JPH03278136A - Method and device for controlling load of computer system - Google Patents

Abstract

PURPOSE:To dynamically control the load of a computer system by observing the length of the queue of a processing requirement received with the computer system at every classification, and determining the number of processing requirements to be processed at every classification based on the result of its observation. CONSTITUTION:This device is provided with a computer 101 for receiving a processing requirement from a terminal, and a computer 102 for executing the processing of the processing requirement, as a hardware constitution. In such a state, the length of the queue of the processing requirement by various classifications is observed, and based on this observed length of the queue of the processing requirement, the processing frequency of the processing requirement by various classifications is calcu lated, and based on this processing frequency, the processing is executed. Accordingly, by such a processing corresponding to the processing frequency, the time when the processing requirement of an arbitrary classification is inputted, and thereafter, output is smoothed. In such a manner, even when the input of the processing requirement of a specific classification is increased, it does not occur that the processing of the processing requirement of its classification is delayed within a range which does not exceed the processing capacity as the whole system.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a load control method and apparatus for a computer system.

(11) [Prior Art] Conventionally, in a computer system, data of various types (for example, by company or by deposit) are randomly sent from terminal devices or other computer systems at various locations.

A technology called queuing control is used to process data processing requests in an orderly manner (for example, online/l-transaction types such as balance verification, etc.). One such technique is to set priorities for processing services depending on the type of individual processing requests, and proceed with processing based on these priorities. Such technology is described, for example, in "Online System" written by Yasuo Ueda and published by Shokodo on June 20, 1986 (page 42 -
(page 56).

In addition, in communication networks such as packet switching networks, for example (
1) Kato et al., "System Configuration Technology, Iwahi Lecture Microelectronics 10", Iwanami, pp. 55-57 (April 1985), (2) JP-A-63-42543 (Packet Flow ``Control method'', (3) JP-A-63-
As described in Publication No. 299550, "Circuit Switching (12) Communication Speed Control Method Using Congestion Control in Networks," communication data within the communication network is A method is used to adjust the amount of communication data input to a communication network depending on how busy the communication network is.

[Problem to be solved by the invention]

The above-mentioned conventional technology does not take into consideration the case where the load suddenly changes. Therefore, when processing requests increase extremely, program bugs are less likely to become apparent due to processing delays, untested operating conditions, or unusual memory usage. There was a problem in that logical flaws in the program could surface and cause inconvenience (), which could lead to system failure.

The above-mentioned increase in processing requests occurs because processing requests are accepted regardless of the amount of unprocessed processing requests remaining within the computer system. One reason for this is that the input speed of processing requests (the number of inputs per unit time) exceeds the maximum processing speed of the computer system (the number of processing requests per unit time (13)). Another cause is when the processing time of an already accepted processing request far exceeds the scheduled time. Due to such causes,
The computer system falls into an overload state (a state in which many unprocessed requests remain in the computer system). In an overloaded state, processing requests have to wait for a long time within the computer system, resulting in an extremely long response time (the time from when a processing request is input until the processing results are returned). . In addition, in an overloaded state, many tasks within a computer system are usually executed in parallel, which increases the probability of timing errors occurring, making it more likely that the system will go down.

On the other hand, the conventional method of adjusting the amount of communication data input to a communication network is aimed at preventing an increase in communication delay time in the communication network and a decrease in communication throughput, and is aimed at preventing overload conditions of a computer system. is not taken into account. As a result, the throughput of the communication network has improved,
(14) The arrival speed of processing requests to a particular computer system connected to this communication network, that is, the input speed of processing requests to the computer system, increases in a direction that encourages the computer system to fall into an overload state. There was a problem that it could work.

An object of the present invention is to provide a method and apparatus for dynamically controlling the load of a computer system.

Another object of the present invention is to solve the problem in a computer system when load imbalance occurs between types of processing requests during system operation, or when the load imbalance changes.
The service is designed to smooth out the time from when a given processing request is input until it is output (hereinafter referred to as "processing time") between each type (to minimize the difference in processing time between types). An object of the present invention is to provide a load control method and device for a computer system that dynamically changes priorities and prevents delays in processing.

Still another object of the present invention is to limit the number of processing requests that the computer system accepts even when processing requests exceed the maximum processing speed of the system, thereby preventing the computer system from becoming overloaded. An object of the present invention is to provide a load control method and device for a computer system that allows the computer system to continue operating without any problems.

Still another object of the present invention is to provide a method and apparatus for shortening the waiting time of processing requests when the waiting time becomes long.

Further objects of the present invention will become apparent from the specification and accompanying drawings.

[Means to solve the problem]

In order to achieve the above object, the present invention is as follows. The length of the queue of processing requests accepted by the computer system is observed for each type, and the number of processing requests to be processed is determined for each type based on the observation results. In this case, the arrival rate of processing requests accepted by the computer system (the number of processing requests that should be accepted by the computer system per unit time) may be used instead of the queue length.

The above object can also be achieved by estimating the processing capacity of the computer system and adjusting the number of processing requests to be accepted (16) so as not to exceed the processing capacity.

Furthermore, the above purpose is to reduce the load on an overloaded computer. The effect based on the above configuration will be explained.

According to the present invention, in a computer system that processes processing requests of a plurality of types, the length of a queue of processing requests for each type is observed, and based on the observed queue length of processing requests, various types of processing requests are processed. The processing frequency of another processing request (the number of processing requests to be processed for the time being) is calculated, and processing is performed based on this processing frequency. Therefore, by processing according to such processing frequency, the time from input to output of any type of processing request (i.e. processing time)
is smoothed. In other words, by processing many processing requests of heavy load types at once based on the processing frequency, the number of types that wait for a long time to be processed decreases, and the processing time becomes uniform among the types. (17) As a result, even if the number of input processing requests of a particular type increases, the processing of the processing requests of the particular type will not be delayed to the extent that the processing capacity of the entire system is not exceeded.

Further, according to the present invention, the number of processing requests accepted by the computer system is adjusted so as not to exceed the processing capacity of the computer system, so the risk of falling into an overload state is reduced.

Further, according to the present invention, in an overloaded computer of a computer system, processing requests waiting to be processed are transferred to other computers and processed in parallel, so processing delays are reduced and response time can be shortened. .

〔Example〕

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 2 shows the hardware configuration of the computer system.

This system consists of the following:

103 is an input terminal for inputting a processing request.

101 receives processing requests from the terminal 103 (18
) It is a calculator. Reference numeral 102 denotes a computer that receives a processing request received by the computer 101 and processes the processing request according to an instruction from the data processing instruction terminal 104. Reference numeral 104 is a data processing instruction terminal used by a specialized operator to input instructions for execution of requested processing. 105 is calculator 1
This is a system control terminal that instructs the operating status of 02 and the degeneration (stopping of its function) of load smoothing control, which will be described later.

A processing request file 106 records processing requests in chronological order. 108 is a processing data file in which processing requests are recorded in a converted form. A processing result file 107 records the results of processing the processing data file according to instructions from the data processing instruction terminal 104.

109 to 117 indicate the connection status of each of these parts.

First, an outline of the processing flow will be described.

Processing requests input from the terminal 103 are sent to the computer 101 and recorded in a processing request file 106 in chronological order. The computer 102 reads the processing request from the processing request file 106, converts it into another format (19), and writes it to the processing data file 108.

Further, the computer 102 has a data processing instruction terminal 104.
The data in the processed data file 108 is processed according to instructions from the processing data file 108, and the processing results are recorded in the processed result file 107. The computer 101 reads the data in the processing result file 107 and outputs it to the terminal 103.

FIG. 3 shows the computer 101 and computer 1 shown in FIG.
Shows programs, tables, etc. in 02. The processing operation of a processing request will be explained in detail with reference to FIG. Note that the solid line connecting the program and the table represents the flow of data, and the solid line connecting the program and the terminal represents the flow of data between the program and the terminal.

A processing request input from the terminal 103 is sent to the computer 101. Processing request reception program 20 of computer 101
1 writes the processing requests to the processing request file 106 in the order received. The type-specific processing request queue 205 of the computer 102 is an area in which multiple types of processing requests are stored by type. The processing request queues for each type are F I F O (First 1n
First 0ut) is a queue managed by
Each of these is logically independent. Also, by type,
It is managed with a type number that corresponds one-to-one with the type, and 1
The number is from to the number of types. The processing request input and load observation program 204 of the computer 102 reads the processing requests written in the processing request file 106 in the order in which they were written, writes them to the processing request queue of the relevant type in the type-specific processing request queue 205, and performs processing request management. Update information regarding processing requests in table 206. The load control and processing request conversion program 208 checks whether there is a processing request in each processing request queue of the type processing request queue 205. As a result, if there is a queue to which one or more processing requests are connected (searching for a queue to which processing requests are connected is called a "processing request queue search"), the load calculation program 209 is started. The load calculation program 209 calculates the number (21) of processing requests of the relevant type (hereinafter referred to as "cycle number") based on the information in the processing request management table 206, and stores the result in the load management table 214. Set to . The load control and processing request conversion program 208 converts the processing request based on the information in the load management table 214 and converts the processing request into the processing data file 108.
, and the value of the number of cycles set in the load management table 214 is decreased by 1. As a result, if the number of cycles is not O, the processing request queue of the type-based processing request queue 205 is not searched, and the next processing request of the type is processed. If the number of cycles is O, the processing request queues in the type-specific processing request queue 205 are searched again, and if there is a queue to which processing requests are connected, the processing requests of the type are converted and the processing data file is Write to 108. The processed data display program 211 periodically displays the contents of the processed data file 108 on the data processing instruction terminal 104. The data processing program 210 processes the information in the processing data file 108 according to instructions from the data processing instruction terminal 104, writes the processing result to the processing result file 107, and sends the processing result to the data processing instruction terminal 104. Send the results. The processing result output program 203 reads the processing results from the processing result file 107 and outputs them to the terminal 103. The write pointer sending program 202 periodically sends to the computer 102 a write pointer indicating the writing position of the most recently written processing request to the processing request file 106. The write pointer writing program 207 receives the sent pointer and stores it in the processing request management table 207.
Write to 6. The system control program 213 manages the execution of programs within the computer 102. When the processing request input and load observation program 204 and the load control and processing request conversion program 208 request suspension of processing or activation after a certain period of time, the programs are suspended or activated in accordance with this request. (The relationship between the system control program 213 and other programs is explained in the third section.
It is omitted in the figure. ) Command processing program 21
2 updates the value of the load control degeneration flag area 5o9 (described later) on the processing request management table 206 in response to an instruction from the system control terminal 105 (23).

Next, parts related to load smoothing control, which is a feature of this embodiment, will be explained in detail with reference to FIG. FIG. 1 is a flowchart showing load smoothing control. In this method, after inputting a processing request (step 1), the length of the queue for each type of processing request is observed (step 2).
Calculate the number of processing requests to be processed (processing frequency) for each type based on the observed queue length of processing requests (Step 3)
, and processes the processing request based on this (step 4).
), further processes (-other processes such as boat fishing, for example post-processing) are added to the above processing results (step 5), and the processing results are output (step 6).

Below, we will first explain the details of the processing request and the structure of tables, etc., and then explain step 1. The detailed explanation of step 2 will be explained using the flowcharts of FIGS. 10(a), (b), and (c), and the detailed operation of step 3 will be explained in FIG.
The flowchart of a) will be explained using (24).

FIG. 4 is a diagram showing detailed contents of a processing request written to the processing request file 106. Processing request file 1
The processing request written in 06 consists of a processing request reception serial number 70 given by the processing request reception program 201 to the processing request input from the terminal 103, and processing request data 7 input from the terminal 103. Further, the processing request data 7 includes a type number 71 and load control and processing request conversion program information 72. The processing request reception serial number 70 is a serial number indicating the order in which processing requests are received by the processing request reception program 201. Type number 71
is a number representing the type of processing request, and corresponds to each type in order starting from 1.

The load control and processing request conversion program information 72 is information that is converted by the load control and processing request conversion program 208, and is information that is not particularly related to load smoothing control.

FIG. 5(a) is a diagram showing details of the type-specific processing request queue 205. Type-specific processing request queue (25) 205 has 1m queue element areas 301i (i
=a-m). FIG. 5(b) is a diagram showing details of the first queue element area of the type-specific processing request queue 205. Cue element (i) area 301
1 consists of a processing request area 310 and a next queue element 1 to area 311 (a pointer to the queue element 1 to be connected next). In the processing request area 310,
As described above, the processing requests stored in the processing request file 106 are stored by type. In the NEXI queue element area 311, there are processing request queues 20 by type.
The position of the next queue element of the same type of processing request queue within 5 is stored. If there is no next queue element, O is set. Type-specific processing request queue 20
Queuing of processing requests to each type of processing request queue in 5 is performed as follows. Using the value of the empty area head pointer area 504 on the processing request management table 206 (described later), secure an area (referred to as "empty area (26) rear") in which no processing requests are stored from the queue element area 301. and set a processing request to this,
This is done by connecting this to the end of the processing request queue for the type of processing request to be queued. Details of the above queue element management method will be described in the explanation of FIG.

FIG. 6 is a diagram showing details of the processing request management table 206. The processing request management table 206 includes a common information area 401 and n entry information areas 402x (
Entry information area 402X
is an area for storing management information of processing requests of type X for each type.

In addition, the common information area 401 stores management information of processing requests regardless of type.

FIG. 7 is a diagram showing details of the common information area 401 in FIG. 6. The common information area 401 includes a total number of types area 501, a total number of areas area 502, a number of empty areas area 503, an empty area first pointer area 504, an empty area last pointer area 505, a WORK pointer area 506, and a write area. Pointer (WP) area 50
7, read (27) pointer (RP) area 508, load control degeneration, festival.

It consists of a flag area 509. Poison type number area 501
The total number (n) of processing request types is stored in . The total number of areas area 502 stores the total number (m) of queue element areas 301 of the type-specific processing queue 205. The number of empty areas area 503 stores the number of empty queue element areas 301i of the type-specific processing request queue 205. The empty area head pointer area 504 stores the position of the head area of the area (empty area) in which no chained processing requests are stored in the type-specific processing request queue 205.

The empty area final pointer area 505 stores the position of the last area of the empty areas chained in the type-specific processing request queue 205.

The WORK pointer area 506 stores the position of the queue element area 301 that should be temporarily saved in order to create a queue of processing requests in the type-specific processing request queue 205. The WP area 507 stores the processing request reception serial number of the processing request written most recently (28) to the processing request file 106. The RP area 508 stores the processing request reception serial number 70 of the processing request last read out from the processing request file 106. In the load control degeneration flag area 509,
A flag indicating whether or not to degenerate load control (ON when degenerate, ie, load control is not performed, and FF when non-degenerate, ie, load control is performed) is stored. Here, "degeneracy" refers to operating the entire system without only one of its many functions working. for example,"
"Degenerate load control" means that we stop dynamically proceeding with processing by observing the load shown in Figure 1, and instead handle each type of processing request in a predetermined fixed order, regardless of the load at each point in time. This refers to the processing of

FIG. 8 shows the processing request management table 206 shown in FIG.
4.02 is a diagram showing details of an arbitrary entry information area 4.02x. The entry information area 402x consists of a type X processing request number area 601, a type X processing request queue head pointer area 602, and a type X processing request queue end pointer area 603 (29). The number of type X processing requests area 601 stores the number of processing requests connected to the type X processing request queue in the type-specific processing request queue 205. The type X processing request queue head pointer area 602 stores the position of the head queue element 3011 of the type X processing request queue in the type-specific processing request queue 205. Type X processing request queue final pointer area 6
03, type X in the type-specific processing request queue 205
The last queue element 301i of the processing request queue of
The location of is stored.

FIG. 9 is a diagram showing details of the load management table 214. The load management table 214 has a type number area 801
, processing request processing number area 802, and type number WORK
It consists of area 803.

The type number area 801 stores the type number of the processing request processed immediately by the load control and processing request conversion program 208. The processing request processing number area 802 stores the (30) number of processing requests (ie, the number of cycles) that the load control and processing request conversion program 208 should process at the present time. Type number W
The ORK area 803 stores the type number of the processing request being processed by the load control and processing request conversion program 208.

Below, the operation of load smoothing control, which is a feature of this embodiment, will be explained in detail. Step 1 and Step 2 in Figure 1
is executed by the processing request input and load observation program 204. The processing request input and the operation of the load observation program 204 will be explained in detail using the flowchart of FIG. 10(a).

First, it is determined whether there is a processing request in the processing request file 106 based on whether the value of the WP area 507 is greater than the value of the RP area 508 (step 901).

If there is no processing request (that is, if the value of the WP area 507 = the value of the RP area 508), the system waits for activation after a certain period of time from the system control program 213 (step 9
02 system control unit 213 timer function). If there is a processing request (if the value of the WP area 507 > the value of the RP area 508), the type (31) Checks whether there is an empty area in the separate processing request queue 205 and the value of the empty area head pointer area 504 is 0. The judgment is made by checking whether or not (step 903).

If there is no empty area (empty area start pointer area 50
If the value of 4 is 0), the system waits for startup from the system control program 213 after a certain period of time (step 904). If there is an empty area, step 905 to secure an empty area (details will be described later) and subtract 1 from the value of the empty area number area 503 (step 906). After this, processing request file 10
Processing request to be read in 6 (value of RP area 508 +
1 (processing request with processing request reception number 7o) is set and written in the previously secured area) (step 907).
, connects the area to the end of the processing request queue of the type (step 908: details will be described later). Furthermore, the value in the processing request count area 601 of the relevant type is increased by 1 (step 9o9), and the value in the RP area 508 is increased by 1 (step 91o).

After this, the process returns to step 901. Through the above processing, one of the processing requests is queued in the processing request queue of the relevant type in the type-based processing request queue 205 (32).

Next, the detailed processing of step 905 is shown in FIG. 10(b).
) will be explained according to the flowchart shown in FIG. First, W.O.
The value of the empty area head pointer area 504 is set in the RK pointer area 506 (step 9051). Next, the queue element area 3 pointed to by the value of the WORK pointer area 506 is added to the empty area head pointer area 504.
The value of the next queue element area 311 of 01i is set (step 9052). Furthermore, (IIO
Next queue element area 31 of queue element area 3011 pointed to by the value of RK pointer area 506
O is set to 1 (step 9053).

Through the above processing, an empty area is secured in the queue element area 301i where no processing requests are stored.

Next, the detailed processing of step 908 is shown in FIG.
) will be explained according to the flowchart.

First, the value (33
) is O (step 9081). If this value is O, set the value of the WORK pointer area 506 in the head pointer area 602 of the processing request queue of the relevant type (Step 9082), and set the value of the WORK pointer area 506 in the last pointer area 603 of the processing request queue of the relevant type. Set the value (step 9083). On the other hand, the value in the processing request count area 601 of the relevant type is 0.
If not (step 9081), the WORK pointer area 5 is placed in the next queue element area 311 of the queue element 301 of the type-specific processing request queue 205 indicated by the value of the processing request queue final pointer 603 of the type.
06 (step 9084), and the value of the WORK pointer area 506 is set in the processing request queue final pointer area 603 of the relevant type (step 908).
5). Through the above processing, the secured area is connected to the end of the processing request queue of the type of processing request to be queued.

The operation of step 4 in FIG. 1 is executed by the load control and processing request conversion program 208 (34). The operation of the load control and processing request conversion program 208 will be described with reference to the flowchart shown in FIG. 11(a) (the entirety of which corresponds to step 4). First, an initial value 1 is set in the type number area 801 of the load management table 214, and processing is started from the type of processing request whose type number is "1" (step 911). Next, it is determined whether or not load control is to be performed (step 912).

When the load control is performed, that is, the value of the load control degeneration flag area 509 is on (“1”, that is, load control is degenerated), or the value of the load control degeneration flag area 509 is off ( rob) and the value of the cycle number area 802 is O (the number of processing requests to be processed for the type is O), the value of the type number area 801 is set in the type number WORK area 803 (step 913). When you are not in a state where load control is performed, that is, when you are in a state other than the above,
The process of step 923 is performed. After processing step 913, type number WORK area 803
Increment the value by 1 (step 914), type number area 8
It is determined whether the value of 01 is larger than the value of the total number of types area 501 (step 915).

If it is larger, the type number WORK area 803 is set to 1 in order to perform the process again from the process type number 1 (step 917), and the process of step 916 is performed. If the value in the type number area 801 is less than or equal to the value in the total number of types area 501, the type number area 8 indicates whether all types have been searched.
It is determined whether the value of 03 and the value of the type number WORK area 801 are equal (step 916). If these values are equal, after increasing the value in the type number area 801 by 1# (step 918), the system control program 21
Wait for activation after a certain period of time (step 919). After this, when started from the system control program 213,
Processing starts from step 912. If you have not performed a search that goes through all the types (if the type in the type number area 801 is not equal to the value (36) in the type number WORK area 803), check whether the value in the number of processing requests area 601 for the type is O. If it is determined to be 0 (step 920), the process of step 914 is performed. If it is not 0, determine whether the value of the load control degeneracy flag area 509 is 0 (
step 921). If the value of the load degeneracy flag 509 is 1, the process of step 923 is performed, and if it is 0, load calculation is performed (step 2 of step 9). After that, the next processing request is read from the type-specific processing request queue 205, converted, and written to the processing data file 108 (step 92
3).

Next, step &) determines whether the value in the load control degeneracy flag area 509 is 0, and if it is 1, the cycle number area 8
02 is set to O to stop load control (step 926
), the process of step 927 is performed. If the value in the load control degeneration flag area 509 is 0, the cycle number area 802
Subtract 1 from the value of (step 925), type number area 80
Set the value of type number WORK area 803 to 1 (
step 927).

(37) Through the above processing, one processing request is extracted from the type-specific processing request queue, processed, and written to the processing data file 107.

Step 3 in FIG. 1 is executed by the load calculation program 209. Operation of load calculation program (step 3)
Or Stellarappa a2゜, ) &, 1 in Fig. 11(a)
11?1(b)E□Row chart.

First, a value Ki determined by the following equation (1) is set (stored) in the cycle number area 802 (step 31
).

L+1 Here, L+: Value of the processing request count area 601 of the relevant type (length of queue for processing requests of type i): Average value of all types of Ll, obtained as follows. , (38) Total number of areas (value of 502) Number of empty areas (value of 503) Total number of types (value of 501) ... (1a) K, processing frequency of processing requests of type 2 (program 208)
is the number of requests to be processed), round down the fractions below the decimal point, and then judge whether the number of cycles (value 802) is O (step 32); if it is not 0, the load calculation process is completed. 0&ba, set 1 in the cycle number area 802 (
Step 33), that is, if 1 is not O in the processing frequency obtained as described above, that value is stored in area 802, and when Kl is O, 1 is set to 1, and
Store in.

Note that when the command processing program 212 is started by inputting a command from the system control terminal 105, the command processing program 212 sets the value of the load control degeneration flag area 509.

(39) According to the above embodiment, the number of processing requests for each type in the processing request queue 205 for each type (type x number of processing requests area 6
01 value LL) is observed, the processing frequency Ki is determined by the above equation (1), and processing requests are processed based on this processing frequency, so the processing time is smoothed between each type.
Processing delays are less likely to occur.

In the above embodiment, a load display program is provided in the computer 102 shown in FIG. 105
The operator issues a command based on this display,
It is also possible to replace the above expression (1) with an expression that gives a different processing frequency. For example, when a large number of processing requests of a specific type are input, the following equation (2) can be used in place of the above equation (1).

I Kl =□ ・・・(
2) Here, the meanings of K+, Lt and L are as described above (40
) It is the same as in the case of equation (1). Also, as in the case of equation (1), the value of + in the processing frequency is rounded down to the decimal point, and K
, becomes O, replace it with 1. By doing this, even if there is a particularly large number of types of processing requests and input requests, this type of processing request will be processed quickly, and there will be no processing delay compared to when formula (1) is used. It will no longer occur.

Note that an appropriate formula may be selected from a plurality of formulas for determining the processing frequency such as formulas (1) and (2) above at the time of generation of the computer system.

In the above embodiment, the processing frequency is determined from the queue length of each type of processing request, as in equation (1) or (2). speed(
The number of inputs per type per unit time) or processing time may be observed, and the processing frequency may be determined from the reception speed or processing time. The details will be described later.

Note that in the above embodiment, the hardware configuration includes the second
As shown in the figure, there is a computer 101 that accepts processing requests from a terminal (41), and a computer 1 that processes the processing requests.
02, but instead of this, the configuration may be as shown in FIG. 12 or 13. In FIG. 12, the computer 101 and the computer 102 in FIG.
9. Also, in Fig. 13, the second
A plurality of computers are provided to perform the function of the computer 102 shown in the figure, and these computers are connected to the computer 101. Here, in addition to the computer 102, a computer 111 having the same functions as the computer 102 is connected to the computer 101. 114 is a data processing instruction terminal, 115 is a system terminal, 116 is a processing request file, 117 is a processing result file, and 118 is a processing data file.

In FIG. 13, each of the computers 102 and 111 can perform load control according to the present invention by checking the load state of the width calculator.

Next, an embodiment in which the processing frequency is calculated from the above-mentioned reception speed (hereinafter referred to as arrival speed) will be described, focusing on the differences from the above-mentioned embodiment.

(42) Figure 14 shows arbitrary entry information in this embodiment of the processing request management table (206) in Figure 6 when the arrival rate of processing requests is used instead of the queue length to calculate the processing frequency. It is a detailed diagram of area 402xx. The entry information area 402xx includes a type-X processing request queue head pointer area 602, a type-X processing request queue last pointer area 603. Arrival time ta Miniria 604 Arrival time t
J-1 area 606. It consists of an arrival time tJ-2 area 606. Type X processing request queue head pointer area 60
2 contains the queue element h (301i) at the head of the processing request queue of type X in the processing request queue 205 by type.
The location of is stored.

In the type X processing request queue final pointer area 603,
The position of the last queue element (301i) of the type X processing request queue in the type-specific processing request queue 205 is stored. t, in the area 604, the processing request that arrived (stored) in the type-specific processing request queue 205 at the closest point in time from the current time (the time when this area was read) is stored at the arrival time (43) (actually this time). This is the time when the area is written 1lrite, and there is some error) is stored. The arrival time of the processing request that arrived at the second closest point to the current time is stored in t,−,605. The tJ-2 area 606 stores the arrival time of the processing request that arrived at the third closest point to the current time.

FIG. 15 is a diagram showing details of the load management table 214 when the arrival rate of processing requests is used to calculate the processing frequency without using the queue length. The load management table 219 includes a type number-rea 801° processing request processing number area 8022 and a type number WORK area 803. The type number area 801 stores the type number of the processing request processed immediately by the load control/processing request conversion program 208. The processing request processing number area 802 stores the number of processing requests (hereinafter referred to as the cycle number) that the load control and processing request conversion program 208 should process at the present time. The type number WORK area 803 stores the type number of the processing request that the load control and processing request conversion program 208 is currently processing. Required (44) The request arrival type number area 804 shows the processing request arrival rate (
The number of types whose storage rate (rate) is not O is stored in the type-specific processing request queue 205. The arrival rate area 805 stores the arrival rate of the type of processing request that the load control/processing request conversion program 208 is attempting to process.

The average arrival rate area 806 stores the average processing request arrival rate for types of processing requests whose arrival rate is not O.

The CPU usage rate area 807 stores the total CPU usage rate of the load control 9 process request conversion program 208 and the load calculation program 209.

The load calculation work area 808 is a work area for load calculation, and stores a type number.

The procedure for determining the processing frequency from the reception speed (arrival speed) will be described below.

First, the processing request input and load observation program 204 uses the information in the entry information area (402xx) in FIG. 14 to execute step 9 in the flowchart in FIG.
Instead of the process in step 09, the process in the flowchart shown in FIG. 16 is performed. In other words, (45) Type X management information section 402 of processing request management table 206
The value of tj-1 area 605 of xx is tA, 2 Area 6
Step 1401), t.

The value of area 604 is set to ta- in area 605 (step 1402). Next, set the current time to t.

It is set in area 604 (step 1403).

Also, the load calculation program 209 uses the information in the entry information area 402xx in FIG. 14 and the information in the load management table 214 in FIG. 15, and uses the information in the entry information area 402xx in FIG. Process the flowchart. That is, if the value in the tA-2 area 606 is O (step 1501), 1 is set in the cycle number area 802 (step 1515), and the load calculation is completed. If, t
If the value of the j-2 area 606 is not 0 (step 15
0 ], ), the arrival speed area 805 has the type number WO
Formula (46) consisting of the value of the entry information area 402xx of the type indicated by the number (type number) stored in the RK area 803 11 Value of tJ-1 area ta-Value of area-ta-2
Step 1502: Set the value found from the area value.
), and sets 1 (type number) in the load division work area 808 (step 1503). Next, it is determined whether the value of the ta-2 area 606 is 0 (step 1504).
, if it is not 0, the value of the request arrival type number area 804 is set to 1.
increase step 1505), average arrival speed area 80
Entry information area 4 of the type indicated by the number (type number) stored in the load calculation work area 808 in the value of 6
After setting the value found from the value of the formula tj area consisting of the value of 02xx (step 1506), the value of the load calculation work area 808 is increased by 1 (step 1507). If the value of the tJ-2 area 606 is 0 (step 1504.), the above step (47) knob 1505. Step 1506 is skipped and step 1507 is performed. Next, compare the value of the load calculation work area 808 and the value of the total number of types area 501,
If the former is less than or equal to the latter (step 1509), again,
Processing of determining whether the value of the t3-2 area 606 is O (
Return to step 1504) and load calculation work area 8
If the value of 08 exceeds the value of the total number of types area 501 (
Step 1509) Divide the value in the average arrival rate area 806 by the value in the requested arrival type number area 804 and set it in the average arrival rate area 806.Step 1510) From the system control program 213, select the CPU usage rate area 80.
After obtaining the total CPU utilization rate of the load control and processing request conversion program 208 and the load calculation program 209 in step 7 (step 1511), enter the formula (processing request processing speed) x (CPU utilization rate area) in the cycle number area 802. value)(
(Processing request processing speed) (value of average arrival speed area)...(3) Set the value found from (step 1512).

However, if the value of equation (3) is a non-integer, the decimal part is cut off and set in the cycle number area 802. Further, the processing request processing speed is a constant determined from the average instruction execution time of the computer and the number of steps of the program. If the value in the cycle number area 802 is O or negative (step 1513), 1 is set in the cycle number area 802 (step 1514), and the load calculation ends. If the value in the cycle number area 802 is not O or negative, the load calculation ends as is. Through the above processing, the type-specific processing request queue (205) for each type of processing request is created.
The processing frequency can be determined from the arrival speed.

Note that the processing frequency described above may be calculated from the following equation instead of from equation (1).

(49) Processing frequency = Value of the number of processing requests area Value of the total number of areas area - Value of the number of empty areas area Value of the total number of types area (4) However, the right side is rounded down to the decimal point. Also, if the processing frequency becomes 0 or negative, replace it with 1.

The above formula is based on the idea that queue lengths larger than the average are processed by the queue length longer than the average. ―;I14 S, Sakae I, Ichigo, Peng Ichi J...3゜Next, it relates to a method and apparatus for controlling the load of a computer system in combination with the above embodiments or alone. An example will be explained. This embodiment prevents the computer system from becoming overloaded by adjusting (controlling) the number of processing requests it receives.

First, an outline of this embodiment will be described.

Based on the maximum number of processing requests V○ that can be processed per unit time by the computer system identified in advance (with V○ as the standard)
The number VI of processing requests to the computer system per unit (50) time is adjusted. When vO is high, VI is also adjusted to be high, but when vO is low, VI is also adjusted to be low.

Here, "identification" refers to determining the quantity of something based on the results of examining data on different species.

Specifically, after inputting excessive processing requests to the computer system in advance, the computer system calculates the reciprocal of the time interval T per unit time when the time interval at which processing of each processing request ends becomes almost constant. The maximum number of processing requests that can be processed in The number VI of processing requests input to the computer system per unit time is adjusted so as not to exceed the maximum number V○ of processing requests that the system can process per unit time. For example, processing requests may not be accepted if necessary.

(51) Once a processing request is received, the number of processing requests input per unit time to the computer system can be adjusted (reduced) by invalidating the input as necessary.

As a result, the number VI of input processing requests to the computer system per unit time is limited to be less than or equal to the maximum number of processing requests vO that the computer system can process per unit time. The inflow of excessive processing requests that exceed the processing speed is prevented, and the computer system is not overloaded.

Examples of the present invention will be specifically described below.

FIG. 18 shows the hardware configuration of the computer system.

This computer system uses a terminal 10101 (such as a card reader, paper tape reader, CRT, etc.) used for inputting processing requests.
(equivalent to the terminal 103 in the above-mentioned embodiment), these terminals 10
Terminal control device 10102.101 controls data input/output to and from computer 10103 and data communication with computer 10103. A computer 10103 that accepts a processing request input from the terminal 10101 and notifies the processing result (52) to the user terminal; a computer 10104 that executes processing of the received processing request; and a computer that controls and manages the computer 10103. control terminal 10
105. Backup data file 1 that saves data to be carried over to the next startup when the computer 10103 is stopped
0106. Request reception file 10107 for storing accepted processing requests. Processing boot file 101 that stores data necessary to execute requested processing.

A processing result file that stores processing results? We have 1010 recruitment.

Next, computer 10103. The configuration of programs and tables in computer 10104 is shown in FIG. Note that arrows between programs, between programs and tables, and between programs and devices indicate the flow of data.

The computer 10103 has a system control program 11031. Request reception program 11032゜Result notification program 11033. Load control program 11034
．． It has (53) load management tables 11035. The system control program 11031 is the computer 1
Controls the startup of the 0103 program, controls the timer, and outputs control messages. That is, the system control program 11031 starts each program in the computer 10103 according to instructions from the control terminal 10105. Also, when an event occurs in the computer 10103,
The other programs in the computer 10103 are notified of what event has occurred. Furthermore, in response to a time acquisition request from each program, each program is notified of the current time. Furthermore, the control terminal 101 sends requested control messages in response to control message output requests from each program.
Output to 05. The request reception program 11032 accepts processing requests input from the terminal 10101. The result notification program 11033 notifies the terminal 10101 of the processing results of the requested processing. The load control program 181034 identifies the maximum processing speed of the computer system shown in FIG. The reading cycle, that is, the reading time interval (in this embodiment, it varies somewhat depending on the operating status of the terminal control device 102, and there are few cases in which a strictly constant value is maintained) is changed.

The load management table 1-81035 stores data necessary for the operation of the load control program 1+1034.

The computer 10104 internally has a system control program 111041. , per-request queue creation program 1610
42. Queue table 1 by request 1043. Processing request selection program 11044. Data processing program 110
It has 4.5. The system control program 11041 is
Controls the startup of programs on the computer 10104. That is, the system control program 11041 starts each program in the computer 10104 when the computer 10104 is powered on. The request-by-request queue creation program 11042 classifies the received processing requests according to the processing content, and queues them in the request-by-request queue table 11043. The request-specific queue table 11043 is a table in which processing requests are queued in FIFO (55) (First In First Out) separately for each processing content of the processing request. The processing request selection program 11044 takes out the queued processing request and requests the data processing program 11045 to process it. The data processing program 11045 executes the requested processing.

Furthermore, FIG. 20 shows the configuration of programs and buffers within the terminal control device 10102. Note that arrows between programs, between programs and buffers, and between programs and devices indicate the flow of data.

The terminal control device 10102 internally has a system control program 11021. I10 control program 11022.
Communication control program 11023° Read cycle control program 11024. Read timing buffer 1102
5. Communication buffer 11026, I10 buffer 1102
Has 7.

The system control program 11021 is the terminal control device 1
Controls the startup of the program 0102 and controls the timer. That is, the system control program (56) ram 11021 starts each program in the terminal control device 10102 when the terminal control device 10102 is powered on. In addition, in response to a time acquisition request from each program, the current time is transmitted to each program. The engineering/○ control program 11022 controls input and output to and from the terminal 10101.

The communication control program 11023 communicates with the computer 10103. I10 buffer 11027 contains I10
Processing request data read from the terminal 10101 by the control program 1.1022 is stored.

The communication control program 11 is stored in the communication buffer 11026.
Processing result data received by 023 from computer 10103 is stored. Read timing buffer 11025
stores the value of the cycle at which processing requests are read from the terminal 10101. The read cycle control program 11024 is
Change the value of the read timing buffer 11025.

When one piece of processing request data is input, the terminal 10101 stores the input processing request data in an internal input buffer and inhibits further input.

(57) Furthermore, when the processing request data from the internal input buffer is read out, the terminal 10101 releases the inhibition of input. Note that the maximum flame is determined for the processing request data, and processing such as input, buffering, transmission, etc. is performed in units of this maximum flame.

Furthermore, details of the load management table 11035 will be explained.

In this embodiment, the maximum processing speed (the maximum number of processing requests that can be processed per unit time) of the computer system shown in FIG. 18 is identified based on the time interval at which the terminal 10101 is notified of the processing results of the requests. That is, when there are almost no processing requests accumulated in the request-specific queue 11043, the average value of the above-mentioned time interval is approximately proportional to the arrival interval of processing requests, and as the arrival interval of processing requests gradually decreases, many processing requests are accumulated in the request-specific queue 11043. When processing requests accumulate, the average value becomes saturated and does not become smaller than a certain value (hereinafter, this state is called an overload state, and the above time interval at this time is called a saturation notification time interval).
behave like that. Based on this behavior, the reciprocal of the saturation notification time interval is identified as the maximum processing speed of the system. The load management table 11035 includes data for realizing the above processing and the read timing buffer 1 of the terminal 10101.
The value for cell 1 is stored in 1025.

FIG. 21 shows details of the load management table 11035. The load management table 11035 includes the processing interval queue table 10041. It is composed of a processing interval management table 10042.

The processing interval queue table 1004.1 has n processing interval areas 10411. In each area 10411,
Stores the time interval for notifying the terminal 10101 of the most recent n processing results in the past. Note that the right end of the processing interval queue 10041 is cyclically managed as being logically connected to the left end.

The processing interval management table 10042 includes an arrival interval setting value area 10421. Saturation processing interval areas 104, 22,
Processing time area a10423. Processing time area b104
24. Processing completion number area 10425, average processing interval area 10426. It has the most (59) new processing interval pointer area 10427. The latest processing interval pointer area 104.27 stores the value of a pointer pointing to the position of the processing interval area 10411 where the latest notification time interval is stored.

Note that the pointer is the processing interval queue table 10041
processing interval area 10411 from left to right
I, tL 2 II,...17 n+1. The average processing interval area 10426 stores the average value of the n values stored in the processing interval queue table 10041. In the processing completion number area 10425,
Stores the number of notifications of processing results to the user terminal 10101. Processing time area a 104.23.

Processing time x, IJ 7 b l O424c, terminal 1
This is a work area for determining the time interval for notification of processing results to 0101. In addition, the saturation processing interval area 1042
2, the saturation notification time interval (the arrival interval of processing requests gradually becomes smaller and the request queue 1104 is calculated using the method described later).
3 stores the processing result (60) (average notification time interval) when the average value of the notification time interval of the processing result is saturated and no longer becomes smaller than a certain value due to the accumulation of many processing requests. . The arrival interval setting value area 10421 stores a value to be set in the read timing buffer 11025 of the terminal control device 10102 (the cycle to be read from the terminal 10101).

In this example, by adjusting the input speed of processing requests to the computer system, the computer system falls into an overload state, and the processing result is returned from the computer system after the processing request is input to the computer system. To prevent the response time from becoming extremely long. For this purpose, the saturation notification time interval is determined by operating the computer system, the upper limit value of the input speed is determined based on the determined saturation time interval, and the input speed to the computer system is adjusted so as not to exceed this.

The operation of this embodiment will be explained below.

The operation of this embodiment is shown in FIG. (,) in FIG. 22 is the system control program 1103 in the computer 10103.
1 operation, the same figure (b) is the load control program 11034
The operation of the system control program 11021 in the terminal control device 10102 and the computer 10104 is shown in FIG.
The operation of systems other than the system control program 11041 is shown. This embodiment starts its operation by turning on the power to each device of the computer system shown in FIG. 18.

When the power is turned on, the system control program 11021 of the terminal control device 10102 starts operating, and the I10 control program 11022. Communication control program 1102
3. Start the read cycle control program 11024.

In addition, the computer 10104 starts operating, and the queue creation program 11042 for each request and the processing request selection program 1
1044° Start the data processing program 11045.

When the power is turned on, the system control program 11031 of the computer 10103 starts operating and waits for an instruction input from the control terminal 10105 (step 100).
01).

Thereafter, when an instruction to start the system is input from the control terminal 10105 (step 10002) (62), the system control program 11031 executes system start processing. In other words, the request reception program 1103
2. Start the result notification program 11033゜load control program 11034 (Step 10003), again.
Waiting for instruction input. Note that the instructions to start the system include:
There are two types, a setting mode and a control mode, and the system control program 11031 starts the corresponding program in the instructed mode.

Furthermore, when an instruction to terminate the system is input from the control terminal 10105 (step 10002), the system control program 11031 executes system termination processing. That is, the system control program 11031 includes a request reception program 11032, a result notification program 110,
33. The load control program 11034 is stopped (step 10004), and the system waits for an instruction input again.

When a time acquisition request is made from another program in the computer 10103 (step l OOO2), the system control program 11031 notifies the requesting program (63) of the current time (step 10005).
), the system waits for an instruction input again.

When an event occurs in the computer 10103, depending on the type of event, the program waiting for that type of event is notified that the event I- is waiting for has occurred (Step 10006), and the program waits for an instruction input again. Become.

After the power is turned on and the operation starts, in this embodiment, the estimation operation (steps 10011 to 100) for estimating the maximum processing speed of the computer system shown in FIG.
13) Processing request processing operation (step 10016.100) of adjusting the input speed of the processing request to the computer system based on the estimated maximum processing speed and executing the processing of the processing request.
19). Here, the former estimated behavior is usually
It is executed separately from the execution of the latter process request process. For example, during maintenance of the computer system shown in Figure 18, or
Estimation operations are generally performed during testing and the like. Furthermore, if there is no change in the configuration, amount of resources, etc. of the computer system shown in FIG. 18, it is not necessary to perform the estimation operation (64) in FIG. 18 every time the computer system starts operating. Therefore, in order to be able to execute the estimation operation and the processing request processing operation in different time periods, in this embodiment, it is possible to choose to stop the operation of the computer system shown in FIG. 18 immediately after the estimation operation. (Step 10014
, 10015).

Further, by giving an instruction from the control terminal 10105, only the latter processing request processing operation can be selected and performed (step 10010).

When the load control program 11034 is started in the setting mode from the system control program 11031 (step 10010), the load control program 11034 sets the terminal control device 10102 so that the reading speed of processing requests becomes the maximum, and the maximum speed of the computer system shown in FIG. Instruct to start input for estimating processing speed (step 10011). That is, the load control program 11034
A message with a code for setting the read timing buffer 11025 and a set data value of 0'' is sent to the program 11025.Then, the system control program (65) requests the program 11031 to increase the maximum processing speed of the computer system shown in FIG. A message to start input for estimation is output to the control terminal 10105.

Note that the communication control program 1 of the terminal control device 10102
When the computer 10103 sends a message with a code for setting into the read timing buffer 11025, the 1023 stores the set data (
Here, IIOIT) is passed to the read cycle control program 11024. Read cycle control program 1102
4 is a timing buffer 11 for reading the passed data.
Set to 025.

When the message to start input for estimating the maximum processing speed of the computer system shown in FIG. 18 is output to the control terminal 10105, a processing request is input from the user terminal 10101 at the fastest possible speed. As a result, the maximum processing speed of the computer system shown in FIG. 18 is identified as described below (step 10012).

When a processing request is input from the user terminal 10101 (66), the computer system of FIG. 18 executes the processing operation of the processing request (step 10019). That is, when one piece of processing request data is input to the user terminal 10101, the input processing request data is stored in an internal input buffer, and subsequent input from the user terminal 10101 is inhibited. On the other hand, the I10 control program 11022 reads processing request data from the input buffer in the user terminal 10101 into the I10 buffer 1027 at the cycle of the value set in the read timing buffer 11025.

When the processing request data in the internal input buffer is read out, the user terminal 10101 releases the inhibition of input. Through the above processing, the read timing buffer 11
By changing the contents of 025, the input speed of processing requests from the user terminal 10101 can be adjusted. Now, the value of the read timing buffer 11025 is “0”, so I10
When the control program 11022 reads the value of the input buffer in the user terminal 10101, the control program 11022 immediately transfers the value to the user terminal 1.
Go read the processing request data (67) in the input buffer in 0101 again. Therefore, when processing request data is input from the user terminal 10101, it is immediately read into the terminal control device 10102. Processing request data is I1
0 buffer 11027, communication control unit 1
1023 immediately sends the read processing request data to computer 10103.

The processing request data sent from the terminal control device 10102 is sent to the request receiving program 1103 of the computer 10103.
Passed to 2. The request reception program 11032 sends the processing request data to the request reception file 1010 in the order received.
Write to 7. In addition, the request reception program 11032 stores new processing request data in the request reception file 10107.
The request queue creation program 11042 of the computer 10104 is informed that the request has been written to the request queue creation program 11042 of the computer 10104.

When receiving notification that new processing request data has been written to the request reception file 10107, the calculation l110104
The request-by-request queue creation program 11042 reads the processing request data newly written (68) to the request reception file 10107. Furthermore, the read data is further classified according to the processing content requested by the processing request, and is connected to the end of the queue in which processing request data of the same processing content is connected in the request-specific queue table 11043. The processing request selection program 11044 periodically searches the request-by-request queue table 11043, extracts one piece of processing request data from it, and passes it to the data processing program 11045. The data processing program 11045 accesses the data in the processing data file 10108 to execute processing corresponding to the passed processing request data,
The processing results are sequentially written into the processing result file 10109. Furthermore, when the data processing program 11045 finishes writing the processing result to the processing result file 10109, the data processing program 11045 writes new processing result data to the processing result file 10109.
109 is notified to the result notification program 11033 of the computer 10103.

Upon receiving notification that new processing result data has been written to the processing result file 10109, (69) the result notification program 11033 writes the processing result file 1.
The processing result data newly written in 0109 is read from this, and the read processing result data is transmitted to the terminal control device 10102. That is, the result notification program 11033 transmits a message having a processing result code and processing result data to the terminal control device 10102. Also,
A processing end event is generated when the transmission of processing result data is finished. When the processing end event occurs, the system control program 11031 of the computer 10103
The load control program 11034 is notified that a processing end event has occurred. Upon receiving notification of the occurrence of the processing end event, the load control program 11034 proceeds with the process of estimating the maximum processing speed of the computer system shown in FIG. 18, as will be described in detail later.

Communication control program 1102 of terminal control device 10102
3 stores the processing result data of the sent message in the communication buffer 11026 (70) when the message having the processing result code is sent from the computer 10103. Processing result data is in the communication buffer 11026
When stored in the I10 control program 11023,
Immediately output the processing result data of the communication buffer 11026 to the user terminal 10101 (step 10019 below)
details of).

While repeating the above processing from inputting a processing request from the user terminal 10101 to outputting a processing result to the user terminal 10101, the load control program 11034 as shown in FIG. Identify the maximum processing speed of the computer system (step 1001
2).

If the maximum processing speed of the computer system in FIG. 18 is not identified, the load control program 11034 uses the identified maximum processing speed to control the computer system from the user terminal 10101 in order to prevent the computer system in FIG. 18 from falling into an overload state. The upper limit value of the input speed of the processing request is determined, and the input timing buffer 11025 in the terminal control device 10101 is determined according to this upper limit value.
The value to be set is determined (step 10013). Specifically, the estimated maximum processing speed VO (detailed operation (7)
1) To explain the operation, the saturation processing interval area 1 of the load management table 11035 is set by the load control program 11034.
0422) to the maximum allowable response time W (the processing result of the processing request input after inputting the processing request from the user terminal 10101
The value VI=VO-1/W, which is obtained by subtracting the reciprocal of the time required to output the processing request from the user terminal 10101, is determined as the upper limit value of the input speed of the processing request from the user terminal 10101. Thereafter, the reciprocal of VI is set as the value to be set in the read timing buffer 11025, and this value (1/VI) is set in the arrival interval setting value area 10422 of the load management table 11035. Furthermore, in general,
According to the logic of queuing, it is known that when the processing speed of a processing request is vo, and the allowable waiting time is W, it is necessary to keep the arrival speed of the processing request below vi. The determination of the upper limit value VI of the input speed of processing requests from the user terminal 10101 (the speed of arrival at the computer system) is based on this theory (72).

After the above processing, the load control program 11034 requests the system control program 11031 to output a message to the control terminal 10105 asking whether or not to temporarily stop the computer system shown in FIG. Wait for input (step 10014). −
Once the instruction to stop is input (step 10015), the load control program 11034 starts the load management table 11.
035 is stored in the save data file 10106 and the process ends (step 10017).

When an instruction to continue processing is input (step 10015)
, the load control program 11034 is executed by the user terminal 101
The terminal control device 10102 is set so that the input speed of the processing request from 01 does not exceed the upper limit value of the input speed of the processing request determined in the above process. After that, an input start instruction message is output to the control terminal 10105 (
Step 10016). That is, the load control program 11034 controls the read timing buffer 11025 (
73) Code and load management table 11 for hesetting
A message with the value of the arrival interval setting value area 10422 of 035 as set data is transmitted to the terminal control device 10102. Thereafter, a request is made to the system control program 11031 to output a message for starting normal input to the control terminal 10105. Note that, as described above, the terminal control device 1010
The set data of the telegram sent to the read timing buffer 11025 is sent to the read timing buffer 11025.

After that, input a processing request from the user terminal 10101,
Execute normal processing request processing (step 10019)
. Note that during normal processing request processing, the I10 control program 11022 of the terminal control device 10102 receives input from the user terminal 10101 at the cycle of the value determined in processing step 10013, which is currently set in the read timing buffer 11025. Read the processed processing request.

This prevents the input speed of processing requests from the user terminal 10101 from exceeding the maximum processing speed of the computer system shown in FIG. 18, and prevents the computer system from falling into the overload state (74) in FIG. Ru.

On the other hand, when the load management program 11034 is started in the control mode (step 10010), the load management table 110 is saved in the save data file 10106.
35 data and load management table 11035
(step 10018). Thereafter, the operation of step 10016 described above is executed.

Next, the load control program 11034 shown in FIG.
The operation of estimating the maximum processing speed (step 10012) will be explained in detail.

FIG. 23 shows the operation for estimating the maximum processing speed of the load control program 11034. Furthermore, load control program 11
FIG. 24 shows an example of how the values of each area of the load management table 11035 are set by the maximum processing speed estimation operation of 034. Note that FIG. 24 shows a case where the number of processing interval areas 10411 is eight.

The load control program 11034 first initializes the load (75) management table 11035 (step 105).
01). That is, the latest processing interval pointer area 104
27 to u I I+, arrival interval setting value area 104
21. Saturation processing interval area 10422° processing time area a10423. Processing time area b10424. Processing completion number area 10425. The average processing interval area 10426 and the total processing interval area 10411 are set to 110" (FIG. 24(a)). From now on, processing end events 1 to (processing end events that occur after notification of the processing result by the result notification program 11033) ) is awaited (step 15021).

When a processing end event occurs, first the value of the processing end number area 10425 is increased by LL 1 jr (step 1
5022).

Next, if this is the first processing completion event, that is, the value of the processing completion number area 10425 is it I I
If it is + (step 10503), the system control program 11031 is requested to obtain the current time (hereinafter, obtaining the current time and the obtained time (76) will be simply referred to as the current time), and the processing time area a10423
, and waits again for a processing end event (step 10504) (FIG. 24(b)).

If it is the second processing end event, that is, if the value of the processing end number area 10425 is 2” (step 1
0505), set the current time in the processing time area b10424 (Step 15061), and set the processing interval value calculated by the following formula to all processing interval areas 104 and 11 of the processing interval queue table 1004.1 ( Step 150 G2).

Processing interval value = (value of processing time area b10424 - value of processing time area a10423) (5) Also, the same processing interval value is set in the average processing interval area 10426 (step 10507) (Fig. 24 (C) ))
. Furthermore, the same processing interval value is set to the saturation processing interval area 10422.
Step 10508), and then set the value of processing time area b 10424 to processing time area a 1042.
3 (77) and waits again for the processing end event (step 1050).
9) (Figure 24(d)).

If it is the third or more processing end event, that is, the value of the processing end area 10425 is 111” or 112.
” (steps 10503 to 10505), first set the current time in the processing time area b10424 and perform the following (step 10510).

First, the average processing interval increment value calculated by the following formula is added to the average processing interval area 10426 (step 105
11) (Figure 24(e)).

Average processing interval increment value = ((value of processing time area b10424 - value of processing time area a10423) - value of the area next to the area indicated by the value of latest processing interval pointer area 10427) ÷ n (6) Latest The area next to the area pointed to by the value of the processing interval pointer area 10427 means that the value of this pointer area 10427 points to the last processing interval area 10411 (the rightmost area) of the processing interval queue table (78) 10041. For example, processing interval queue table 1
This refers to the processing interval area 10411 (leftmost area) at the beginning of 0041, and if not, the processing interval area 1041 pointed to by the value of this pointer area 10427.
This refers to the processing interval area 10411 next to (right) 10411.

Note that this process is based on the recent past n including the latest processing time interval.
The average value of the processing time intervals is calculated.

Next, the value of the latest processing interval pointer area 10427 is set to point to the next area of the area in which the value is set.After step 10512), the value of the latest processing interval pointer area 10427 points to the processing interval area 10411.
The processing interval value calculated by the above equation (5) is set (step 10513) (FIG. 24(f)). Through this process, the latest past n processing time intervals including the latest processing time interval are set in the processing interval queue table 10041.

After that, the absolute value of the difference between the value of the saturation processing interval area 10422 and the value of the (79) average processing interval area 10426 is calculated using 1 as the value obtained below (value of the saturation processing interval area 10422 x 1) If it is larger (step 10514), set the value of the average processing interval area 1042G to the saturation processing interval area 10422.Step 10515), set the value of the processing time area b10424 to the processing time area a10423, and execute the processing end event again. Waiting (step 10516) (Figure 24 (g)
)). Note that 1 above is the data processing program 110.
45 and the other programs shown in FIG. 19 are determined according to the degree of variation in time required to process one processing request, and is usually about several percent (kl).
The value is approximately 0.05). In addition, by using methods such as simulation and testing using the system itself shown in Figure 18, several thousand or more processing requests are processed in an overload state, and the value of the saturation processing interval area 10422 is set to finish processing one processing request. Each time you do this, record it and calculate the difference between the maximum value of the recorded value and the value of the most stepped value) ÷ (average of the recorded value (80
) Average value x 2) may be set to 1.

While repeating the above, the saturation processing interval area 10422
When the absolute value of the difference between the value of and the value of the average processing interval area 10426 becomes within (value of the saturated processing interval area 10422 x 1) (step 10514), the process ends.

With the above processing, in the saturation processing interval area 10422,
Processing time interval (operating at full processing capacity) when the computer system is overloaded and the processing time interval (time from one processing end event to the next processing end event) can no longer become smaller. (processing time interval) is calculated.

That is, the reciprocal of this time interval indicates the maximum processing speed of the system shown in FIG. In addition, saturation processing interval area 1
User terminal 1 at a shorter time interval than the value found in 0422.
If a processing request is input from 0101, the system will be overloaded. Therefore, by setting the read cycle of processing requests from the user terminal 10101 to a value longer than this value, it is possible to prevent the system from becoming overloaded (81).

Note that the read cycle of processing requests from the user terminal 10101 determined in step 100'13 of FIG. 22 to prevent the computer system of FIG. 18 from falling into an overload state is shown in FIG. 25, which will be explained below. It may be determined by any method.

That is, in step 10012, saturation processing interval area 1
After finding the value at 0422, step 10601 is not executed, but instead a value suitably larger than the obtained value is set in the arrival interval setting value area 10421, and the process is temporarily stopped. Next, start the load control unit 11034 in control mode, and from the user terminal 10101,
A processing request is input at an interval shorter than the time interval set in the arrival interval setting value area 10421, and the response time from inputting the request until the processing result is returned is observed for a while (step 10602). The observed response times are at maximum load.

If this is greater than the desired response time (step 10
603), in order to shorten and improve the response time (82), the value in the arrival interval setting value area 10421 is appropriately increased in order to reduce the request input speed, and the above-described operation is repeated (step 10605). If the response time is smaller than the desired response time (step 10603), there is still room to accept processing requests, so the value of the arrival interval setting value buffer 10046 is appropriately reduced and the above-described operation is repeated (step 10606).

Repeat the above and end when the desired response time is reached (
Step 10604). Note that the above operation may be performed automatically following step 10012 using the load control program 11034.

In addition, in Fig. 18, two computers 10103゜computer 1
0104, but the computer 10103
Processing data file 1o1o8゜Processing result file 1
0109 to create a queue by request program 110
42. A queue 11043 for each request, a processing request selection program 11044, a data processing program 11045 provided in the computer 10103, and a computer 10104 removed.
It is also possible to use a system consisting of one computer. Furthermore, a new (83) processing result file 10109. Processing data file 1
0108. Another computer is provided that waits for the control terminal 10105 and is connected to the request reception file 10107, and this computer is connected to the computer 10103 through a new communication line, and the same program as that of the computer 10104 is stored in the computer. Here, when a processing request is input, the computer 10103 selects the computer 10103 according to the type of the processing request.
04 or a newly connected computer to inform that the processing request has been accepted and execute the processing.

Furthermore, in each of the computer systems described so far, each computer may be configured to include a backup computer.

As described above, according to this embodiment, processing requests are no longer input to the system at a speed exceeding a certain upper limit value, and the system can continue to operate without falling into an overload state.

Next, still another embodiment of the present invention will be described.

(84) In the above-mentioned embodiment, by suppressing the input speed of processing requests (the speed at which they reach the computer system) to a value slightly smaller than the maximum processing speed (reciprocal of the saturation notification time interval) V○ of the computer system shown in FIG. , which prevents the computer system shown in FIG. 18 from falling into an overload state. With this method, the number of processing requests remaining in the computer system of FIG. 18 can be kept small, so the processing waiting time is relatively small and the response time is stable. On the other hand, the amount of processing requests accepted (the number of processing requests read per unit time) tends to be kept low. Therefore, below, we will explain an example in which the overall amount of processing requests accepted is larger even if there is some variation in response time (the response time may be long or short depending on the input time of the processing request). .

In the above embodiment, the upper limit VI of the input speed of processing requests is set to 1.
(The value is the maximum processing speed of the system
). In contrast, in this embodiment, two upper limit values (adjustment values, limit values) for the input speed of processing requests are set, one of which is a value v that is slightly smaller than the maximum processing speed vO of the system (85). 2,
The other one is set to a value VI that is slightly larger than ■O. At a certain time, when a processing request using VI□ is inputted as the upper limit value, and as a result a tendency is observed that the number of accumulated processing requests in the computer system increases, the upper limit value is switched to v□2. As a result, when a tendency for the number of retained processing requests to decrease is observed, the upper limit value is switched to VI.

Since the upper limit value is alternately switched in this way, the total amount of bone to be processed can be made larger than in the previous embodiment. The above is an overview of this embodiment.

The hardware configuration of this embodiment is the same as that of the previous embodiment.
This is the computer system shown in Figure 8. Also, part of the load management table 11035 of the embodiment, load control program 1
Part of the operation of 1034, request reception program 11032
Except for some of the operations, this embodiment is the same as the previous embodiment. In the following, parts that differ from the previous embodiment in operation will be explained.

FIG. 26 shows the load management table 11 of this embodiment, which is replaced with the load management table (86) 11035 of the above-mentioned embodiment.
Details of 037 are shown.

The load management table 11037 is the management table 1007
1. Arrival Interval Queue Table 10072 Processing Interval Queue Table 10041, which stores information for managing the arrival interval of processing requests and the completion interval of processing.

The information for managing the processing end interval consists of the same part as the load management table 11035 of the above-described embodiment and the processing interval variation area 17108, and the same numbers are used for the same parts. That is, the processing interval queue table 1004
1. Processing interval area 10411, management table 1007
1 saturation processing interval area 10422. Processing time area a
10423° Processing time area b10424. Processing completion number area 10425, average processing interval area 10426. The latest processing interval pointer area 10427 is exactly the same as the area of the processing interval management table 10042 of the above-described embodiment, and is used for the same purpose. In addition, the processing interval change area 17108 has (87) values in the average processing interval area 10426 between the time when the processing result of the processing request is notified to the user terminal 10101 and the time when the processing result of the next processing request is transmitted. Stores the amount of change, which is how much the value has changed.

This load management table 11037 further has a part for managing the time interval at which processing requests arrive. The inter-arrival queue table 10072 has n inter-arrival areas 1
It has 0721. Each area 10721 stores the arrival time intervals (intervals at which they actually arrive at the computer system) of the most recent n processing requests in the past. Note that the right end of the arrival interval queue table 10072 is cyclically managed as being logically connected to the left end. The latest arrival interval pointer area 17107 stores the value of a pointer pointing to the position of the arrival interval area 10721 where the latest arrival time interval is stored. Note that the pointer points the arrival interval area 10721 of the arrival interval queue table 10072 from the left to "1", '2' II 31j, ... 11 n31
point like this. The average arrival interval area 17106 stores the average value of the values stored in the arrival interval queue table (88) 10072.

In the request arrival number area 17105, the user terminal 1010
Stores the number of processing requests arriving from 1. Arrival time area a17103. The arrival time area b17104 is a work area for determining the arrival time interval of processing requests. Arrival interval setting value 1 area 17101. Arrival interval setting value 2 area 17102 includes terminal control device 10
A value to be set in the read timing buffer 11025 of 102 is stored. In addition, as described later, the value of arrival interval setting value 1 area 17101 is changed to the value of arrival interval setting value 1 area 17102.
be smaller than the value of The mode flag 17109 is a flag indicating that the load control program 11036 has shifted to the processing interval monitoring mode. At the time of IZ I II,
It indicates that it is in the processing interval monitoring mode, and indicates that it is not in it OII (details of turning on and off the flag will be described later).

In this embodiment, even if a processing request is temporarily input at a speed exceeding the system's maximum processing speed (saturation processing speed, i.e., speed corresponding to the saturation processing interval) V○ (89), the subsequent input If the speed is kept to a certain level lower than the maximum processing speed, it is possible to reduce the number of unprocessed processing requests that remain in the computer system shown in Figure 18, and it is possible to prevent the computer system shown in Figure 18 from falling into an overload state. It is based on the idea that there is. According to this embodiment, processing requests exceeding the maximum processing speed of the system can be input for a certain period of time, and the overall amount of processing requests accepted can be improved.

The load control program 11036 of this embodiment, which replaced the load control program 11034 of the previous embodiment, executes this control. Let me explain this.

Note that the request reception program 11032a (
(not shown) is the request reception program 11 of the above-mentioned embodiment.
In addition to the operation of step 032, after writing the processing request data to the request reception file 10107, an operation of generating a request arrival event is performed.

The operation of the load control program 11036 of this embodiment is the same as the operation of the load control program 11034 of the above-described embodiment shown in FIG. 22, except that the contents of the operations of step 10013 and step 10016 are different. . The different steps will be explained below.

Step 1 of the load control program 11036 of this embodiment
0013 (Load control program 110 of the above-mentioned embodiment
(corresponding to step 10012 of No. 34) will be explained.

When the maximum processing speed (VO) of the computer system shown in FIG. A VI2) similar to the VI of the first embodiment is determined. Thereafter, the reciprocal number (1/VL) of the determined upper limit value is set in the arrival interval setting value 2 area 17102. Further, a value calculated by the following formula is set in the arrival interval setting value 1 area 17101.

Area 17101 value = (Saturation processing interval area 10422 value Arrival interval setting value 2 Area 17102 value) + Saturation processing interval area 10422 value (91) Here, the area 10422 value is 1/VO, area 171
02 value is 1/VL, area 17101 value is almost 1
/VI□ is given. Therefore, the above area 17101 value (
More precisely, the reciprocal) is the limit amount of the input speed of processing requests (VO
-VI2) is a value that allows the input speed of processing requests to exceed the maximum processing speed of the computer system in FIG. 18 by approximately the same amount. In addition, if the response speed deteriorates extremely, the excess amount (value in parentheses in the above formula) should be changed to 1/2.1/3.
You can make it smaller like...

Furthermore, step 10016 of the load control program 11036 of the J actual flow example (corresponding to step 10016 of the load control program 11034 of the above-mentioned embodiment) will be explained.

FIG. 27 shows a load control program 11036 of this embodiment.
The operation of step 10016 is shown.

Furthermore, step 100 of the load control program 11036
Example (92) of how the value of the load management table 11037 is set by the operation in area 10721.16. The number of processing interval areas 10411 is
In each case, 8 cases are shown.

The load control program 11036 first initializes the load management table 11037 (step 10801).

That is, the latest arrival interval pointer area 17107 and the latest processing interval pointer area 10424 are set to 111'',
Arrival time area a17103.

Arrival time area b 17104. Processing time area a10
423. Processing time area b10424. Request arrival number area 17105. Processing completion number area 10425, average arrival interval area 17106. Saturation processing interval area 10422
．． Processing interval change area 17108. Average processing interval area 10426, total arrival interval area 10721, and total processing interval area 10411 are set to II O11. Further, the values of the arrival interval setting value 1 area 17101 and the arrival interval setting value 2 area 17102 are values set in advance based on the value of the previous saturation processing interval 10422 (1/v○) as described above, In this example, the 1/VO(93) value is "17.5" (not shown), the 17101 value is r15J, and the 17102 value is "20".

Also, set the mode flag 17109 to 0'' (second
Figure 8(a)).

Next, the value of the arrival interval setting value 1 area 17101 is transferred to the read timing buffer 1100 of the terminal control device 10102.
5, and then the control terminal 1010
Output an input start instruction message to Step 5 (Step 10)
802). That is, the load control program 11036 transmits to the terminal control device 10102 a message in which the code for setting the read timing buffer 1025 and the value of the arrival interval setting value 1 area 17101 of the load management table 11035 are set data. Thereafter, a request is made to the system control program 11031 to output a message for starting normal input to the control terminal 10105. Note that, as described above, the set data of the message sent to the terminal control device 10102 is set in the read timing buffer 11025.

(94) After that, input a processing request from the user terminal 10101,
Execute normal processing request processing (step 10019)
.

Note that from now on, the load control program 11036 will respond to a request arrival event (a request arrival event that is generated after the request is accepted by the program 11032a (not shown)) or a processing end event (a request arrival event that occurs after the request is accepted by the program 11032a (not shown)) or a processing end event (by the result notification program 11033). The process waits for the occurrence of the process end event (which occurs after notification of the process result) (step 1080).
3).

Here, when a request arrival event occurs (step 18
041), first set the value of request arrival number area 17105
I II is increased (step 18042).

Next, if this is the first request arrival event, that is, if the value of the request arrival number area 17105 is 1'' (step 10805), the current time is set to the arrival time area a.
17103 and waits for an event again (step 10806) (FIG. 28(b)).

If it is the second request arrival event, that is, (95
) If the value of the request arrival number area 17105 is 1721+ (step 10807), the current time is set to the arrival time area b1.
7104 (Step 18081), the arrival interval value calculated by the following formula is set in the arrival interval queue table 100.
72 total arrival interval area 10721 (step 18082).

Arrival interval value = (value of arrival time area b 17104 value of arrival time area a 17103) (7) Also, set the same arrival interval value in average arrival interval area 17106 (step 10809) (Fig. 28(C) )
. Thereafter, the value of the arrival time area b17104 is set in the arrival time area a17103, and the process waits for an event again (step 10810).

If it is the third or more request arrival event, that is, if the value of the request arrival number area 17105 is not "1" or "2" (steps 10805 to 10807), first,
Set the current time in the arrival time area (96) b17104 and do the following (step 108)
1.1).

First, the average inter-arrival interval increment value calculated by the following formula is added to the average inter-arrival interval area 17106 (step 108
12).

Average arrival interval increment value = ((value of arrival time area b 17104 value of arrival time area a 17103) - value of the area next to the area pointed to by the value of latest arrival interval pointer area 17107) ÷ n (8) Latest arrival The area next to the area pointed to by the value of the interval pointer area 17107 is the arrival interval if the value of this pointer area 17107 points to the last arrival interval area 10721 (-Koseki area) of the arrival interval queue table 10072. Refers to the arrival interval area 10721 (leftmost area) at the beginning of the queue table 10072,
Otherwise, it refers to the (97) arrival interval area 10721 next (to the right) of the arrival interval area 10721 pointed to by the value of this pointer area 17107.

Note that this process calculates the average value of the latest n arrival time intervals including the latest arrival time interval.

Next, set the value of the latest arrival interval pointer area 17107 to point to the next area of the area in which the deposit is made. After step 10813), the arrival interval area 10721 pointed to by the value of the latest arrival interval pointer area 17107
The arrival interval value calculated by equation (7) is set (step 10814) (FIG. 28(d)). Through this process, the latest past n arrival time intervals including the latest arrival time interval are stored in the arrival interval queue table 10072.

After that, if the value of request arrival number area 17105 does not exceed n (step 10815), arrival time area b
The value of 17104 is set in the arrival time area a17103 and the process waits for an event again (step 10816).

Even if the value of request arrival number area 17105 exceeds n (
Step 10815), if the processing interval monitoring mode (98) is set, that is, the mode flag 1710
If 9 is tL I II (step 10817), the value of the arrival time area b 17104 is set in the arrival time area a 17103, and the process waits for an event again (step 10816).

If the value of the request arrival number area 17105 exceeds n (step 10815) and the processing interval monitoring mode is not set, that is, the mode flag 17109 is set to rr O.
If u (step 10817), further perform the following.

First, the value of the average arrival interval area 17106 (the average value of the actual arrival interval) is set in the arrival interval setting value 1 area 17101 (
7) If it is smaller than the value (1/VL) (step 10818), this indicates that the actual arrival rate of requests exceeds the maximum processing speed (VO) of the system, so the arrival interval setting value 2 area 17102 value 1/
The input speed is suppressed by instructing to set VL (larger than the value 1/V1. of the same area 17101) in the read timing buffer 11025 of the terminal control device 10102 (99) (step 10819).

That is, the load control program 11036 includes a code for accessing the read timing buffer 1 and the arrival interval setting value 2 area 1 of the load management table 11037.
A message with the value of 7102 as set data is transmitted to the terminal control device 10102. Note that, as described above, the data from cella 1 of the message sent to the terminal control device 10102 are set in the read timing buffer 11025. After that, the state shifts to the processing interval monitoring mode, that is, the mode flag 17109 is set to 1'' (step 10820), and the value of the arrival time area b 17104 is set to the arrival time area a 17103, and the event wait is resumed ( Step 10821) (Fig. 11(e)
)).

Note that if the value of the average arrival interval area 17106 is not smaller than the value of the arrival interval setting value 1 area 17101 (step 10818), the value of the arrival time area b 17104 is simply set in the arrival time area a 17103, and the event wait is resumed. (Step (100) Step 10821).

On the other hand, when a processing end event occurs (step 180
41), First, set the value of the processing completion number area 10425 to II
I II increases.

Next, if this is the first processing end event, that is, if the value of the processing end number area 10425 is "1" (step 18232), the current time is set to the processing time area a.
10423 and waits for an event again (step 10824) (FIG. 28(f)).

If it is the second processing end event, that is, if the value of the processing end number area 10425 is II 2 II (
Step 10825), process the current time in time area b10
424 (Step 18261), the processing interval value calculated by the following formula is added to the processing interval queue table 1004.
1 in the total processing interval area 10411 (step 18262).

Processing interval value mi (value of processing time area b10424 - value of processing time area a10423) (9) (1
01) Also, the processing interval value is set in the average processing interval area 10426 (step 10827).

Thereafter, the value of the processing time area b10424 is set in the processing time area a10423, and the process waits for an event again (step 10828) (FIG. 28(g)).

If it is the third or more processing end event, that is, the value of the processing end number area 10425 is not It I IT or "2" (step 18232゜10825
) - First, the current time is set in the processing time area b10424 and the following is performed (step 10829).

First, the average processing interval increment value calculated by the following formula is added to the average processing interval area 10426 (step 108
30).

Average processing interval increment value = ((value of processing time area b10424 - value of processing time area a10424) - value of the area next to the area pointed to by the value of latest processing interval pointer area 10427) ÷ n ... (10) ( 10
2) The area next to the area pointed to by the value of the latest processing interval pointer area 10427 means that the value of this pointer area 10427 points to the last processing interval area 10411 (-Koseki area) of the arrival interval queue table 10041. For example, referring to the arrival interval area 10411 (the area with the first person) at the beginning of the processing interval queue table 10041,
Otherwise, it refers to the next (right) arrival interval area 10411 of the processing interval area 10411 pointed to by the value of this pointer area 10427.

In addition, the same average processing interval increment value is calculated in the processing interval change area 1.
7108 (step 10831).

This value corresponds to the time differential value of the average processing interval.

Next, set the value of the latest processing interval pointer area 10427 to point to the next area of the area that is being separated.After step 10832), the value of the latest processing interval pointer area 10427 points to the processing interval area 10411
The processing interval value calculated by equation (9) is set (step 10834) (FIG. 28(h)).

After that, if the value of the processing completion number area 10425 does not exceed n (step 10835), processing time area b
The value 10424 is set in the processing time area a10423 and the process waits for an event again (step 10836).

Even if the value of area 10425 exceeds n after processing is completed (
Step 10835), if it is not the processing interval monitoring mode, that is, if the mode flag 10109 is 0'' (step 10837), the processing time area b 104
．． The value of 24 is set in the processing time area a10423 and the process waits for an event again (step 10836).

If the value of the post-processing area 10425 exceeds n (step 10835) and the processing interval monitoring mode is set, that is, if the mode flag 17109 is "1" (
Step 10837), further perform the following.

First, if the value in the processing interval change area 17108 is negative (step 10838), this indicates that the system is coming out of the overload state, so the value in the arrival interval setting value 1 area 17101 is changed again to the terminal control device 10102. The input speed is increased again by sending an instruction to the read timing buffer 11025 and setting it in the read timing buffer 11025 (step 10839). That is, the load control program 11036 controls the read timing buffer 1025
Code and load management table 11 for setting
A message with the value of arrival interval setting value 1 area 17101 of 035 as set data is transmitted to the terminal control device 10102.

After that, step 10 cancels the processing interval monitoring mode, that is, sets the mode flag 17109 to "O".
840), sets the current time in the processing time area a10423 and waits for an event again (step 10841)
(Figure 28(i)).

Note that if the value in the processing interval change area 17108 is not negative (step 10838), the processing time area b104
By simply setting the value of 24 in the processing time area a10423, the process waits for an event again (step 10841).

(105) In the above manner, the acceptance of input processing requests at an input speed greater than or equal to the maximum processing speed of the system and the acceptance of input processing requests at an input speed less than the maximum processing speed are switched depending on the situation. This makes it possible to increase the overall amount of processing requests accepted while preventing the system from becoming overloaded.

Note that the arrival interval setting value 1 area 17101 and the 2 area 1 determined in step 10013 of the operation of this embodiment
The value of 7102 may be determined by the method shown in FIG.

That is, in step 10012, saturation processing interval area 1
After finding the value in 0422, do not execute step 10013. Instead, set a value that is appropriately smaller than the found value in the arrival interval setting value 1 area 17101, and a value that is suitably larger than the found value in the same area 2. set as a temporary value) - then stop processing (step 10901
).

Next, start the load control program 1036 in control mode, input a processing request from the user terminal 10101 at a time interval shorter than the time interval set in the interval setting value 1 area 17101, and input the request. The response time from when the process is processed until the processing result is returned is observed for a while (step 10902).

If the observed response time is gradually increasing (that is, if the number of requests remaining in the system is increasing) (step 10903), the user terminal 101
Stop inputting requests from user terminal 10101, change the value of arrival interval setting value 2 area 17102 to a larger value (lower the upper limit of input speed), and input requests from user terminal 10101 again in the same way. The observation is restarted at a speed (a speed corresponding to a time interval shorter than the time interval set in the area 17101) (step 10904).

If the observed response time no longer gradually increases (step 10903), it is determined whether the response time is a desired one (step 10905). If it is larger than the desired value (step 10905), user terminal 1
Stop inputting requests from 0101 and increase the value of arrival interval 1 area 17101 a little (step 10906).
, once again, return the value of the arrival interval setting value 2 area 17102 to a value slightly larger than the value of the saturation processing interval area 10422, and set the time interval shorter than the time interval set in the total arrival interval setting value 1 area 17101 from the user terminal 10101. Observe by inputting the processing request again at intervals (Step 1)
0907).

If the response time is less than desired (step 109
05), since it can withstand a slightly larger load, it prevents the input of requests from the user terminal 10101, makes the value of the arrival interval setting value 1 area 17101 a little smaller (step 10908), and sets the arrival interval setting value 2. Area 1
The value of 7102 is returned to the original value, and a processing request is again input from the user terminal 10101 at a time interval shorter than the time interval set in the total arrival interval setting value 1 area 17101, and observation is performed.

After repeating the above, if the response time becomes satisfactory (step 10905), the process ends (step 109).
09).

As described above, setting values of appropriate sizes for the arrival interval setting 1 area 17101 and the arrival interval setting value 2 area 17102 can be obtained by trial and error. Note that the above operations may be automatically performed following step 10012 using the load control program 11036.

According to the embodiments described above, there is an effect that the amount of processing requests accepted by the system can be increased while the system continues to operate without falling into an overload state.

In the above embodiment, the input speed of processing requests to the computer system shown in FIG. 18 was adjusted by changing the read cycle of processing requests from the user terminal 10101 in the terminal control device 10102. In this method, the operator of the user terminal 10101 is not directly informed that the system is overloaded, so the operator of the user terminal 10101
From the operator's point of view, in extreme cases, phenomena such as suddenly becoming difficult to enter input into the computer system shown in Fig. 18, and the ratio (109) of keys of the user terminal 10101 being locked increase, occur. This may cause instability in the computer system shown in the figure. One possible way to deal with this is to notify the operator of the user terminal 10101 that the computer system shown in FIG. 18 is overloaded. This example will be explained below.

In this embodiment, the terminal control device 10102 in FIG.
The rest is the same as the previous embodiments.

FIG. 30 shows the configuration of programs and buffers in the terminal control device 10112 of this embodiment, which is replaced with the terminal control device 10102 of FIG. 18. I10 control program 110
22. Communication control program 11023, read timing buffer 11025, communication buffer 11026. I1
0 buffer 11027 is the same as that in FIG.

Further, the time buffer 11128 is stored in the user terminal 1010.
This is a work buffer that stores the time when data is input from 1.

The system control program 11021 starts other programs (110) of the terminal control device 110112 when the power is turned on. Further, the current time is notified in response to a time acquisition request from another program of the terminal control device 10112.

The input amount control program 11124 adjusts the input speed and notifies the operator of the user terminal 10101 that the system is overloaded.

That is, the I10 control program 11022 transfers processing request data from the buffer in the user terminal 10101 to the ■10 buffer 110 at the maximum processable speed (minimum cycle).
Load it on 27. The input amount control program 11124 is
When processing request data is input to the I10 buffer 11027, the system control program 110 stores the current time.
21 and compares it with the value in the time buffer 11128.

The difference (the difference between the two processing request times before and after the current time) is determined by the read timing buffer 11.
If the value is smaller than 025, a message to the effect that the input processing request data is invalid is sent to the communication buffer 11026.
to output to user terminal 10101, I10
By Near 711027 (7) Discard the data.

If it is larger, the data in the I10 buffer 11027 is passed to the communication control program 11023 and transmitted to the computer 10103, and the value in the time buffer 11128 is updated to the current time. Also, as in the case of the terminal control device 10102 in FIG. 18, when a message with a read timing buffer set code is sent from the computer 10103, the read timing buffer is
Change the value of 025.

Note that input timing determination (time comparison) in the input amount control program 11124 and user terminal 10101
The computer 10103 may perform functions such as notifying data discard to the computer 10103.

According to the embodiments described above, input restrictions can be implemented while notifying the operator of the user terminal 10101 that the computer system shown in FIG. 18 is in an overload state, and there is an effect that unnecessary anxiety is not caused.

In the above embodiment, the computer system load state shown in FIG. 18 was determined mainly by the size of the request processing completion time interval, but instead, the load state was determined by the number of processing requests remaining in the system. You may.

In this case, the number of pending requests can be known by incrementing and decrementing the counter for pending requests upon arrival of a processing request, notification of processing results, and the like. And based on this judgment result,
What is necessary is to determine the upper limit value of the input speed of processing requests.

In these methods, it is necessary to create a criterion for determining the load state of the system using another method, but the effort required to determine the load is reduced compared to the previous embodiments.

In the above embodiment, the upper limit value of the input speed of processing requests was manually set based on the maximum processing speed that can be processed by the system. It may be set automatically.

As detailed above, when a processing request occurs at a speed exceeding the maximum processing speed of the actual computer system, the computer system can restrict acceptance of processing requests to prevent the computer system from falling into an overload state. Therefore, it is possible to prevent the response speed to the user from becoming extremely slow (113) or from abnormal conditions such as system failure.

In addition, based on the maximum processing speed of the computer system identified in advance, the acceptance of processing requests is limited to prevent overload while accepting as many processing requests as possible.
This has the effect of increasing the amount of processing requests accepted.

Furthermore, other embodiments will be described. FIG. 31 shows the overall configuration of the adaptive computer system handled in this embodiment. This system consists of a user terminal 201 used for inputting and outputting processing requests.
01, Computer 20 that accepts processing requests from user terminals
102 and its backup computer 20'l O3
, a computer 20104 . which receives a processing request from the computer 20102 and executes the processing of the request based on an instruction from the system terminal 20108 . It consists of a computer 20105, respective backup computers 20106, 20107, and a system terminal 20108 used by a specialized operator to input instructions for execution of requested processing and to monitor processing status. (114) Processing request reception files 20203, 20204 for recording accepted processing requests, computer 20104゜201
Processing request pending file 20230.05 that temporarily stores processing requests to be sent to 05. It has a processing data file 20211 for processing requests and a processing result file 20211 for recording processing results. User terminal 201
The processing request input from 01 is sent to the computer 20102, and the computer 20102
The processing request is sent to the computer 104 or the computer 20105. Computer 20104 (computer 20105) processes requests according to instructions from system terminal 20108, and sends the results to computer 20102 and system terminal 20108. Computer 20102 transmits the above information to user terminal 20101. Please note that the backup computer 20106. Calculator 20107 is computer 2, respectively.
0104. If the computer 20105 goes down, it is the computer that takes over the processing.

Below, the details of each part of the computer system shown in FIG. 31 are shown in FIG. 32, and the operation will be described in detail.

Calculator 20106. Calculator 20107 is computer 201
04, computer 20105 has exactly the same parts and performs the same operations, so the details are omitted here. Processing requests input from the user terminal 20101 are sent to the computer 2010
Sent to 2. Upon receiving a processing request from the user terminal 20101, the processing request reception/sending control unit 20201 of the computer 20102 determines the processing request type and sends the processing request to the computer 20102.
If the request type is to be processed in step 104, the processing request is written to processing request reception file (1) 20203, and computer 2
If the request type is to be processed in step 0105, it is written to the processing request reception file 20204. Also, start up the DB/DL system. In response to this, DB/DL system 2
0202 transmits the contents of processing request reception file (1) 20203 or processing request reception file (2) 20204 to computer 20104 and computer 20105, respectively. Load judgment/processing system switching control unit 20 of computer 20104
When a processing request arrives from the computer 20102, the processing request 205 checks the amount of data of the request type retained in the request type queue 20209, which queues processing requests for each processing request type, and determines whether the amount of data is below a certain threshold. If there is, control is transferred to the request type queue creation unit 20208 and the processing request is queued in the request type queue 20209.

If the amount of data for the request type exceeds the threshold, all the data for the request type is transferred to the request type queue 2 of the computer 20106 using inter-channel coupling 20220 as described later.
Copy to 0114. Request type queue creation unit 202
08 queues the processing requests in the request type queue 20209, the processing request conversion registration unit 20206 takes out the processing requests one by one from the request type queue 20209, converts them into the data format given by the data processing unit 20221, and then A request is made to the shared DB exclusive control unit 20207 and this is registered in the processing boot file 20212. At this time,
If the computer 20106 is not accessing the processing data file 20212, the shared DB exclusive control unit 20207 uses the DB/DC (117) system 20210 to access the processing cheater file 202.
Access 12. If the computer 20106 is accessing the data file 20212, wait for the access to complete.
access. In addition, the data processing unit 20221 reads the processed data file 20212 at regular time intervals,
This can be sent to the system terminal 20108. The specialized operator constantly monitors the processing data file 20212 on the system terminal 20108, and
A data processing instruction is issued from 108. calculator 20104
In response to this instruction, the data processing unit 20221 performs data processing and writes the processing results to the processing result file 2011. The DB/DC system 20210 sends the information newly written to the processing result file to the system terminal 20108 and computer 20102. The notification output unit 20223 of the computer 20102 sends the above information to the user terminal 201.
Output to 01.

FIG. 33 shows the request type queue 20 of the computer 20104.
Processing requests of the request type (118) that are accumulated in the request type queue 20 of the computer 20106 are sent to the request type queue 20 of the computer 20106.
218 is a diagram showing details of a method for copying to 218. The request type queue (20209 or 20218) is a buffer that stores processing requests sent from the computer 20102 for each request type on a first-come, first-served basis.
) is cyclically extracted from the beginning and processed.

In this figure, a processing request of the request type indicated by ILVT of the request type queue 20209 of the computer 20104 is sent to the computer 201Q6 via the inter-channel coupling 20220.
The request type is copied to a queue prepared in advance for the request type in the request type queue 20218.

Next, the operation of the load judgment/processing system switching control unit (20205 or 20214) will be explained in detail using the state transition table shown in FIG. Here, the position of the action being explained is indicated by (event number, state number). Also, the transition destination state is shown in the small box at the bottom right corner of the action. When the computer 20104 receives a processing request queuing instruction (event 1) while it is in the normal processing request queuing instruction waiting state (state 1), the request type queue 2
If the amount of data for the request type in 0209 does not exceed the threshold, the processing request is simply queued, but if it does, an instruction to stop sending the processing request is sent to the computer 201024 , waits for a response to stop sending a processing request (state 2) (1.1). In this state, when the processing request sending stop completion response (event 5) is returned, all the processing requests accumulated in the human output buffer, etc. of the computer 20102 and the computer 20104 are transferred to the request type queue 20.
In order to confirm that it has been written to the computer 209, the last data in the request type queue is sent to the computer 201.
02 and waits for a data integrity response via communication # (state 3).
) becomes (5.2).

In this state, a response (message #
Data integrity response (event 4) by computer 201
02, the computer 20106 sends a copy permission request for the centralized data in the request type queue 20209, and waits for a request type queue copy permission response (state 5). In addition, if there is a notification from the computer 20102 that the communication numbers do not match, the communication number will be returned again.
Send confirmation instructions and wait for data integrity response via email (
The state becomes state 4) (4.4).

When waiting for request type queue permission response (state 5), waiting for request type queue copy permission response (event 7)
Immediately upon receiving the processing request data, the computer 20106
and waits for a request type queue copy completion response (state 3) (7.5).

In this state, when a request type queue copy completion response (event 3) is received, an instruction to change the processing request destination is sent to the computer 20.
102 and waits for a normal processing request queuing instruction (state 1) (3.3).

On the other hand, the load judgment/processing system switching unit 20 of the computer 20106
214 normally receives a request type queue copy request (event 6) from the computer 20104 when it is in the state of waiting for processing request queuing (121), and if the above input is received at this time, the request type queue copy request (event 6) is received from the computer 20104. Waiting for a copy permission response (event 7) is returned to the computer 20104, and the processing request data is being copied (state 6).
6.1). When copy data is received in this state (event 8), this data is copied to the request type queue 20218 (8°6). Furthermore, when the copy is completed, a request type queue copy completion response (event 3) is sent,
Waiting for normal processing request queuing instruction (state 1)
(2.6) Also, waiting for response to stop sending processing request (state 2), waiting for response to copy completion of request type queue (state 2),
State 3) Waiting for data integrity response with 2 #s (state 4)
, request type queue copy permission response waiting state (state 5)
), if there is a queuing instruction (event 1) for a processing request, queuing is performed ((1, 2) to (1,
5)).

Next, the processing request reception/sending control unit 2 of the computer 20120
The operation of 0201 will be explained using FIG. The note numbers are the same as in FIG. 34.

Normally, it is in the state (state 1) of waiting for a processing request input (12
2) When a processing request is input (event 1), this data is sent to the processing request reception file (20203 or 20204).
write to. At this time, the processing request reception file to be written is selected based on the request type (1, 1). When the computer 20104 receives a processing request stop sending instruction (event 2) while waiting for a processing request input (state 1), it stops sending a processing request, returns a processing request sending stop completion response, and waits for a confirmation instruction ( State 2) is reached (2.1). When a message confirmation instruction (event 3) is received in this state, the message number sent is compared with the message number of the most recently sent data, and
If they match, it returns an affirmative response to the data integrity response using the communication number and waits for an instruction to change the processing request destination. If they do not match, it returns a negative response to the data integrity response using the communication number and waits for an instruction to confirm the communication number (state 2). ) remains (3.2). When the communication number is waiting for a request type destination change instruction (state 3), when a processing request destination change instruction (event 3) is received, the processing request destination is changed to computer 20106,
The data in the processing request pending queue is written to the processing request reception file, and the process waits for input of the processing request (4.3). Also,
If a processing request is input while waiting for a communication number inquiry or an instruction to change the processing request destination, this processing request is sent to this file. Write it to the processing request pending file 20230 ((1, 2L (1, 3)).

In addition, the request type queue 20209 of the computer 20104
When the number of processing requests exceeds the threshold, the processing requests are not automatically transferred to the computer 20106, and the processing requests are transferred to the request type queue 2.
By installing a device that constantly monitors 0209, the operator can
A trigger for the process for load distribution described above (a series of processes for transferring the load of computer 20104 to computer 20106) may be provided. Figure 36 shows the system configuration.

The system console 20231 monitors the load on each computer and provides a trigger for load distribution.

Note that the computer on which the load of the computer 20104 is distributed may be the computer 20107 or the computer 20105 instead of the computer 20106, and the load status of these computers may be checked to distribute the load to a computer with a smaller load. This selection may be specified by the user during system generation.

Note that the request type queue 20214 of the computer 20106
The type of processing request of the computer 20104 to be copied to may be other than the intensive request type. At this time, select the request type that exceeds the second threshold among the intensive request types and send it to the computer 20106.
It may be copied to the request type queue 20214.

According to the above embodiment, when the amount of data in the request type queue 20209 of the computer 20104 exceeds the threshold,
Since the centralized request type (or other request type) is transferred to the computer 20106 and processed in parallel, there is an effect that delays in processing request registration processing can be eliminated.

In addition, as shown in FIG. 37, odd-numbered processing requests from the beginning of the intensive request type are moved to computer 20106, and even-numbered processing requests are moved to computer 20107, so that the order of order processing request registration processing does not change. The processed data file may be updated while the computer 20106 and computer 20107 (125) DB/DC systems (20210 and 20219) are synchronized. For example, as a method for synchronization, a processing request message is stored in a processing data file, and the processing request conversion registration unit refers to it and performs synchronization control by comparing it with the processing request message to be registered next. conduct. According to this, the processing request registration processing is performed in parallel by the two standby machines with almost no load, so there is an effect that high-speed processing becomes possible.

Note that, as shown in FIG. 38, the processing data file is not shared and the computer 20104. The computer 20106 may have separate processing data files with the same content, and update both processing data files from normal times. At this time,
DB/DC of computer 20104 and computer 20106
The systems (20210 and 20219) have access information for the processing data files of the commercial computers, and perform the remote data access function (basic function of the distributed DBMS (126)) when there is an instruction to update the processing data file of the processing request conversion registration unit. and update the image processing data file at the same time. Note that the same process is performed when distributing the load. In this way, by having separate processing data files, even if the other file should be destroyed, the other file can be operated, which has the effect of improving reliability.

Note that the load on the computer 20104 is not distributed to the computer 20106, and the request type queue 20 in the computer 20104 is
By determining the load status from the number of processing requests for each request type in 209 and changing the number of processing requests to be registered at one time in the processing request conversion registration unit 20206 for each request type, the computer 20104 can be processed independently. The delay in processing request registration processing may be resolved by The processing of the processing request conversion registration unit 20206 at this time will be explained with reference to FIG. The processing request conversion registration unit 20206 registers the request type side queue 20
The load of the request type to be processed is determined from the number of processing requests in 209 (step 20801), and the number of processing requests to be processed per processing request registration is determined (step 20802).
. Next, processing request registration processing is performed based on this (step 20803).

By doing so, there is an effect that even a small-scale system without a backup system can improve the efficiency of processing request registration processing to some extent.

According to this embodiment, when the amount of data in the processing request queue of a computer in the computer system exceeds a threshold, processing request registration processing is performed to transfer the intensive request type or other request types to another computer for parallel processing. This has the effect of eliminating delays.

〔Effect of the invention〕

According to the present invention, the number of processing requests input to a computer system is adjusted so as not to exceed its processing capacity, and there is an effect that the risk of falling into an overload state is reduced.

[Brief explanation of drawings]

FIG. 1 is a flowchart showing the basic operation according to an embodiment of the present invention. FIG. 2 is a hardware configuration diagram of a computer system to which an embodiment of the present invention is applied. FIG. 3 is a configuration diagram of each program and each table applied within the computer system of FIG. 2. FIG. 4 is a diagram showing details of the contents of the processing request file. FIG. 5 is a detailed diagram of the type-specific processing request queue. FIG. 6 is a detailed diagram of the processing request management table. FIG. 7 is a detailed diagram of the common information area of the processing request management table. FIG. 8 is a detailed diagram of the entry area of the processing request management table. 9th
The figure is a detailed diagram of the load management table. Figure 10(a)
-(c) are flowcharts showing detailed operations of the processing request input and load observation program. Figure 11(a)~
(b) is a flowchart showing detailed operations of the load control and processing request conversion program and the load calculation program. FIG. 12 is a system configuration diagram of a computer system in which there is one computer, and FIG. 13 is a system configuration diagram of a computer system in which a plurality of computers are connected to a computer that accepts processing requests. FIG. 14 is a detailed diagram of the entry information of the processing request management table when the processing frequency is determined from the arrival rate of processing requests. FIG. 15 is a detailed diagram of the load management table when the processing frequency is determined from the arrival rate of processing requests (129). FIG. 16 is a flowchart of a process performed in place of the process of step 909 in the flowchart of FIG. 10(a) when determining the processing frequency from the arrival rate of processing requests. FIG. 17 is a flowchart used in place of the flowchart in FIG. 11(b) when the processing frequency is determined from the arrival rate of processing requests. FIG. 18 is a hardware configuration diagram of another embodiment. Figure 19 is the first
FIG. 9 is a diagram showing a program configuration in the computer of FIG. 8; FIG. 20 is a diagram showing the program configuration within the terminal control device. FIG. 21 is a detailed diagram of the load management table shown in FIG. 19. FIG. 22 is a flowchart showing the operation in the configuration of FIG. 18. FIG. 23 is an operation flow diagram of the load control program of FIG. 19. FIG. 24 is an explanatory diagram of the operation of the load control program. FIG. 25 is a flow diagram of one method for determining the set values of the load management table. FIG. 26 is a detailed diagram of a load management table according to still another embodiment of the present invention. FIG. 27(a) (130) is an explanatory diagram of the operation of the mouse. FIG. 29 is a flow diagram of one method for determining the set values of the load management table. FIG. 30 is a detailed diagram of a terminal control device of still another embodiment. FIG. 31 is an overall configuration diagram of a computer system according to still another embodiment of the present invention. FIG. 32 is a system configuration diagram of the computer system. FIG. 33 is a conceptual diagram showing the transfer of centralized request type data to a computer to another computer. FIG. 34 is a state transition table of the load judgment/processing system switching section of the computer. 35th
The figure is a state transition table of the processing request reception/sending control unit. Figure 36 shows how to monitor the load status on the system console.
FIG. 2 is a system configuration diagram when switching processing systems. FIG. 37 is a conceptual diagram showing that odd-numbered requests from the head of the intensive request type are transferred to a certain computer, and even-numbered requests are transferred to another computer. FIG. 38 is a diagram showing a system configuration when processing data files are not shared between computers. 39th
The figure is a flowchart of processing by the processing request conversion registration unit. 101... Calculator, 102... Calculator, 103...
- Processing request input terminal, 104... Data processing instruction terminal, 105... System control terminal, 106... Processing request file, 107... Processing result file, 108
... Processing data file, 201... Processing request reception program, 202... Write pointer transmission program, 203... Processing result output program, 204...
- Processing request human power/load observation program, 205... Processing request queue by type, 206... Processing request management table, 207... Write pointer writing program,
208...Load control/processing request conversion program, 20
9... Load calculation program, 210... Data processing program, 211... Processed data display program,
212... Command processing program, 213... System control program, 214... Load management table, 401... Common area of processing request management table, 4
02x...Processing request management table type entry, 5
09...Load control degeneracy flag, 802...Number of processing requests. a φ hard (υ) 3θ11 foil country 2θ2 4ρ2X built 10 (2) (kyu) Kikuzu 2/4 mel4 (2) circulation 5 (2) 2/4 base built 23 evil (A) plan 2z kuni CB)

Claims (1)

[Scope of Claims] 1. A load control method for a computer system that has a storage device that stores processing requests of a plurality of types, and that retrieves the processing requests stored in the storage device and executes the corresponding processing. Then, at least one of the queue length of processing requests stored in the storage device and the arrival rate of processing requests stored in the storage device is observed for each type of processing request, and the observed processing request Determining the number of processing requests to be processed for each type of processing request based on at least one of the length of the queue and the arrival rate of processing requests, and based on the determined number of processing requests to be processed, A load control method for a computer system, comprising: executing a process corresponding to a process request stored in the storage device for each type of process request. 2. In the load control method for a computer system according to claim 1, the number of processing requests to be processed for each type of processing request is determined based on the queue length and type of processing requests of the desired type. A load control method for a computer system, characterized in that the load control method is determined from the average length of a queue for each processing request. 3. The load control method for a computer system according to claim 1, further comprising controlling execution of the observing step, the determining step, and the executing step in response to a command from an input device. Features: Load control method for computer systems. 4. In the load control method for a computer system as set forth in claim 1, furthermore, if the length of the queue of processing requests for each type exceeds a predetermined length, acceptance of a new processing request is prohibited. A load control method for a computer system characterized by suppressing load. 5. A load control method for a computer system, which has a storage device for storing processing requests, and retrieves the processing requests stored in the storage device and executes the corresponding processing, the computer system processing requests per unit time. A computer system characterized by estimating in advance a possible substantially maximum number of processing requests, and adjusting acceptance of new processing requests per unit time so as not to exceed the estimated number of processing requests. Load control method. 6. The load control method for a computer system as set forth in claim 5, characterized in that the adjustment of acceptance of the new processing request is performed by controlling the time interval of acceptance of the new processing request. A method for controlling the load of a computer system. 7. The load control method for a computer system according to claim 5, wherein the adjustment of acceptance of the new processing request is performed by invalidating the processing request stored in the storage device. A method for controlling the load of a computer system. 8. In the computer system load control method as set forth in claim 5, estimating the substantially maximum number of processing requests that can be processed by the computer system per unit time increases processing requests to the computer system. The number of processes that can be processed per unit time is the number of processes that are completed per unit time when the time interval for completing processes corresponding to the frequently input processing requests becomes substantially constant. A method for controlling load on a computer system, characterized in that the number of processing requests is set to a substantially maximum number. 9. In the computer system load control method as set forth in claim 5, estimating the substantially maximum number of processing requests that can be processed by the computer system per unit time increases processing requests to the computer system. The number of processes that can be processed per unit time is the number of processes that are completed per unit time when the frequency is input and the end time interval of processes corresponding to the frequently input processing requests reaches a desired time interval. A method for controlling load on a computer system, characterized in that the number of processing requests is set to a substantially maximum number. 10. The load control method for a computer system according to claim 5, further comprising determining a number of other processing requests that is smaller than the substantially maximum number of processing requests that can be processed by the computer system per unit time; Adjustment of reception of new processing requests is such that, if a processing request exceeding the above-mentioned substantially maximum number of processing requests is input, the number of processing requests accepted per unit time is reduced to below the number of other processing requests determined above. If the processing end time interval corresponding to the processing requests stored in the storage device starts to decrease, the number of processing requests accepted per unit time is reduced to the substantially maximum number of processing requests. A load control method for a computer system, characterized in that the load is controlled to be as follows. 11. A load control device for a computer system, comprising a storage device that stores processing requests of a plurality of types, and a computer system that retrieves the processing requests stored in the storage device and executes the corresponding processing, The computer system includes means for observing at least one of the length of a queue of processing requests stored in the storage device and the arrival rate of processing requests stored in the storage device for each type of processing request; means for determining the number of processing requests to be processed for each type of processing request based on at least one of the length of a queue of processed processing requests and the arrival rate of processing requests; A load control device for a computer system, characterized in that, based on the number of processing requests to be processed, a process corresponding to a processing request stored in the storage device is executed for each type of processing request. 12. In the load control device for a computer system according to claim 11, the number of processing requests to be processed for each type of processing requests is determined based on the queue length and type of processing requests of the type to be obtained. A load control device for a computer system, characterized in that the load control device calculates the load from the average length of a queue for each processing request. 13. The load control device for a computer system according to claim 11, further comprising an input device connected to the computer system, and in response to a command from the input device, the observing means and the determining means A load control device for a computer system, characterized in that the computer system is provided with means for controlling the operation of the computer system. 14. In the load control device for a computer system according to claim 11, furthermore, if the length of the queue of processing requests for each type exceeds a predetermined length, acceptance of a new processing request is prohibited. A load control device for a computer system, characterized in that it includes a means for suppressing the load. 15. A load control device for a computer system, comprising a storage device that stores processing requests, and a computer system that retrieves the processing requests stored in the storage device and executes the corresponding processing, and the computer system includes: means for estimating in advance a substantially maximum number of processing requests that can be processed by the computer system per unit time; and means for accepting new processing requests per unit time so as not to exceed the estimated number of processing requests. A load control device for a computer system, comprising: means for adjusting; 16. In the load control device for a computer system according to claim 15, the means for adjusting the acceptance of new processing requests is means for controlling the time interval of acceptance of the new processing requests. A load control device for a computer system featuring features. 17. In the load control device for a computer system according to claim 15, the means for adjusting reception of new processing requests is means for invalidating processing requests stored in the storage device. A load control device for a computer system characterized by: 18. In the load control device for a computer system according to claim 15, the means for estimating the substantially maximum number of processing requests that can be processed by the computer system per unit time includes: means for frequently inputting processing requests, and when the completion time interval of processing corresponding to the frequently input processing requests becomes substantially constant, the number of processing completed per unit time is determined as the number of processing completed per unit time. 1. A load control device for a computer system, comprising: means for determining a substantially maximum number of processing requests that can be processed at one time. 19. In the load control device for a computer system according to claim 15, the means for estimating the substantially maximum number of processing requests that can be processed by the computer system per unit time includes: means for frequently inputting processing requests; and when the finishing time interval of processing corresponding to the processing request inputted frequently reaches a desired time interval, the number of processing completed per unit time is calculated as the number of processing completed per unit time. A load control device for a computer system, comprising: means for determining a substantially maximum number of processing requests that can be processed at one time. 20. The load control device for a computer system according to claim 15, further comprising means for determining a number of processing requests smaller than the substantially maximum number of processing requests that can be processed by the computer system per unit time. and the means for adjusting the acceptance of new processing requests adjusts the number of processing requests to be accepted per unit time to the other predetermined number if a processing request exceeding the substantially maximum number of processing requests is input. means for controlling the number of processing requests to be equal to or less than the number of processing requests; and when the time interval for finishing processing corresponding to the processing requests stored in the storage device starts to decrease, the number of processing requests accepted per unit time is controlled to be equal to or less than the number of processing requests; 1. A method for controlling load on a computer system, comprising: means for controlling the number of processing requests so that the number of processing requests is less than or equal to the maximum number of processing requests.