JPH08249199A - Inter-process delay time restricting method - Google Patents

Abstract

PURPOSE: To restrain inter-process delay time in a computer system starting a pseudo process whose content is the same as that of a real process concerned after the real process is started to be within a prescribed value. CONSTITUTION: Wherever the individual processes and pseudo processes are executed S11 and S12, the difference of the number of the processes is calculated S13. Whether the difference exceeds a prescribed value or not is judged S14. When it does not exceed, the subsequent real process is executed. When the difference exceeds the threshold on the other hand, a delay time restricting process is executed S15. The delay time restricting process is to terminate the delayed pseudo process S16 and to update the pseudo process to the content of the present process S17. Then, the process is shifted to process after S12 on the pseudo process, and whether the number of execution times exceeds the prescribed number of times or not is judged S18 on the real process. When the number is less than the prescribed number of times, the process after S11 are repeated, and the process is terminated when the number exceeds the prescribed one.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a computer system which starts a pseudo process having the same contents as the real process after the start of the real process, for example, a method for eliminating inter-process delay time in a message distributed processing system.

[0002]

2. Description of the Related Art For example, in a distributed computer system, if some failure occurs in a program process (hereinafter referred to as a process) for realizing a service on a specific computer, it is impossible to continue the process on another computer. Measures are usually taken to prevent the situation. As an example of this means, when a master process and a backup process of the same content are grouped and a plurality of groups cooperate to perform a predetermined message transmission service, when the master process of one group is abnormal, the backup process concerned A method is known in which the processing up to that point is switched and executed to send a message to another group. In a computer system that adopts such a method, only the master process sends a message to another group,
Although the backup process has the same contents, it does not actually send a message but only performs a pseudo sending process.

FIG. 8 shows a processing procedure in such a computer system. FIG. 8A shows the procedure of the master process,
The same (b) shows the procedure of the backup process. Referring to FIG. 8A, each time the master process transmits a message to another group (S61 (S is a processing step, the same applies hereinafter)), the master process notifies the backup side of the transmission completion (S62). It is determined whether notification has been given a predetermined number of times, that is, whether all backup processes have been notified (S6).
3) If the notification is given, the process is terminated.

On the other hand, the backup process is shown in FIG.
As shown in (b), it is determined whether or not the transmission completion notification has arrived from the master process side (S64), and when it has arrived, dummy transmission processing is performed (S65). In this way, the backup process does not overtake the master process because it performs the dummy transmission process upon receiving the transmitted notification from the master process.

When an abnormality occurs in the master process, the process tries to switch to the backup process. The message sending process in the backup process is performed after the master process finishes the dummy sending process for all the message sendings that have been processed. Be seen. That is,
The switching process is not completed until all the dummy transmission processes are completed.

[0006]

As described above, the backup process does not overtake the master process for the same contents, but the processing timing may be delayed more than the master process for some reason. When the number of message transmission processes and the number of dummy transmission processes associated therewith are large, the delay time is cumulatively increased. In this case, the timing of switching from the master process to the backup process is significantly delayed, and the message transmission process to other groups may be temporarily interrupted. Such a problem commonly occurs in a computer system in which a pseudo process having the same contents as the real process is started after the real process is started and the pseudo process is switched to the real process when an abnormality occurs in the real process. It was not taken.

In view of the above problems, an object of the present invention is to provide a computer system in which a delay time between processes becomes a problem.
It is to provide a technique for keeping the inter-process delay time always below a certain value.

[0008]

In order to solve the above-mentioned problems, the present invention provides a processed real process in a computer system for starting a pseudo process having the same contents as the real process after starting each real process. The difference between the number of pseudo processes and the number of pseudo processes, terminates the pseudo process when the detected value exceeds a predetermined value, and performs a predetermined inter-process delay process, for example, a real process with a clear process in the relevant area of the pseudo process. Provided is an inter-process delay time elimination method for executing a copying process.

Further, a master process for executing a predetermined message transmission process and a dummy transmission process having the same contents as the message transmission process are executed when the master process is normal, and the dummy transmission process is executed when the master process is abnormal. In a computer system that has a backup process that switches and executes message transmission processing instead, and message transmission processing always precedes dummy transmission processing,
Detects the difference between the number of message transmissions processed by the master process and the number of dummy transmissions processed by the backup process, and when the detected value exceeds a specified value, the master process terminates the backup process and performs a specified inter-process delay process. For example, there is provided an inter-process delay time eliminating method for executing a process of copying an already processed master process (message transmission process or the like) to a corresponding area of a backup process.

[0010]

When the delay transmission processing of the dummy transmission processing (pseudo process) on the backup process side with respect to the message transmission processing (real process) on the master process side, it is necessary to detect the degree of delay and set it to an appropriate frequency. is there. This is because if the delay resolution processing is performed frequently, the time required for the delay resolution processing itself becomes long and the normal processing does not proceed promptly. Therefore, in the present invention, highly reliable delay detection is performed by using the difference between the number of message transmissions processed by the master process and the number of dummy transmissions processed by the backup process as an index of inter-process delay time. For example, FIG.
As shown in, after execution of each actual process (S1
1) When the corresponding pseudo process is executed (S1
2) The computer system calculates the difference between the number of real processes and the number of pseudo processes (S13), and determines whether this difference exceeds a predetermined threshold value (S14). Execute the real process of. On the other hand, if the difference exceeds the threshold value, delay time elimination processing is executed (S15). In this delay time elimination process, the delayed pseudo process is first ended (S16), and the pseudo process is updated to the content of the current real process (S17). After that, the pseudo-process is moved to the processing of S12 and thereafter, and it is determined whether or not the number of executions of the actual process exceeds a predetermined number (S18). S1 if the specified number of times is not exceeded
The processing after 1 is repeated. Further, when the number of times exceeds the predetermined number, the processing is ended. By doing so, it is possible to keep the inter-process delay time below a certain value while suppressing the execution of excessive delay elimination processing.

The actual process is used as a master process for transmitting a message and a backup process for performing dummy transmission as a pseudo process, and at the time of delay elimination processing, the content of the message transmission processing that has been processed in the master process is copied to the corresponding area of the backup process. Thus, there is no need to newly generate or prepare a process dedicated to delay cancellation, and it is possible to quickly cancel delay. This copy process can be easily executed by, for example, a process generation system call, and after the delay is resolved, the process is started from the state immediately before the delay resolution process, so that the processes up to that point are not wasted. In addition, the loss of time due to the delay elimination process is minimized.

[0012]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will be described below in detail with reference to the drawings. FIG. 2 is an explanatory diagram of a process for realizing a specific message transmission service, in which one group is composed of a master process 1 which is an actual process and two master processes (A, B) 2 and 3 which are pseudo processes. It shows an example in the case of configuration.

Each process has a service processing unit 1
1, 21, 31 and the message transmission control unit 12, 22,
32, and each message transmission control unit 12, 22, 3
2 is a message transmission management unit 121, 221, 321
And message transmission operation units 122, 222, and 322.

Message transmission management units 121, 221, 3
21 stores a group list (grouplist) which is a process list in each group.
It manages essage-count (mc) and group-info [pid] (GI). mc indicates the number of message transmissions (the number of dummy transmissions) for which the own process is completed, and GI
Indicates the number of transmissions (the number of dummy transmissions) that the other process indicated by [pid] has completed.

Message transmission management units 121, 221, 3
Reference numeral 21 detects the delay of the backup process and eliminates the delay of the backup process using the mc and GI. Further, the message transmission operation units 122, 222, 32
2 is a process for actually transmitting a message to the line in the case of a master process, a dummy transmission process in the case of a backup process, and further performing a notification process between the processes. In addition, each process is
As shown in the figure, it is assumed that the computer runs on each computer connected by a line.

FIG. 3 is a service processing procedure diagram in the configuration of FIG. 2, in which the master process 1 sequentially sends a message to another group according to the flow of the service processing, and the message transmission completion notification is sent to the backup process 2
3 to each of the backup processes 2 and 3 and the dummy transmission completion notification from the master process 1
It shows how to notify. In the figure, "mc = p +"
1 ... ”is mc of the message transmission management unit of the process every time message transmission or dummy transmission is completed.
"1" ... is added to the value of
[A] = p + 1 ”indicates that the dummy transmission completion notification has been added once to the master process 1 from the backup process (A) 2.

The processing procedure of FIG. 3 and the delay time elimination processing will be described with reference to FIGS. FIG. 4 is a processing procedure diagram in the master process 1. For convenience, one backup process (eg backup process (A)
Although an example of communication with 1) is shown, the same processing is performed for the other backup process (B) 2. The same process is executed when each backup process 2 or 3 is switched to the master process.

Referring to FIG. 4, the master process 1 is
The GI of the message management unit 121 of its own process is searched to grasp the delay of the backup process 2 (S21).
That is, the delay on the backup side is detected from the difference between the processed message transmission number mc and the dummy transmission number based on the dummy transmission completion notification notified from the backup process 2. It is determined whether or not this delay (difference in the number of transmissions) is behind a predetermined threshold value (S22), and if it is behind, delay elimination processing is performed (S23). This process will be described later. If the delay is less than or equal to the threshold, the message is sent to the process of another group (S24). Then, the transmitted message information is accumulated and the value of mc of the message management unit 121 is added by "1" (S2
5) The message transmission completion notification is sent to the backup process 2 (S26). It is determined whether or not this message transmission completion notification has reached a predetermined number of times, that is, whether or not notification has been sent to all backup processes (S27), and when the number of notifications has reached the predetermined number of times, the processing ends.

On the other hand, in the backup process 2, FIG.
As shown in FIG. 1, first, it is determined whether the message transmission completion notification has arrived from the master process 1 (S31), and if it has arrived, the dummy transmission processing corresponding to the message processing is started (S32), and the dummy transmission processing is started. When the transmission process ends, m of the message transmission management unit 221 of the own process
The value of c is incremented by "1" and the master process 1 is notified of the dummy transmission information, that is, the value of mc (S).
34).

The master process 1 is, as shown in FIG.
The dummy transmission information, that is, the value of mc is received from the backup process 2 (S41) and is stored in the message transmission management unit 121 of its own process. Specifically, the corresponding GI is updated (S42). In this way, the master process 1 can know the dummy transmission information that the backup process 2 has finished. Therefore, when transmitting a message to another group, the master process 1 always confirms the dummy transmission information obtained as described above. If the delay is smaller than the predetermined threshold value, the subsequent message is transmitted as it is.

Next, the delay elimination processing in S23 will be described in detail. FIG. 7 is a schematic diagram of the procedure of the delay elimination process according to the present embodiment. Referring to FIG. 7, when the master process 1 detects the delay of the backup process 2 as described above, the master process 1 terminates the delayed backup process (S51), and the GI of the message transmission management unit 121 of its own process determines that Items related to the process are deleted (S52). Next, a new process is created. This is generated, for example, by the master process 1 copying memory contents of its own current process (mc value and other related variables). The copying process can be easily realized by using a fork-type process generation system call (S53). Then, this new process is added to the GI (S54), and the value of each corresponding GI is set to the same value as the mc value of the master process itself (S55). Finally, the group list is updated, the deleted process and the added process are notified to the backup process 2, and a GI update is requested (S56). After that, the message transmission processing to other groups and the message transmission completion notification to each backup process 2 and 3 are continued.

As described above, in this embodiment, the master process 1 uses the number of message transmissions and the backup process 2,
Since the number of dummy transmissions processed in 3 is constantly monitored, and the delay elimination processing is started when the difference in the number of transmissions by each process exceeds the threshold value, the number of excessive delay elimination processings is suppressed. On the other hand, the inter-process delay can always be kept within a certain time. Further, since the message transmission contents that the master process 1 has already processed are copied to the backup processes 2 and 3 during the delay elimination processing, the inter-process delay can be eliminated easily and quickly. As a result, the master process 1 to the backup process 2
The switching time to (3) is reliably shortened as compared with the conventional case, and the interruption of service does not occur.

The above is an example in which one master process 1 and two backup processes 2 and 3 are arranged for each group, but the number of processes and the group configuration may be arbitrary. Further, as long as the pseudo process having the same content starts after the start of the actual process, the present invention can be similarly applied to the service including the message transmission process. The processes may be executed in parallel on the same computer.

[0024]

As is apparent from the above description, according to the present invention, the difference between the numbers of processing of the real process and the pseudo process, for example, the numbers of transmissions of the master process and the backup process, is kept within a certain value. Therefore, the interprocess delay time can be efficiently eliminated. Therefore, the switching of processes due to a failure or the like becomes quick, and the reliability of the computer system is improved. In particular, when distributed processing of one service is performed by multiple groups, such as in a distributed computer system, it is possible to prevent the processing of other groups from being stagnated or unable to continue due to the occurrence of an error in a specific group. , Its effect is remarkable.

[Brief description of drawings]

FIG. 1 is a schematic procedure diagram of an inter-process delay time elimination method of the present invention.

FIG. 2 is an explanatory diagram of a process for realizing a message transmission service.

FIG. 3 is an explanatory diagram of a service processing procedure between the processes shown in FIG.

FIG. 4 is a processing procedure diagram of a master process according to the present embodiment.

FIG. 5 is a processing procedure diagram of a backup process according to the present embodiment.

FIG. 6 is a reception processing procedure diagram of a master process according to the present embodiment.

FIG. 7 is a procedure diagram of delay time elimination processing according to the present embodiment.

FIG. 8A is a diagram showing a processing procedure of a conventional master process of this type, and FIG. 8B is a diagram showing a processing procedure of a backup process.

[Explanation of symbols]

 1 Master Process 2, 3 Backup Process 11, 21, 31 Service Processing Unit 12, 22, 32 Message Transmission Control Unit 121, 221, 321 Message Transmission Management Unit 122, 222, 322 Message Transmission Operation Unit

Claims (4)

[Claims] 1. In a computer system in which a pseudo process having the same contents as the real process is started after the start of each real process, a difference between the number of processed real processes and the number of pseudo processes is detected. A method for eliminating inter-process delay time, which comprises terminating a pseudo process when a predetermined value is exceeded and executing a predetermined inter-process delay process. 2. The inter-process delay time elimination method according to claim 1, wherein the inter-process delay process is a process of copying a processed real process to a corresponding area of the pseudo process. 3. A master process for executing a predetermined message transmission process, a dummy transmission process having the same contents as the message transmission process when the master process is normal, and a dummy transmission process when the master process is abnormal. Instead, in a computer system that has a backup process that switches and executes message transmission processing, and message transmission processing always precedes dummy transmission processing, in a computer system, the number of message transmission processing processed by the master process and the dummy transmission processed by the backup process. Detect the difference from the number,
A method for eliminating inter-process delay time, characterized in that, when the detected value exceeds a predetermined value, the master process terminates the backup process and executes a predetermined inter-process delay process. 4. The method for eliminating inter-process delay time according to claim 3, wherein the inter-process delay process is a process for copying a master process that has already been processed to a corresponding area of a backup process.