JPH02266441A - Process end monitoring system - Google Patents

Abstract

PURPOSE:To reduce overhead by setting up limit time generated first of all as check time, and only when the current time the check time, executing the time-over processing of a process whose limit time is less than the current time. CONSTITUTION:Without controlling time up to time-over in each process, the limit time determining time-over is controlled in each process, the first limit time out of respective limit times is set up as the check time 5213 and the current time 5212 is compared with the check time 5213 at a fixed period. When the current time 5212 exceeds the check time 5213, the time-over processing for an unended process whose limit time is smaller than the current time 5212 is executed. Since the time-over processing is executed in accordance with the limit time of the process proposed to be ended at first without executing time-over processing in each fixed period, useless time-over processing is not generated.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to an electronic computer system that processes a plurality of processes in parallel. The present invention relates to a method of monitoring whether the elapsed time from when each process is started to when it ends does not exceed a predetermined elapsed time limit value.

[Conventional technology]

Conventionally, to monitor the end time of a process in a computer system, as described in Japanese Patent Laid-Open No. 58-192152, each process has a time counter, and the monitoring time is set in the time counter when the process is started. Every time there is a timer interrupt that occurs at regular intervals, 1 is subtracted from the time counters of all running processes, and a check is performed to see if the value after subtraction is 0. The method was to detect a process that caused a timeout as a timeout.

[Problem to be solved by the invention]

In the above conventional technology, each time a timer interrupt occurs, the time counter is updated and the time counter is checked to 0 for all processes running at that time, and if any abnormality occurs, However, no consideration was given to the fact that processes exceeding the elapsed time limit would not generate acid, and there was a problem in that wasteful checks were performed every time a timer interrupt occurred.

An object of the present invention is to reduce as much as possible the amount of processing that must be performed each time a timer interrupt occurs, and as a result, reduce the overhead for monitoring process elapsed time per process. be.

[Means to solve the problem]

In the conventional method, the remaining time until a limit value of elapsed time has passed (hereinafter referred to as time-over) has been managed for each process. Therefore, it was necessary to update the remaining time of all processes every time a timer interrupt occurred at a fixed time interval.

Therefore, instead of managing the time until timeout for each process, we manage the limit time at which timeout should occur for each process, set the earliest limit time among them as the check time, and check the current time at regular intervals. The check times are compared, and if the current time exceeds the check time, time-over processing is performed for unfinished processes whose limit time is smaller than the current time. Also, if each process ends before the limit time, the timeout processing at the check time will be canceled and the next check time will be the earliest among the limit times of the currently running processes. This is what I did.

[Effect]

In this method, what is managed for each process is not the remaining time until a timeout occurs, but the time at which a timeout occurs. Therefore, after setting the time at which the process will time out at the start of the process, there is no need to perform a countdown operation.

In addition, instead of processing a process that has timed out at regular intervals, the earliest limit time in the process is set as the check time, and the current time and check time are only compared at regular intervals. However, when the check time is exceeded, time-over processing is performed, and at the same time, the earliest limit time from among the currently executing processes is set as the next check time.6 In other words, instead of processing time-over at regular intervals, In line with the time limit of a process that is scheduled to end early.

By performing time-over processing, wasteful time-over processing will not occur.

〔Example〕

An embodiment of the present invention will be described below with reference to FIGS. 1 to 10.

FIG. 1 shows an overall diagram of an online system that implements the present invention. Online controller 1.

When a request for business processing is received from the terminal device 2, the database 3 is searched, updated, etc. according to the contents of the request, and the results are returned to the terminal device 2. In the online controller 1, there are a plurality of tasks 4 that execute business processes, and the plurality of business processes are executed in parallel. Although the database 3 is shared by a plurality of tasks 4, the plurality of tasks 4 cannot update the database 3 at the same time. Therefore, when task 4 starts using database 3, it prohibits other tasks 4 from using database 3 and then starts using database 3, and when task 4 finishes using database 3, Allow task 4 to use database 3. If use of database 3 is prohibited when task 4 starts using database 3, task 4 is forced to wait until use of database 3 is permitted.

In such an online system, there is a possibility that an abnormality may occur in task 4 during execution of business processing, and task 4 may enter an infinite waiting state. If the task 4 enters the w-limit wait state while prohibiting other tasks 4 from using the database 3, the other tasks 4 will be unable to use the database 3.

Therefore, task 4 that is in an infinite wait state must be forcibly terminated, and if task 4 is prohibited from using database 3, other tasks 4 must be able to use database 3. . Therefore, for each type of business process, the maximum elapsed time (hereinafter referred to as the limit elapsed time) from when Task 4 starts executing the business process until it finishes is determined, and when the limit elapsed time has elapsed. Task 4 that does not finish the business process even after the completion of the task (hereinafter referred to as time-over)
If so, it is assumed that an abnormality has occurred, and the task 4 is forcibly terminated. The timer management 5 of the online controller 1 performs control to detect the task 4 that has timed out. The timer management 5 receives and receives triggers (hereinafter simply referred to as timer interrupts) at regular intervals from the operating system 6 (hereinafter simply referred to as O8) and monitors whether there are any tasks that have timed out. .

Figure 2 shows the configuration of timer management. Timer management consists of a task 51 that specializes in timer management (hereinafter simply referred to as timer task) and data 52 necessary for timer management (hereinafter simply referred to as timer data). Consists of. Timer task 5
1 exists only in the online controller 1.

The timer task 51 refers to the timer data 52 and checks whether there is any task 4 that has timed out among all the tasks in the online controller 1. If there is a task 4 that has timed out, the timer task 51 executes that task 4. Force termination. However, the timer task 51 does not always monitor whether there are any tasks that have timed out, but rather determines a certain time and checks whether there are any tasks that have timed out only at that time. The time at which the timer task checks whether there are any tasks that have timed out is called the check time.

There are two timings for determining the check time, and the methods will be described below.

The first timing is at the previously set check time,
This is when the timer task 51 is started, and among the tasks 4 that are executing business processing at that time and are not timed out, check the time at which the timeout is scheduled for the task 4 that is most likely to timeout the earliest. shall be.

The second timing is when task 4 starts executing the business process, and task 4, which starts the business process, adds the limit elapsed time to the current time and determines the time when the time will expire (hereinafter referred to as the limit time). If the limit time of the task 4 is later than the check time, the check time is not changed, but if the limit time of the task 4 is earlier than the check time, the limit time of the task 4 is checked. Time.

The timer task 51 is activated at regular intervals by a timer interrupt from O8. The timer task 51 started by a timer interrupt checks whether the check time has been reached, and only when the check time has been reached, checks whether there is any task 4 that has timed out, and when the check is finished, resets the check time. and stops processing at the next timer interrupt. When the timer task 51 is activated by the timer interrupt of O8, if the check time has not yet been reached, the process is stopped at the next timer interrupt without processing anything.

FIG. 3 shows the structure of the timer data 52. The timer data 52 includes a system table 521 that manages information regarding the entire online system, and a task table 522 that manages information regarding each task. The task table 522 is created for each task 4. Each task 4 is
The address of the task table 522 of the task 4 is held in a general-purpose register, and the task table 522 is stored at any time.
can be referred to.

FIG. 4 shows the configuration of the system table 521.

The system table 521 has a first task table address 521 for connecting task tables 522 in a chain.
1, a current time 5212 that displays the current time, and a check time 5213 that checks whether the timer task 51 has timed out any task. The current time 5212 is updated by the timer task 51 at regular intervals.

FIG. 5 shows the configuration of the task table 522. The task table 522 has a next task table address 5221 for connecting the task tables 522 together in a chain.
It consists of a system table address 5222 pointing to the system-pull 521, and a limit time 5223 at which the task has passed its limit elapsed time and is timed out. However, the limit time 5 for tasks that are not executing business processing
223 is set to 0.

Next, the flow of processing when the execution of business processing by task 4 starts, the flow of processing when execution of business processing by task 4 ends, and the flow of processing of timer task 751 will be described.

FIG. 6 shows the flow of processing at the start of business processing execution. 6o
-, the system table address 5222 is obtained by referring to the task table 522 based on the address of the task table held in the general-purpose register. In 6b,
Receives the specified value of the limit elapsed time for the business process that the task will execute from now on.

In step 6c, the current time 5212 is obtained from the system table. At step 6d, the limit time of the task is obtained by adding the current time obtained at step 6c and the limit elapsed time obtained at step 6b, and is set in the limit time 5223 of the task table 522. In 6e, the check time 5213 on the system table 521 is compared with the limit time 5223 obtained in 6d,
Only when the limit time 5223 is smaller, processing is performed to set the limit time 5223 obtained in 6d to the check time 5213 on the system table 521 in 6f.

FIG. 7 shows the flow of processing at the end of business processing execution.

At Mt. 7, the limit time 5223 on the task table 522 of the task is cleared to 0 based on the task table address of the task held in the general-purpose register.

FIG. 8 shows the flow of processing of the timer task 51.

The timer task waits for a trigger from O8 at regular time intervals and is activated at regular time intervals.

When the timer task starts, first, 8b receives the current time from O8, and stores the current time 5212 on the system table.
to set. In addition to receiving and receiving the current time from the OS each time, the current time update process performed here may also include a method of simply adding 1 to the current time 5212. In this case, the only change is the time and time unit used in timer management, and other methods remain unchanged.

Next, in 80, the current time obtained in 8b is compared with the check time 5213 on the system table, and if the check time has not been reached, the process is stopped at peak 8 without processing anything. If the check time has been reached, the maximum possible time is set in the check time 5213 on the system table in step 8d. Next, in 88, the first task table address 5211 is obtained from the system table, and while tracing the chain of task tables 522, the address of the task table 522 that performs the next process on all task tables 522 is obtained. Check whether the limit time 5223 is set in the task table 522, and if it is not set.

In 912, the next task table 52 is executed without performing any processing.
Find address 2 and move on to checking the next task table. If the limit time 5223 is set, 8h
Move to. In 8h, the current time obtained in 8b and the limit time 5
223, and if the limit time 5223 has been reached, the limit time 5223 on the task table is cleared in 81, and the O8 is contacted to forcibly terminate the task.

If the limit time has not been reached, the limit time 5223 of the task in question is compared with the check time 5213 on the system table in 8j, and if the limit time occurs earlier than the check time, the limit time of the task in 8j is compared. 5223 is set to check time 5213 on the system table. When the above processing is completed for the task table, 8
In Q, the next task table address 5221 is obtained from the fully processed task table, and the process moves to the next task table. When the processing for all task tables is completed and it is determined that there is no next task table address at 8f, the process continues until the trigger is received at fixed time intervals from the next ○S.
Processing stops at the mountain.

In this method, the current time 5212 on the system table is updated only when the timer task is activated.

and does not check whether the check time 5213 has been reached. Therefore, there is a possibility that an error equal to plus or minus the activation interval of the timer task may occur in the time handled by timer management. If it is necessary to reduce the error, the activation interval of the timer task can be shortened, but the overhead for timer management increases accordingly.

Therefore, the activation interval of the timer task needs to be set to the longest value that satisfies the accuracy required for timer management.

The effects of this embodiment are shown below.

First, let's look at how the check interval by the 5-timer task changes depending on the average business process occurrence interval, business process execution time, and limit elapsed time.

It is shown in FIGS. 9 and 10.

FIG. 9 shows an example where the difference between the limit elapsed time and the actual business process execution time is smaller than or equal to the average business process occurrence interval. In this case, at the limit time of Taskle, Taskle +1
Business processing is being executed.

Therefore, the check by the timer task only needs to be performed at the limit time of each task. In other words, the check interval = average business processing occurrence interval.

Next, the difference between the limit elapsed time and the actual business processing execution time is
FIG. 10 shows an example of a case where the interval is larger than the average business process occurrence interval. In this case, at the limit time of the task, the business processes of some tasks started after the task have already finished. Therefore, the earliest among the limit times of the tasks being executed at that time becomes the next check time.

In other words, from 0 or more, which is the check interval = limit elapsed time - actual business process execution time, the check interval is expressed in a general formula as follows: check interval = MAX (average business process occurrence interval, limit elapsed time - actual business process execution time) time).

In this embodiment, unless an abnormality occurs, the limit elapsed time =
Specify a value that is sufficiently larger than the time expected to be required to execute the actual business process as the limit elapsed time so that it does not take the actual business process execution time. Therefore, in this embodiment, the check processing by the timer task can be performed at the time interval determined by the limit elapsed time - the actual business process execution time, and this interval is determined by the increase in the number of business processes occurring per unit time. Even if you do, it won't change.

Next, as shown in the section on conventional technology, a time counter is provided for each task that executes business processing.

When starting the execution of a business process, set the limit elapsed time to the time counter of the task in question, and each time a timer interrupt occurs, the time counters of all tasks running the business process will be decremented by 1. The timer management overhead of this embodiment will be compared with a method (hereinafter referred to as a conventional method) in which a time counter that reaches O is detected as a time over.

First, define variables as follows.

S: Overhead at the start of business processing E: Overhead at the end of business processing C: Overhead that always occurs every time a timer task is started (including O8 overhead) A: Is there a timeout for one task during execution of business processing? Overhead for checking whether or not business processing is being executed B: Overhead for skipping checking for tasks that are not executing business processing M: Total number of tasks in the online controller D = Timer task activation interval N: Business processing occurrence per unit time Number T: Average execution time for one business process L: Limit elapsed time When a timer interrupt occurs and the timer task starts,
The number of tasks P that is executing business processing is P=T−N
However, T −N<M Therefore, the number of tasks Q that is not executing business processing is Q=M
-P=M-T-N. First, the conventional method is used. Calculating timer management overhead X per unit time: X = (S + E)
・N+ (A-P+C)/D= (S+E) ・N
+ (A-T-N 〇)/D, which is 0th order. Next, find the timer management overhead per unit time according to this embodiment. The check interval U of the timer task is U=LT. Overhead R that occurs only at check time
is R=A-P+B-Q =A-T-N+B・(M-T-N). Therefore, the timer management overhead Y per unit time is Y= (S+E) ・N+C/D+R/U= (
S+E) ・N+C/D +(A-T-N+B・(M-T-N))/(L
=T).

Now, calculate the overhead by assigning the following values to each variable.

5=20 (step) E=5 (step) C=1000 (step) A=10 (step) B=5 (step) M=300 (task) D=1 (second) N=50 (transaction) T =5 (seconds) L=50 (seconds) The timer management overhead using the conventional method is x=47
50 (steps) The timer management overhead by Honjiteikan is Y = 230
0 (steps), and according to this embodiment, the timer management overhead is reduced to less than half.

〔Effect of the invention〕

According to the present invention, only one time counter is updated every time a timer interrupt occurs.

There is no need to update a counter for the number of running processes. Also, to check whether there is a process that has exceeded the elapsed time limit, when a timer interrupt occurs, first check whether the check time has been reached, and only when the check time has been reached. All processes need to be checked to see if the elapsed time limit has been exceeded, and there is no need to check all processes every time a timer interrupt occurs. In other words, the number of counters that are updated each time a timer interrupt occurs is reduced, and the number of times that all processes are checked to see if they have exceeded the elapsed time limit is reduced, making it easier to monitor process completion times. overhead is reduced.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the relationship between the process state and the table state according to an embodiment of the present invention, FIG. 2 is a block diagram of the online system of the embodiment, FIG. 3 is a block diagram of timer management, and FIG. Figure 5 shows the configuration of timer data, Figure 5 @ shows the configuration of the system table, Figure 6 shows the configuration of the task table, Figure 7 shows the flow of timer management processing at the start of business processing, and Figure 8 shows the end of business processing. FIG. 9 is a flow chart of timer task processing, and FIG. 10 is a time chart when the difference between the limit elapsed time and the actual processing time is smaller than the average processing occurrence interval. 1... Online controller, 2... Terminal device, 3... Database, 4
...Task, 5...Timer management. 6...Operating system. 51...Timer task, 52...Timer data, 5
21...System table, 522...Task table. 5211...Start task table address. 5212... Current time, 5213... Check time, 5221... Next task table address, 5
222...System table address, 5223...
Limit time. Kudzu! Figure map, 5゛volume map collection map, maple map collection map

Claims (1)

[Claims] 1. In the process of monitoring whether the elapsed time from startup to termination of each process of a computer system that processes multiple processes in parallel does not exceed a predetermined limit value, the limit time at which each process should be considered time-out , the earliest occurring limit time is set as the check time, the check time is compared with the current time at regular intervals, and only when the current time exceeds the check time, time-over treatment is performed for processes whose limit time is less than or equal to the current time. A process termination monitoring method characterized by reducing overhead by performing the following steps.