index data is used until the next measuring timing, messages are concentrated on one of the computers whose load has been believed to be the minimum value, which would cause a seesaw phenomenon in the load. Also, the CPU utilization rate is not appropriate in a computer system including SMP computers. On the other hand, the response time as the second load index is used for switching the selected computer, which would be useful if the overhead is decreased. However, when a message requesting a transaction process arrives, the response time thereof is not obtained at this time, so that the response time is not an appropriate load index.

[0007] The first prior art load-distributed computer system using the CPU utilization rate as a load index is also disclosed in JP-A-2001-34591. In JP-A-2001-34591, the distribution of transactions are basically concentratedly controlled by using a round-robin discipline. In this case, when a selected computer is overloaded, another computer is selected. Also, when all the computers are overloaded, one of the computers having the minimum load is selected. Whether or not one computer is overloaded is determined by whether or not the load of the computer such as the CPU utilization rate thereof is higher than an upper limit. In this case, the overload amount is represented by the number of transactions over the upper limit of the CPU utilization rate. For example, if the CPU utilization rate is 90% , the upper limit is 60% and the number of executing-transactions is 10, the overload amount is

10·(0.9/0.6−1)=5

[0008] That is, the determination of overload is basically carried out by the CPU utilization rate. However, as explained above, the reliability for a dynamic control is not high, and this determination is not appropriate in a SMP computer. Additionally, the number of transactions is assumed to be linear with the CPU utilization rate, which contradicts with a queue theory in which the number of clients has no relationship to the utilization rate particularly with a high load. Therefore, when all the computers are overloaded, the load distribution may be not appropriate. Further, this technology is not intended to decrease the minimum response time for arrival transactions and for the entire system.

[0009] In a second prior art load-distributed computer system (see JP-A-10-27168), each computer stores a response time of the latest transaction and multiplies this response time by the number of executed transactions therein, to obtain a load index. That is, when a message requesting a transaction process arrives, the load indexes of all the computers are calculated, and this message is sent to one of the computers having the minimum load index.

[0010] In the above-described second prior art load-distributed computer system, however, it is questionable whether the load index obtained based upon the response time of the latest transaction represents a typical response time of the computer. This response time reflects the congestion of the computer and the job characteristics such as a pure process time and a CPU time/input and output time ratio of the latest transaction. If the job characteristics are the same for all messages, the above-mentioned load index is appropriate. However, since there are actually different job characteristics in transaction, the above-mentioned load index is not appropriate. Also, no consideration is given to nonhomogeneous SMP computers In a third prior art load-distributed computer system (see: JP-A-7-302242), a load is periodically detected and a load tendency Tr is calculated by

T r =(W2−W1)/(T2−T1)

[0011] where W1 is a load detected at time T1; and

[0012] W2 is a load detected at time T2. When a message requesting a transactionprocess is accepted, after adefinite time period Ti had passed, it is determined whether or not Tr·Ti≦Wt where Wt is a predicted load is satisfied. If Tr·Ti≦Wt, the subject computer carries out this transaction process. Otherwise, the requesting message is sent to another computer which has a lower load.

[0013] In the above-described third prior art load-distributed computer system, however, it is unclear what the predicted load is defined by. Also, the predicted load Wt, which is calculated by the linear extrapolation method, is not a good predicted value. On the other hand, a predicted load is actually determined microscopically, not macroscopically. That is, a predicted load is dependent upon the current state of the computer, an end timing of an executing-transaction, a timing for receiving a new transaction and the like. Therefore, since the predicted load of the third prior art load-distributed computer system is dependent upon the past load, the predicted load is not appropriate.

[0014] The above-described first, second and third prior art load-distributed computer systems are intended to minimize the response time of a message requesting a transaction process which has just arrived, not to minimize the response time of the entire system.

[0015] In a fourth prior art load-distributed computer system (see: Hisao Kameda et al., “Optimal Load Balancing in Distributed Computer Systems”, Springer-Verlag, pp. 230-232, 1997), if the number of CPUs in each computer sharing a disk appartus is 1, a response time is defined by the utilization time of the CPU and the input/output time of the file apparatus. When a plurality of transactions are carried out, a queue of transactions may be generated before each computer. In the fourth prior art load-distributed computer system, the following two load indexes are defined:

f i =s i ·(ni+1)2

F i =S i ·(ni+1)

[0016] where i is a computer number,

[0017] s is an average service time of the CPU of a transaction;

[0018] n is the number of transactions in the CPU. The load indexes are introduced by using an equilibrium state of an open-type queue network model where the number of CPUs is 1, to minimize the average response time. Actually, the formula Fi=si·(ni+1) represents an averge time of a transaction in the CPU, and therefore, an average response time is represented by this averge time plus an average time of the input/output time. The two load indexes have been proved to be optimum in view of a static load distribution.

[0019] In the above-described fourth prior art load-distributed computer system, however, since a dynamic control requires control at every moment, si and ni should use current values, not average values. Note that it is possible to measure a current value of ni, but it is impossible to directly measure a current value of si, since si reflects the characteristics of a job mix executed in the computer i. Thus, si is used as an equilibrium average value. An equilibrium average value may be used if the job characteristics are the same and have a small deviation. However, since there are actually different job characteristics, a dynamic control using the equilibrium average value may deteriorate. Additionally, the above-mentioned load indexes fi and Fi are applied to computers each having one CPU, but are not applied to SMP computers.

SUMMARY OF THE INVENTION

[0020] It is an object of the present invention to provide a dynamic load-distributed computer system having a plurality of closely-coupled computers each having a plurality of processors (CPUs), capable of improving the response characteristics such as the average and deviation characteristics of the response time.

[0021] Another object is to provide a load-distributing method for the above-mentioned dynamic load-distributed computer system.

[0022] According to the present invention, in a method for distributing a transaction in a computer system including a plurality of computers each formed by processors, a plurality of estimated loads are calculated for the computers. Then, a plurality of estimated process time expansion ratios are calculated for the computers, in accordance with the estimated loads and the numbers of processors in the computers. Then, a plurality of load indexes one for each of the computers are calculated in accordance with the estimated process time expansion ratios. Finally, the transaction is distributed in accordance with the load indexes.

[0023] Also, in a load-distributed computer system including a plurality of computers each including processors, a relay distributing unit for receiving a message requesting a transaction process, and a channel communication network connected between the computers and the relay distributing unit, each of the computers is constructed by-a first load data measuring unit for measuring a load of each of the computers, and a transaction processing unit for processing transactions. Also, the relay distributing unit is constructed by a second load data measuring unit for receiving the load to calculate an estimated load in accordance with the load, a load data storing unit for storing the estimated load data, and an executingcomputer selecting unit for calculating a plurality of estimated process time expansion ratios one for each of the computers, in accordance with the stored estimated loads and the number of processors in each of the computers, calculating a plurality of load indexes, one for each of the computers, in accordance with the estimated process time expansion ratios, selecting one of the computers having the minimum load index in accordance with the load indexes, and causing the transaction to be processed by the transaction processing unit of the selected computer.

[0024] Further, in a load-distributed computer system including a plurality of computers each including processors, each for receiving a message requesting a transaction process, and an exchange/store unit connected to the computers, each of the computers is constructed by a first load data measuring unit for measuring a load of each of the computers, a transaction processing unit for processing transactions, a second load data measuring unit for receiving the load to calculate an estimated load in accordance with the load, a load data storing unit for storing the estimated load data, an executing-computer selecting unit for calculating a plurality of estimated process time expansion ratios one for each of the computers, in accordance with the stored estimated loads and the number of processors in each of the computers, calculating a plurality of load indexes, one for each of the computers, in accordance with the estimated process time expansion ratios, determining whether or not each of the computers is to process the transaction process in accordance with the load indexes, causing the transaction process to be processed by the transaction processing unit when the each of the computers is to process the transaction process, and transmitting the message via the exchange/store unit to another computer when each of the computers is not to process the transaction process.

BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The present invention will be more clearly understood from the description set forth below, with reference to the accompanying drawings, wherein:

[0026] FIG. 1 is a block circuit diagram illustrating a first embodiment of the load-distributed computer system according to the present invention;

[0027] FIG. 2 is a circuit diagram of one computer of the computer apparatus of FIG. 1;

[0028] FIG. 3A is a flowchart for explaining the operation of the load data measuring unit of the computer of FIG. 1,

[0029] FIG. 3B is a flowchart for explaining the operation of the load data measuring unit if the relay distributing of FIG. 1,

[0030] FIG. 4, is a flowchart for explaining the operation of the executing-computer selecting unit of FIG. 1;

[0031] FIG. 5 is a table of the load data storing unit of Fig.

[0032] FIG. 6 is a table of load indexes used in the first embodiment of the present invention;

[0033] FIG. 7 is a flowchart illustrating a modification of FIG. 4 in which the load index Lp2 of FIG. 6 is used;

[0034] FIG. 8A is a flowchart illustrating a modification of FIG. 4 in which the load index La1 of FIG. 6 is used;

[0035] FIG. 8B is a flowchart illustrating a modification of FIG. 8A in which the load index L2 of FIG. 6 is used;

[0036] FIG. 8C is a flowchart illustrating a modification of FIG. 4 in which the load index La3 of FIG. 6 is used;

[0037] FIG. 9A is a flowchart illustrating a modification of FIG. 4 in which the load index Lk1 of FIG. 6 is used;

[0038] FIG. 9B is a flowchart illustrating a modification of FIG. 8A in which the load index Lk2 of FIG. 6 is used;

[0039] FIG. 9C is a flowchart illustrating a modification of FIG. 4 in which the load index Lk3 of FIG. 6 is used;

[0040] FIG. 10 is a block circuit diagram illustrating a second embodiment of the load-distributed computer system according to the present invention;

[0041] FIG. 11A is a flowchart for explaining the operation of the load data measuring unit lit of the computer of FIG. 10;

[0042] Fig. 11B is a flowchart for explaining the operation of the load data measuring unit 1i5 of the relay distributing of FIG. 10;

[0043] FIG. 12 is a flowchart for explaining the operation of the executing-computer selecting unit of FIG. 10;

[0044] FIG. 13 is a block circuit diagram illustrating a third embodiment of the load-distributed computer system according to the present invention; and

[0045] FIGS. 14 and 15 are graphs showing the effect of the present invention as compared with the prior art.

DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0046] A first embodiment of the load-distributed computer system according to the present invention will now be explained with reference to FIGS. 1, 2, 3A, 3B, 4, 5, 6, 7, 8A, 8B, 8C, 9A, 9B and 9C.

[0047] In FIG. 1, which illustrates a first embodiment of the load-distributed computer system according to the present invention, reference numeral 1 designates a computer apparatus formed by a plurality of computers 11, 12, . . ., 1n each including a load data measuring unit lit, a transaction processing unit 1i2 and a communication unit 1i3. The computer group is connected to a file apparatus 2 formed by a plurality of disk units through a single high speed channel communication network 3.

[0048] Also, a relay distributing apparatus 4 is connected via the high speed channel communication network 3 to the computer apparatus 1. The relay distributing apparatus 4 is formed by a load data storing unit 41, a load data measuring unit 42, an executing-computer selecting unit 43 and a communication unit 44.

[0049] The load-distributed computer system of FIG. 1 is connected via a communication network 5 to a large number of external terminal units 6.

[0050] In FIG. 2, which illustrates one computer 1i of the computer apparatus 1 of FIG. 1, the computer 1i is a symmetrical type multi-processor (SMP) which includes a plurality of central processing units CPUI, CPU2, . . . , CPUmi which share a main storage unit MSU and the disk units of the file apparatus 2. That is, the computers 11, 12, . . . , 1n do not share one main storage unit. Generally, the resource utilization characteristics of one transaction is dependent upon the utilization of the CPUs and the number of input/output operations of the main storage unit MSU and the file apparatus 2. Therefore, when a plurality of transactions are carried out in the computer 1i, a conflict may occur in the resources, i.e., the main storage unit MSU and the file apparatus 2, so that transactions are queued at the computer 1i. In this case, the higher the resource utilization characteristics of the computer 1i, the longer the queue of transactions. As illustrated in FIG. 2, only one series of transactions are queued at the computer 1i, and its operating system assigns the queue of transactions to the CPUs homogeneously. In this case, the input/output operation characteristics of the CPUs for the file apparatus 2 are homogeneous in view of the performance, so that the processing time of the file apparatus 2 can be assumed to be equal for the computer 1i.

[0051] In FIG. 2, Ni designates the number of transactions in total which are currently assigned to the computer 1i. That is, the transaction number Ni includes the number of transactions queued at the computer 1i, the number of transactions which are currently being executed in the computer 1i and the number of transactions stored in the file apparatus 2 executed by the computer 1i. Also, Pi designates the number of transactions which are currently executed in the computer 1i. That is, the transaction number Pi includes the number of transactions queued at the computer 1i and transactions which are currently being executed in the computer 1i. Further, Di designates the number of transactions stored in the file apparatus 2 executed by the computer 1i. Therefore, generally, Ni=Pi+Di.

[0052] In the computer system of FIGS. 1 and 2, in the computer 1i, at every predetermined time period, the load data measuring unit 1i1 measures a load data and transmits it via the communication unit 1i3 to the relay distributing apparatus 4. On the other hand, in the relay distributing apparatus 4, the load data measuring unit 42 receives the load data from the computer 1i, and then, changes the load data into an estimated load data and stores it in the load data storing unit 41.

[0053] Also, in the computer system of FIGS. 1 and 2, one transaction (job) is processed as follows. First, a message requesting a transaction process is transmitted from one of the terminal units 6 via the communication network 5 to the relay distributing apparatus 4. Then, the executing-computer selecting unit 43 calculates estimated expansion ratios of the computers 11, 12, . . . , In from the estimated load data and determines in which the computer 1i the transaction process is to be executed in accordance with the estimated expansion ratios. Then, the executing-computer selecting unit 43 sends the above-mentioned message to the determined computer such as 1i. Then, after the computer 1i carries out this transaction process, the computer 1i transmits a response message to the relay distributing apparatus 4, so that this response message is transmitted via the communication network 5 to the one of the terminal units 6.

[0054] The load measuring unit 1i1, the transaction processing unit 1i2 and the communication unit 1i3 of the computer 1i are usually realized by using software (programs). In this case, when the computer 1i is started up, such software is loaded thereinto. Similarly, the load data storing unit 41, the executing-computer selecting unit 42, the load measuring unit 42 and the communication unit 44 of the relay distributing apparatus 4 are usually realized by using software (programs). In this case, when the relay distributing apparatus 4 is started up, such software is loaded thereinto.

[0055] The operations of the load data measuring unit 1i1 and the load data measuring unit 42 are explained next with reference to flowcharts as illustrated in FIGS. 3A and 3B.

[0056] The flowchart as illustrated in FIG. 3A is carried out by the load data measuring unit 1i1 at every predetermined time period smaller than 100 msec. In this case, the smaller the predetermined time period, the higher the accuracy of load distribution. On the other hand, the smaller the predetermined time period, the smaller the overhead. The predetermined time period is preferably about 10 msec.

[0057] First, at step 301, the load data measuring unit 1i1 calculates the number Npi of transactions in total currently-assigned to the computer 1i.

[0058] Next, at step 302, the load data measuring unit 1i1 calculates the number Ppi of transactions which are currently being executed in the computer 1i.

[0059] Next, at step 303, the load data measuring unit 1i1 transmits the current load data Npi and Ppi to the load measuring unit 42 of the relay distributing unit 4.

[0060] Thus, the operation of the flowchart of FIG. 3A is completed by step 304.

[0061] The flowchart of FIG. 3B is carried out by the load data measuring unit 42 every time the load data measuring unit 42 receives the current load data Npi and Ppi from the load data measuring unit 1i1.

[0062] First, at step 311, the load data measuring unit 42 calculates an estimated number Nei by

N ei ←a·N pi +(1−a)·Nei

[0063] where “a” is a definite value (0<a≦1). Then, the estimated number Nei is stored in the load data storing unit 41 as shown in FIG. 5.

[0064] Next, at step 312, the load data measuring unit 42 calculates an estimated number Pei by

P ei ←a·P pi +(1−a)·Pei

[0065] Then, the estimated number Pei is stored in the load data storing unit 41 as shown in FIG. 5.

[0066] Thus, the flowchart of FIG. 3B is completed by step 313.

[0067] In FIG. 3B, the estimated load data Nei and Pei are reflected by the past values of Npi and Ppi. In this case, the smaller the value “a”, the larger the ratio of the past values to the current values in the estimated load data Nei and Pei. If the above-mentioned time period is small, the value “a” is 0.1, for example, to increase the reliability of the estimated load data Nei and Pei.

[0068] The operation of the executing-computer selecting unit 43 is explained next with reference to a flowchart as illustrated in FIG. 4, which is carried out every time one of the terminal units 6 transmits a message requesting a transaction process via the communication network 5 and the communication unit 44 to the executing-computer selecting unit 43.

[0069] First, at step 401, a value i is initialized, i.e.,

i←1

[0070] Next, at step 402, the executing-computer selecting unit 43 calculates a corrected number Nri of the estimated number Nei by

N ri <w·N pi +(1−w)·Nei.

[0071] where w is a definite value (0<w≦1). In this case, the value w is preferably about 0.8. That is, the estimated numbers Nei and Pei are obtained at a predetermined timing, while the corrected number Nri is obtained as an optimal value at this timing in accordance with the estimated numbers Nei and Pei.

[0072] Next, at step 403, it is determined whether Nri≧Nei is satisfied. As a result, if Nri≧Nei, the control proceeds to step 404 which calculates a corrected number Pri by

P ri ←P ei ·(Nri−Nei)

[0073] That is, the corrected number Pri is increased by (Nri−Nei )

[0074] On the other hand, at step 403, if Nri≧Nei, the control proceeds to step 405 which calculates a corrected number Pri by

P ri ←P ei ·N ri /N ei

[0075] That is, the corrected number Pri is decreased by the ratiori/Nei.

[0076] Next, at step 406, a corrected number Nsi per one CPU is calculated by

N si ←N ri /mi

[0077] Also, a corrected number Psi per one CPU is calculated by

P si ←P ri /mi

[0078] Next, at step 407, it is determined whether or not Nsi>Psi is satisfied. As a result, if Nsi>Psi) the control proceeds to step 408 which calculates an estimated expansion ratio Epi by

E pi ←X /(X−Psi·Psi)

[0079] where X is Nsi·(Psi+1).

[0080] The above-mentioned estimated expansion ratio E can be proved as follows (see formulae (8.1) and (8.3) of page. 228 of Hisao Kameda et al., “Optimal Load Balancing in Distributed Computer”, Springer-Verlag, 1997). Here, the transmission time of transactions not including waiting time for the file appartus 2 is assumed to be exponentially distributed, i.e., have a Poison distribution. The following parameters are defined:

[0081] P: the average number of transactions in one computer;

[0082] D: the average number of transactions in the file apparatus for that computer;

[0083] N: the average total number of transactions for that computer;

[0084] T: estimated process time per one transaction;

[0085] t: pure process time per one transaction;

[0086] s: CPU utilization time per one transaction; and

[0087] d: input/output time per one transaction.

[0088] In this case,

N=P+D   (1)

t=s+d   (2)

[0089] Also, a CPU utilization ratio R is defined by

P+ 1=1/(1−R)

∴R=P/(P+1)  (3)

[0090] Further, due to no waiting time for the file apparatus 2,

d/s=D/R   (4)

[0091] From the formulae (1), (2), (3) and (4), 1$\begin{array}{cc}\begin{array}{c}s=\mathrm{Pt}/\left(P+D+\mathrm{PD}\right)\\ =\mathrm{Pt}/\left(N\left(P+1\right)-P·P\right)\end{array}& \left(5\right)\end{array}$

[0092] On the other hand, the estimated process time T is given by 2$\begin{array}{cc}\begin{array}{c}T=s\left(P+1\right)-d\\ =\mathrm{sP}+t\end{array}& \left(6\right)\end{array}$

[0093] Therefore, 3$\begin{array}{cc}\begin{array}{c}T=P·\left(\mathrm{Pt}/N\left(P+1\right)-P·P\right)+t\\ =N\left(P+1\right)t/\left(N\left(P+1\right)-P·P\right))\\ =\mathrm{Xt}/\left(X-P·P\right)\end{array}& \left(7\right)\end{array}$

[0094] where X=N(P+1).

[0095] Thus, an estimated expansion ratio E is 4$\begin{array}{cc}\begin{array}{c}E=T/t\\ =X/\left(X-P·P\right)\end{array}& \left(8\right)\end{array}$

[0096] In Formula (8), if N and P are replaced by Nsi and Psi, respectively, the above-mentioned estimated expansion ratio Epi can be obtained.

[0097] In the symmetrical-type multi-processor (SMP), the numbers Nsi and Psi per one CPU are used. The estimated expansion ratio Epi using Nsi and Psi are not accurate for the symmetrical-type multi-processor; however, the estimated expansion ratio Epi is actually useful in practice. Note that this is quite different from a case where 1/ml of the load of the same job mix is applied to one CPU.

[0098] Also, the above-mentioned estimated expansion ratio Epi is based upon an equilibrium state. In the equilibrium state, the estimated expansion ratio Epi can be obtained by the estimated numbers Nei and Pei in view of a static load distribution. In this embodiment, however, the estimated expansion ratio Epi is obtained by the corrected numbers Nri and Pri affected by the current number Npi in view of a dynamic load distribution.

[0099] Returning to FIG. 4, at step 407, if Nsi<Psi, the control proceeds to step 409 which causes the estimated expansion ratio Epi to be Nsi+1.0.

[0100] The operations at steps 402 through 409 are repeated for all the computers 1i (i=1, 2, . . . , n) by steps 410 and 411.

[0101] Next, at step 412, the executing-computer selecting unit 43 selects one of the computers having the minimum estimated expansion ratio Epi.

[0102] Next, at step 413, the executing-computer selecting unit 43 sends a message requesting a transaction process to the computer selected at step 412. As a result, the selected computer carries out a transaction process.

[0103] Thus, the flowchart of FIG. 4 is completed by step 414.

[0104] In the above-described embodiment, the estimated expansion ratio Epi(=Lpi) is used as a load index. However, as shown in FIG. 6, an estimated expansion ratio Nri·Epi/mi per one CPU (=Lp2) can be used as a load index. Note that the load indexes Lp1 and Lp2 are both estimated expansion ratios before scheduling, i.e., before distributing a message requesting a transaction process.

[0105] When the load index Lp2 of FIG. 6 is used, the flowchart of FIG. 4 is modified to a flowchart as illustrated in FIG. 7. That is, at step 701, an estimated expansion ratio Lp2i per one CPU is calculated by

L p2i ←N ri ·E pi /mi

[0106] In this case, at step 412, the executing-computer selecting unit 43 selects one of the computers having the minimum estimated expansion ratio Nri·Epi/mi.

[0107] Also, as shown in FIG. 6, an estimated expansion ratio Eni (=Lai) per one transaction after scheduling, i.e., after distributing a message requesting a transaction process whose characteristics are unknown or not expected.

[0108] When the load index Lai is used, the flowchart of FIG. 4 is modified to a flowchart as illustrated in FIG. 8A. In this case, steps 406 to 409 of FIG. 4 are replaced by steps 801 to 805.

[0109] At step 801, the number Nni of total transactions for the computer 1i and the number Pni of transactions in the computer 1i are increased by

Nni←Nri+w

Pni←Pri+w

[0110] where w is the same value at step 402 of FIG. 4.

[0111] Next, at step 802, a corrected number Nsi′ per one CPU is calculated by

Nsi′←Nni/mi

[0112] Also, a corrected number Psi′ per one CPU is calculated by

Psi′←Pni/mi

[0113] Next, at step 803, it is determined whether or not Nsi′>Psi′ is satisfied. As a result, if Nsi′>Psi′, the control proceeds to step 408 which calculates an estimated expansion ratio Eni by

Eni←X′/(X′−Psi·Psi)

[0114] where X′ is Nsi′·(Psi+1).

[0115] On the other hand, at step 803, if Nsi′≦Psi′ the control proceeds to step 805 which causes the estimated expansion ratio Eni to be Nni′+1.0.

[0116] In this case, at step 412, the executing-computer selecting unit 43 selects one of the computers having the minimum estimated expansion ratio Eri.

[0117] Also, as shown in FIG. 6, an estimated expansion ration Nni·Eni/mi per one CPU (=La2) can be used as a load index. Note that the load index La2 is an estimated expansion ratio after scheduling, i.e., after distributing a message requesting a transaction process whose characteristics are unknown or not expected.

[0118] When the load index La2 of FIG. 6 is used, the flowchart of FIG. 8A is modified to a flowchart as illustrated in FIG. 8B. That is, step 806 is added to steps of FIG. 8A. At step 806, an estimated expansion ratio La2i per one CPU is calculated by

La2i←Nni·Eni/mi

[0119] In this case, at step 412, the executing-computer selecting unit 43 selects one of the computers having the minimum estimated expansion ratio La2i.

[0120] Also, as shown in FIG. 6, an increased estimated expansion ratio mi(La2−Lp2) per one CPU (=La3) can be used as a load index. Note that the load indexes La3 is also an estimated expansion ratios after scheduling, i.e., after distributing a message requesting a transaction process whose characteristics are unknown or not expected.

[0121] When the load index La3 of FIG. 6 is used, the flowchart of FIG. 4 is modified to a flowchart as illustrated in FIG. 8C. In this case, steps 801 to 805 and 807 are added to FIG. 4.

[0122] Steps 801 to 805 are carried out in the same way as in FIG. 8A.

[0123] Next, at step 807, an increased estimated expansion ratio La3i per one CPU is calculated by

La3i←(Nni·Eni−Nri·Epi)/ml

[0124] In this case, at step 412, the executing-computer selecting unit 43 selects one of the computers having the minimum increased estimated expansion ratio La3i.

[0125] Also, as shown in FIG. 6, an estimated expansion ratio Eki(=Lki) per one transaction after scheduling, i.e., after distributing a message requesting a transaction process whose characteristics are known or expected.

[0126] When the load index Lki is used, the flowchart of FIG. 4 is modified to a flowchart as illustrated in FIG. 9A. In this case, steps 406 to 409 of FIG. 4 are replaced by step 901.

[0127] At step 901, Nni/ml and Pni/ml are calculated. That is, a ratio C of a CPU utilization time s to a pure process time t for a transaction is estimated. If C1 is a value of the ratio C in the computer 1i before scheduling is represented by using formula (5), i.e.,

C 1 =P ri /(Nri·(1+Pri)−Pri·Pri))

[0128] Also, C2 is assumed to be a value of the ratio C in the computer 1i after scheduling a message having a ratio C0 in the computer 1i. In this case, there are Nri transactions having an average ratio of C1, and one transaction having the ratio C0 is added thereto. As a result, the number of transactions in the computer 1i is represented by Nni whose average ratio C is C2 is represented by

C 2 =(Nri·C1+C0)/Nni

[0129] On the other hand, if formula (5):

s/t=P /(NP+N−P·P)

[0130] is applied to an after-scheduling state, since s/t=C2,

C 2 =(A·y+A−y2)=y

[0131] where A=Nni/mi=(Nni+w)/mi

y=P ni /mi.

[0132] That is,

C 2 y 2 +(1−C2·A)y−C2·A=0

[0133] when this equation is solved, Pni/mi (=A) which is in this case positive is obtained.

[0134] Then, Nsi and Psi are calculated by

Nsi←Nni/mi

Psi←Pni/mi

[0135] Then, similar steps as steps 407, 408 and 409 of FIG. 4 calculate an estimated expansion ratio Eki.

[0136] In this case, at step 412, the executing-computer selecting unit 43 selects one of the computers having the minimum estimated expansion ratio Eki Further, as shown in FIG. 6, an estimated expansion ratio Nni·Eni/mi per one CPU (=Lk2) can be used as a load index. Note that the load index Lk2 is an estimated expansion ratio after scheduling, i.e., after distributing a message requesting a transaction process whose characteristics are known or expected.

[0137] When the load index Lk2 of FIG. 6 is used, the flowchart of FIG. 9A is modified to a flowchart as illustrated in FIG. 9B. That is, step 902 is added to step of FIG. 9A. At step 901, an estimated expansion ratio Lk2i per one CPU is calculated by

Lk2←Nni·Eki/mi

[0138] In this case, at step 412, the executing-computer selecting unit 43 selects one of the computers having the minimum estimated expansion ratio Lk2i.

[0139] Further, as shown in FIG. 6, an increased estimated expansion ratio mi(Lk2−Lk2) per one CPU (=Lk3) can be used as a load index. Note that the load indexes Lk3 are also estimated expansion ratios after scheduling, i.e., after distributing a message requesting a transaction process whose characteristics are known or expected.

[0140] When the load index Lk3 of FIG. 6 is used, the flowchart of FIG. 4 is modified to a flowchart as illustrated in FIG. 9C. In this case, steps 901 and 903 are added to FIG. 4.

[0141] Step 901 is carried out in the same way as in FIG. 9A.

[0142] Next, at step 902, an increased estimated expansion ratio Lk3i per one CPU is calculated by

Lk3i←(Nni·Eki−Nri·Epi)/ml

[0143] In this case, at step 412, the executing-computer selecting unit 43 selects one of the computers having the minimum increased estimated expansion ratio Lk3i.

[0144] Note that an expansion ratio E designates a ratio of an actual process time including a waiting time to a pure process time of one transaction, and Ei is an expansion ratio in the computer 1i. If the pure process time of a transaction is the same in all the computers 11, 12, . . . , 1n, when this transaction is allocated to one of the computers having the minimum expansion ratio, the process time of this transaction can be minimized. An estimated expansion ratio reflects program characteristics of transactions (job mix) currently being executed in the subject computer including the CPU characteristics and input/output characteristics thereof. Also, when the formula at step 408 is used, an estimated expansion ratio can be obtained by only observed data without knowing the individual characteristics of transactions currently being executed in the subject computer. The formula at step 408 basically belongs to the fourth prior art and is derived by modifying the formula Fi=si(ni+1)2; however, the estimated expansion ratio according to the first embodiment is reflected by the characteristics of the job mix as well as the total process time including the CPU process time and the input/output time thereof, thus improving the accuracy and dynamic characteristics of the computer system. Note that, 5since the formula at step 408 is based upon an equilibrium average state, a perfect reliability cannot be expected.

[0145] In the first embodiment, one of the eight load indexes Lp1, Lp2, La1, La2, La3, Lk1, Lk2 and Lk3 is used. In this case, all the eight load indexes can be calculated and stored in the load data storing unit 41, and as occasion demands, one of the eight load indexes can be selected. Note that, since the load indexes La3 and Lk3 are helpful in decreasing the increased process time by scheduling a transaction, it is expected to minimize the entire process time of the subject computer. In this case, if the characteristics of a given transaction are known or expected, the load index Lk3 should be selected. Otherwise, i.e., if the characteristics of a given transaction are unknown or not expected, the load index Lk3 should not be selected, and instead of this, the load index Lk2 should be selected.

[0146] Thus, in the first embodiment, the relay distributing unit 4 supervises the load states of all the computers 1i in real time. Also, when the relay distributing unit 4 receives a message requesting a transaction process, the relay distributing unit 4 calculates an estimated expansion ratio of each of the computers and sends this message to one of the computers having the minimum expansion ratio. As a result, since a concentrated control by the relay distributing unit 4 can be realized, the overhead can be suppressed and an excellent load distribution can be realized.

[0147] In FIG. 10, which illustrates a second embodiment of the load-distributed computer system according to the present invention, the relay distributing unit 4 of FIG. 1 is not provided. Instead of this, one load data storing unit 1i4, one load data measuring unit 1i5 and one executing-computer selecting unit 1i6 corresponding to the load data storing unit 41, the load data measuring unit 42 and the executing-computer selecting unit 43, respectively, of FIG. 1 are provided, and an exchange/store unit 10 is connected to the computer 11, 12, . . . , 1n. Note that the hardware of the computer 1i is also illustrated in FIG. 2.

[0148] In the computer system of FIG. 10, in each of the computers 1i, at predetermined time periods, the load data measuring unit 1i1 measures a load data. Then, the load data measuring unit 1i5 changes the load data into an estimated load data and stores it in the load data storing unit 1i1. Simultaneously, the estimated load data is transmitted via the exchange/store unit 10 to the load data storing units of all the other computers. Similarly, start and end of processing of each transaction in one computer are transmitted to the other computers. Thus, the latest estimated load data of all the computers are stored in all the computers.

[0149] Also, in the computer system of FIG. 10, a transaction (job) is processed as follows. First, a message requesting a transaction process is transmitted from one of the terminal units 6 via the communication network 5 to a predetermined computer such as 1i. Then, the load data measuring unit 1i5 calculates an estimated expansion ratio of the subject computer 1i. Then, the executing-computer selecting unit 1i6 determines whether the subject computer 1i carries out this message or the subject computer 1i sends this message via the exchange/store unit 10 to another computer, in accordance with the estimated expansion ratio of the subject computer 1i. When the subject computer 1i is determined to carry out the message, the transaction processing unit 1i2 of the subject computer 1i carries out the transaction process. On the other hand, when the other computer is determined to carry out the message, the transaction processing unit thereof carries out the transaction process. After the transaction process is completed, a response message is transferred from the subject computer or the other computer via the exchange/store unit 10 to all the other computers.

[0150] The operations of the load data measuring unit 1i1 and the load data measuring unit 1i5 are explained next with reference to flowcharts as illustrated in FIGS. 11A and 11B.

[0151] The flowchart as illustrated in FIG. 11A is carried out by the load data measuring unit lit at predetermined time periods.

[0152] First, at step 1101, the load data measuring unit 1i1 calculates the number Npi of transactions in total which are currently assigned to the computer 1i.

[0153] Next, at step 1102, the load data measuring unit 1i1 calculates the number Ppi of transactions which are currently being executed in the computer 1i.

[0154] Next, at step 1103, the load data measuring unit lit transmits the current load data the Npi and Ppi to the load measuring unit 1i5.

[0155] Thus, the operation of the flowchart of FIG. 11A is completed by step 1104.

[0156] The flowchart of FIG. 11B is carried out by the load data measuring unit 1i5 every time the load data measuring unit 1i5 receives the current load data Npi and Ppi from the load data measuring unit 1i1.

[0157] First, at step 1111, the load data measuring unit 1i5 calculates an estimated number Nei by

Nei←a·Npi+(1−a)·Nei

[0158] where “a” is a definite value (0<a≦1). Then, the estimated number Nei is stored in the load data storing unit 1i4. Simultaneously, the estimated number Nei is transmitted via the exchange/store unit 10 to the other computers and is stored in the load data storing units thereof.

[0159] Next, at step 1112, the load data measuring unit 1i5 calculates an estimated number Nei by

Pei←a·Ppi+(1−a)·Pei

[0160] Then, the estimated number Pei is stored in the load data storing unit i4. Simultaneously, the estimated number Pei is transmitted via the exchange/store unit 10 to the other computer and is stored in the load data storing units thereof.

[0161] Thus, the flowchart of Fig. 11B is completed by step 1113.

[0162] In FIG. 11B, the estimated load data Nei and Pei are the same as those of FIG. 3B.

[0163] The operation of the executing-computer selecting unit 1i6 is explained next with reference to a flowchart as illustrated in FIG. 12, which is carried out every time one of the terminal units 6 transmits a message requesting a transaction process via the communication network 5 and the communication unit 1i3 to the executing-computer selecting unit 1i6.

[0164] In FIG. 12, steps 1201 and 1202 are added to the steps of FIG. 4.

[0165] That is, after all the estimated expansion ratios Ep1, Ep2, . . . , Epn are calculated, at step 1201, it is determined whether or not the estimated expansion ratio Epi of the subject computer 1i is smaller than a threshold value TH such as 1.3. As a result, only when Epi<TH, does the control proceed to step 1202 which operates the transaction processing unit 1i2 of the subject computer 1i, so that the transaction process is carried out in the computer 1i. Otherwise, i.e., if Epi≧TH, the control proceeds to steps 412 and 413.

[0166] At step 412, the executing-computer selecting unit 1i6 selects one of the computers having the minimum estimated expansion ratio Epi.

[0167] Then, at step 413, the executing-computer selecting unit 43 sends a message requesting a transaction process to the computer selected at step 412.

[0168] After the transaction processing unit of any computer completes the transaction process, this transaction processing unit sends a response message via the exchange store unit 10 to the other computers.

[0169] Note that, at step 412, if the computer 1i, is determined to have the minimum estimated expansion ratio in spite of the estimated expansion ratio Epi being large (Epi≧TH), the computer 1i is selected after all.

[0170] In FIG. 12, at step 1201, the threshold value TH is a definite value such as 1.3; however, the threshold value TH can be dependent upon the minimum estimated expansion ratio Epmin. For example, TH=1.3·Epmin. In this case, at step 1201, the minimum estimated expansion ratio Epmin is calculated in advance.

[0171] Even in the second embodiment, any of the other load indexes Lp2, La1, La2, La3, Lk1, Lk2 and Lk3 of FIG. 6 can be used instead of the load index Lp1(=Epi). For example, if the load index La3 or Lk3 is used, the threshold value TH is a smaller value such as 0.02.

[0172] Further, in FIG. 12, at step 1201, the threshold value TH is dependent upon the predetermined time period of carrying out the flowcharts as illustrated in FIGS. 11A and 11B. In this case, the longer this time period, the larger the threshold value TH. This is because the increased predetermined time period deteriorates the reliability of the load index.

[0173] In FIG. 10, if the exchange/store unit 10 has a high speed storage, the function of the load data storing units 114, 124, . . . , 1i4, . . . , 1n4 of all the computers can be moved to the exchange/store unit 10, which would decrease the overhead by renewing the load data. Also, a message requesting a transaction process can be temporarily stored in the exchange/store unit 10, and in this case, the computers can access the messages stored in the exchange/store unit 10 asynchronously.

[0174] Thus, in the second embodiment, since the relay distributing unit 4 of FIG. 1 is not provided, the manufacturing cost can be decreased as compared with the first embodiment. Although the load distribution is not centralized, when a subject computer receives a message requesting a transaction process, the subject computer calculates an estimated extension ratio and determines one of the computers to execute the transaction process, thus decreasing the overhead and realizing a high response characteristic.

[0175] In FIG. 13, which illustrates a third embodiment of the load-distributed computer system according to the present invention, a relay provisional-distributing unit 7 is provided between the computers 11, 12, . . . , 1n and the communication network, 5 of FIG. 10, in order to distribute a message requesting a transaction process among the computer 11, 12, . . . , in statically or quasi-statically in accordance with the following methods in combination or individually.

[0176] 1) distribution by dividing the terminal units 6 into a plurality of groups each corresponding to one of the computers 11, 12, . . . , 1n;

[0177] 2) distribution by sequentially messages arrived at the relay provisional-distributing unit 7 to the computers 11, 12, . . . , 1n in accordance with the numbers of CPUs therein; and

[0178] 3) stochastic distribution in accordance with result data.

[0179] In the distribution 1), each of the terminal units 6 is allocated in advance to one of the computers 11, 12, . . . 1n. In the second embodiment, the distribution may be adopted; however, the allocation of the terminal units 6 to the computer 11, 12, . . . , In is fixed. On the other hand, in the third embodiment, the allocation of the terminal units 6 to the computers 11, 12, . . . , 1n can be changed by the relay provisional-distributing unit 7, to realize an optimum allocation.

[0180] In the distribution 2), if a sequence of m1, m2, . . . , mn messages arrive at the relay provisional-distributing unit 7, the first ml messages to arrive at the relay provisional-distributing unit 7 are allocated to the computer 11, the second m2 messages arrived at the relay provisional-distributing unit 7 are allocated to the computer 12, . . . , the final mn messages to arrive at the relay provisional-distributing unit 7 are allocated to the computer In. Particularly, when the job characteristics of a large number of transactions belong to the same class, the balancing effect of the load can be optimized.

[0181] In the distribution 3), the ratio of messages allocated to the computer 1i is determined in accordance with the load thereof periodically such as at every 1 sec or 10 sec.

[0182] The operation of the computer system of FIG. 13 is substantially the same as that of the computer system of FIG. 10. Only if the distribution 3) is adopted, does the relay provisional-distributing unit 7 receive the load states of the computers 11, 12, . . . , 1n periodically.

[0183] Thus, in the third embodiment, since the relay provisional-distributing unit 7 has a limited function, the manufacturing cost can be also decreased as compared with the first embodiment. Also, since the relay provisional-distributing unit 7 carries out a provisional load-distribution to realize a quasi-static load-distribution, the improvement of the response performance such as the average and deviation characteristics of the response time can be expected. Additionally, since a balancing load-distribution is provisionally realized at the relay provisional-distributing unit 7, the fluctuation of unbalancing of load-distribution between the computers 11, 12, . . . , in can be suppressed, which would decrease the transfer of messages from one computer to another.

[0184] The effect of the present invention as compared with the prior art is explained next with reference to FIGS. 14 and 15 which show the average and deviation characteristics, respectively, of the response time. In FIGS. 14 and 15, a computer system is assumed to have eight computers whose CPU numbers are 4, 1, 1, 1, 1, 1, 2 and 2, and messages are assumed to be randomly generated. Also, Nc designates a no load-distribution state; Nc designates a dynamic load-distribution state using Nc/m as a load index where Nc is the number of transactions currently being executed and m is the number of CPUs; and Lp1, La2 and Lk3 designate load-distribution state using the load indexes Lp1, La2 and Lk3 according to the present invention. As shown in FIGS. 14 and 15, the average and deviation characteristics of the load-distribution according to the present invention are both improved as compared with the prior art.

[0185] As explained hereinabove, according to the present invention, the response performance characteristics such as the average and deviation characteristics of the response time can be improved.

Claims

1. A method for distributing a transaction in a computer system including a plurality of computers each formed by processors, comprising the steps of: calculating a plurality of estimated loads one for each of said computers; calculating a plurality of estimated process time expansion ratios one for each of said computers, in accordance with said estimated loads and the numbers of processors in said computers; calculating a plurality of load indexes, one for each of said computers, in accordance with said estimated process time expansion ratios; and distributing said transaction in accordance with said load indexes.

2. The method as set forth in claim 1, wherein said estimated load calculating step calculates said estimated loads at predetermined time periods.

3. The method as set forth in claim 1, wherein said estimated load calculating step comprises the steps of: calculating a plurality of current loads one for each of said computers at predetermined time periods; and calculating said estimated loads by Le←a·Lp+(1−a)·Le where Le is an estimated load, Lp is a current load, and “a” is a definite value (0<a≦1).

4. The method as set forth in claim 1, wherein each of said estimated loads comprises a first estimated load relating to the number of transactions currently assigned to one of said computers and a second estimated load relating to the number of transactions currently being executed in the processors of said one of said computers.

5. The method as set forth in claim 1, wherein said estimated process time expansion ratio calculating step comprises a step of correcting each of said estimated loads by the number of transactions currently assigned to said computers, so that said estimated process time expansion ratios are calculated in accordance with said corrected estimated load and the numbers of processors in said computers.

6. The method as set forth in claim 1, wherein said load indexes are said estimated process time expansion ratios.

7. The method as set forth in claim 1, wherein said load indexes are said estimated process time expansion ratios per processor in said computers.

8. The method as set forth in claim 1, wherein said load index calculating step calculates said load indexes before distributing said transaction to said computers.

9. The method as set forth in claim 1, wherein said load indexes calculating step calculates said load indexes before distributing said transaction to said computers.

10. The method as set forth in claim 1, wherein said load index calculating step calculates said load indexes each being a load index per processor in one of said computers after distributing said transaction to said computers.

11. The method as set forth in claim 1, wherein said load index calculating step calculates said load indexes each being a load index per processor in one of said computers before distributing said transaction to said computers.

12. The method as set forth in claim 1, wherein said load index calculating step comprises the steps of: calculating a first load index per processor in one of said computers before distributing said transaction to said computers; calculating a second load index per processor in one of said computers after distributing said transaction to said computers; and calculating a difference between said second and first load indexes as one of said load indexes.

13. The method as set forth in claim 1, wherein said transaction distributing step comprises the steps of: selecting one of said computers having the minimum load index; and causing said transaction to be processed by said one of said computers.

14. The method as set forth in claim 1, wherein said transaction distributing step comprises the steps of: determining whether or not one of said load indexes is smaller than a threshold value; and causing said transaction to be processed by one of said computers having said one of said load indexes, when said one of said load indexes is smaller than said threshold value.

15. The method as set forth in claim 14, wherein said threshold value is dependent upon the minimum one of said load indexes.

16. A load-distributed computer system comprising: a plurality of computers each including processors; a relay distributing unit for receiving a message for requesting a transaction process; and a channel communication network connected between said computers and said relay distributing unit, each of said computers comprising: a first load data measuring unit for measuring a load of said each of said computers; a transaction processing unit for processing transactions, said relay distributing unit comprising: a second load data measuring unit for receiving said load to calculate an estimated load in accordance with said load; a load data storing unit for storing said estimated load data; an executing computer selecting unit for calculating a plurality of estimated process time expansion ratios one for each of said computers, in accordance with said stored estimated load and the number of processors in each of said computers, calculating a plurality of load indexes, one for each of said computers, in accordance with said estimated process time expansion ratios, selecting one of said computers having the minimum load index in accordance with said load indexes, and causing said transaction process to be processed by said transaction processing unit of said selected computer.

17. The system as set forth in claim 16, wherein said second load data measuring unit calculates said estimated loads at predetermined time periods.

18. The system as set forth in claim 16, wherein said second load data measuring unit comprises: means for calculating a plurality of current loads one for each of said computers at predetermined time periods; and means for calculating said estimated loads by Le←a·Lp+(1−a)·Le where Le is an estimated load, Lp is a current load, and “a” is a definite value (0<a≦1).

19. The system as set forth in claim 16, wherein each of said estimated loads comprises a first estimated load relating to the number of transactions currently assigned to one of said computers and a second estimated load relating to the number of transactions currently being executed in the processors of said one of said computers.

20. The system as set forth in claim 16, wherein said executing computer selecting unit comprises means for correcting each of said estimated loads by the number of transactions currently assigned to said computers, so that said estimated process time expansion ratios are calculated in accordance with said corrected estimated load and the numbers of processors in said computers.

21. The system as set forth in claim 16, wherein said load indexes are said estimated process time expansion ratios.

22. The system as set forth in claim 16, wherein said load indexes are said estimated process time expansion ratios per processor in said computers.

23. The system as set forth in claim 16, wherein said executing computer selecting unit calculates said load indexes before distributing said transaction to said computers.

24. The system as set forth in claim 16, wherein said executing computer selecting unit calculates said load indexes before distributing said transaction to said computers.

25. The system as set forth in claim 16, wherein said executing computer selecting unit calculates said load indexes each being a load index per processor in one of said computers after distributing said transaction to said computers.

26. The system as set forth in claim 16, wherein said executing computer selecting unit calculates said load indexes each being a load index per processor in one of said computers before distributing said transaction to said computers.

27. The system as set forth in claim 16, wherein said executing computer selecting unit comprises; means for calculating a first load index per processor in one of said computers before distributing said transaction to said computers; means for calculating a second load index per processor in one of said computers after distributing said transaction to said computers; and means for calculating a difference between said second and first load indexes as one of said load indexes.

28. The system as set forth in claim 16, wherein said executing computer selecting unit comprises: means for selecting one of said computers having the minimum load index; and means for causing said transaction to be processed by said transaction processing unit of said one of said computers.

29. A load-distributed computer system comprising: a plurality of computers each including processors each for receiving a message for requesting a transaction process; and an exchange/store unit connected to said computers, each of said computers comprising: a first load data measuring unit for measuring a load of said each of said computers; a transaction processing unit for processing transactions, a second load data measuring unit for receiving said load to calculate an estimated load in accordance with said load; a load data storing unit for storing said estimated load data; an executing computer selecting unit for calculating a plurality of estimated process time expansion ratios one for each of said computers, in accordance with said stored estimated load and the number of processors in each of said computers, calculating a plurality of load indexes, one for each of said computers, in accordance with said estimated process time expansion ratios, determining whether or not said each of said computers is to process said transaction process in accordance with said load indexes, causing said transaction process to be processed by said transaction processing unit when said each of said computers is to process said transaction process, and transmitting said message via said exchange/store unit to another computer when said each of said computer is not to process said transaction process.

30. The system as set forth in claim 29, wherein said second load data measuring unit calculates said estimated loads at predetermined time periods.

31. The system as set forth in claim 29, wherein said second load data measuring unit comprises: means for calculating a plurality of current loads one for each of said computers at predetermined time periods; and means for calculating said estimated loads by Le←a·Lp+(1−a)·Le where Le is an estimated load, Lp is a current load, and “a” is a definite value (0<a≦1).

32. The system as set forth in claim 29, wherein each of said estimated loads comprises a first estimated load relating to the number of transactions currently assigned to one of said computers and a second estimated load relating to the number of transactions currently being executed in the processors of said one of said computers.

33. The system as set forth in claim 29, wherein said executing computer selecting unit comprises means for correcting each of said estimated loads by the number of transactions currently assigned to said computers, so that said estimated process time expansion ratios are calculated in accordance with said corrected estimated load and the numbers of processors in said computers.

34. The system as set forth in claim 29, wherein said load indexes are said estimated process time expansion ratios.

35. The system as set forth in claim 29, wherein said load indexes are said estimated process time expansion ratios per processor in said computers.

36. The system as set forth in claim 29, wherein said executing computer selecting unit calculates said load indexes before distributing said transaction to said computers.

37. The system as set forth in claim 29, wherein said executing computer selecting unit calculates said load indexes before distributing said transaction to said computers.

38. The system as set forth in claim 29, wherein said executing computer selecting unit calculates said load indexes each being a load index per processor in one of said computers after distributing said transaction to said computers.

39. The system as set forth in claim 29, wherein said executing computer selecting unit calculates said load indexes each being a load index per processor in one of said computers before distributing said transaction to said computers.

40. The system as set forth in claim 29, wherein said executing computer selecting unit comprises; means for calculating a first load index per processor in one of said computers before distributing said transaction to said computers; means for calculating a second load index per processor in one of said computers after distributing said transaction to said computers; and means for calculating a difference between said second and first load indexes as one of said load indexes.

41. The system as set forth in claim 29, wherein said executing computer selecting unit comprises means for determining whether or not one of said load indexes is smaller than a threshold value, thus determining whether or not said each of said computers is to process said transaction process.

42. The system as set forth in claim 41, wherein said threshold value is dependent upon the minimum one of said load indexes.

43. The system as set forth in claim 29, further comprising a relay provisional-distributing unit, connected to said computers, for statically or quasi-statically distributing said message to said computers.

44. The system as set forth in claim 43, wherein said relay provisional-distributing unit distributes said message to said computers by dividing a plurality of terminal units connected to said relay provisional-distributing unit into a plurality of groups each corresponding to one of said computers.

45. The system as set forth in claim 43, wherein said relay provisional-distributing unit distributes said message to said computers by sequentially said message to said computers in accordance with the numbers of processors therein.

46. The system as set forth in claim 43, wherein said relay provisional-distributing unit stochastically distributes said message to said computers.