JP2006172065A - Checkpoint collection method, system and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a technique capable of reducing a user's burden of designating a condition for determining a timing for collecting checkpoints in a system resource usage state and describing a checkpoint collection command in a job.
In a checkpoint collection method for acquiring checkpoint information for restarting a job when a failure occurs during job execution, information indicating a system resource usage status for each job is acquired and stored in a storage device. The storage step, the conditions for determining whether or not checkpoint collection is necessary according to the system resource usage status for each job, and the stored system resource usage status information are read from the storage device to checkpoint collection. A step of determining necessity and a step of collecting checkpoints by storing checkpoint information in a storage device based on the determination result.
[Selection] Figure 1

Description

  The present invention relates to a checkpoint collection technique for acquiring checkpoint information during job execution for restarting a job when a failure occurs.

  In the checkpoint collection method that acquires checkpoint information for restarting a job when a failure occurs during job execution, there is a method of collecting job checkpoints periodically at certain time intervals. (For example, refer nonpatent literature 1). In addition, there is a method of determining whether or not to collect a checkpoint according to a calculation result with respect to a job execution time and a checkpoint collection time (see, for example, Patent Document 1).

JP 2004-94422 A E.N.Elnozahy, Lorenzo Alvisi, Yi-Min Wang, D.B.Johnson, `` A Survey of Rollback-Recovery Protocols in Message-Passing Systems, Technical Report CMU-CS-96-181, Department of Computer Science, Carnegie Melon University, 1996)

  Since any of the above checkpoint collection methods used the actual time as the checkpoint collection condition based on the job execution time or checkpoint collection time, it was not possible to evaluate how much the job was actually processed, and the load status of the computer system In some cases, appropriate checkpoints were not taken. In addition, it is necessary to describe an instruction for collecting checkpoints in the job program, which is a burden when creating the job program.

  The object of the present invention is to solve the above-mentioned problems, specify the conditions for determining the timing for collecting checkpoints in the usage status of system resources, and reduce the user's burden of writing a checkpoint collection command in the job. Is to provide new technology.

  Another object of the present invention is to provide a technique capable of changing checkpoint collection conditions in accordance with the use status of system resources of a job being executed.

  The present invention collects checkpoints in accordance with the use status of system resources in a computer system that acquires during execution of a job checkpoint information for restarting a job when a failure occurs.

  In the present invention, first, an input of a condition for determining whether or not it is necessary to collect a checkpoint according to the usage status of system resources for each job is received and stored in a storage device. At this time, as the condition, for example, an input of a conditional expression indicating a CPU usage accumulated time, a threshold value of the accumulated amount of input / output data, or a composite conditional expression obtained by combining these conditional expressions is received and stored in the storage device. .

  Next, when the job is executed, information indicating the usage status of the system resource for each job is acquired and stored in the storage device. At that time, information on the accumulated CPU usage time and the accumulated input / output data, which are the elements of the stored conditions, is acquired and stored in the storage device, and the information indicating the usage status of the system resources is updated.

  Then, after updating the information indicating the use status of the system resource as described above, the stored condition and the updated information on the use status of the system resource are read from the storage device to determine whether or not it is necessary to collect a checkpoint. Based on the determination result, the checkpoint information is stored in the storage device and the checkpoint is collected. That is, a logical calculation is performed by substituting the updated CPU usage cumulative time and input / output data cumulative amount into the conditional expression, and the CPU usage cumulative time and input / output data cumulative amount is a condition for collecting checkpoints. If the condition is satisfied, the checkpoint information of the job is stored in the storage device.

  According to the present invention, it is possible to reduce a user's burden of designating a condition for determining the timing of collecting checkpoints by using system resources and describing a checkpoint collection command in a job.

(Embodiment 1)
A computer system according to the first embodiment that acquires checkpoint information for restarting a job when a failure occurs according to the use status of system resources will be described below.

  FIG. 1 is a diagram showing a schematic configuration of a computer system 1 of the present embodiment. As shown in FIG. 1, the computer system 1 according to the present embodiment includes a job execution management unit 20, a checkpoint collection execution unit 21, a scheduler 22, a CPU monitoring unit 23, and an I / O monitoring unit 24. ing.

  The job execution management unit 20 is a processing unit that manages the execution of jobs by controlling the operations of the checkpoint collection execution unit 21, the scheduler 22, the CPU monitoring unit 23, and the I / O monitoring unit 24. The checkpoint collection execution unit 21 includes a condition table 12 and a compound conditional expression 13 that indicate conditions for determining whether or not checkpoint collection is necessary according to the system resource usage status for each job, and system resource usage status information. Is a processing unit that reads out the resource status table 14 indicating whether or not checkpoints are necessary, stores checkpoint information in the magnetic disk device based on the determination result, and collects checkpoints.

  The scheduler 22 is a processing unit that performs job state transition from the executable state job queue 3 to the execution state job queue 4. The CPU monitoring unit 23 is a processing unit that monitors and records the CPU time allocated to each job, and acquires CPU usage status information such as CPU usage accumulated time as information indicating the usage status of system resources for each job. The resource monitoring unit stores the information in the resource status table 14 in the memory.

  The I / O monitoring unit 24 is a processing unit that monitors and records the data amount of I / O processing performed by each job. As information indicating the usage status of system resources for each job, the input / output data cumulative amount, etc. This is a resource monitoring unit that acquires the usage status information of the I / O device and stores it in the resource status table 14 in the memory.

  Programs for causing the computer system 1 to function as the job execution management unit 20, the checkpoint collection execution unit 21, the scheduler 22, the CPU monitoring unit 23, and the I / O monitoring unit 24 are recorded on a recording medium such as a CD-ROM and magnetically recorded. It is assumed that after being stored in a disk or the like, it is loaded into a memory and executed. The recording medium for recording the program may be a recording medium other than the CD-ROM. The program may be used by installing it from the recording medium into the information processing apparatus, or the program may be used by accessing the recording medium through a network.

  As shown in FIG. 1, the computer system 1 of this embodiment includes a job execution management unit 20, an executable job queue 3, and an execution job queue 4. The job execution management unit 20 performs check according to function. The point collection execution unit 21, the scheduler 22, the CPU monitoring unit 23, and the I / O monitoring unit 24 are configured.

  The computer system 1 has a resource management mechanism for appropriately distributing resources to jobs. In this embodiment, the CPU monitoring unit 23 and the I / O monitoring unit 24 correspond to the mechanism, and the CPU monitoring unit 23 monitors and records the CPU time allocated to each job, and the I / O monitoring unit 24 monitors and records the data amount of I / O processing performed by each job.

  The executable state job queue 3 and the execution state job queue 4 manage job information 10-1, 10-2,..., 10-M, respectively. Transition between the queues of the job information 10 is performed by the scheduler 22. The job information 10 includes a job ID 11, a condition table 12, a compound conditional expression 13, and a resource status table 14.

  The user 5 inputs the job and the checkpoint collection condition associated therewith to the job execution management unit 20, and the job execution management unit 20 that receives the job generates job information 10 in the executable job queue 3 and collects the checkpoint. The condition table 12 and the compound conditional expression 13 are initialized according to the conditions.

  The information in the resource status table 14 is updated by the CPU monitoring unit 23 and the I / O monitoring unit 24 after the job information 10 is transitioned from the executable state job queue 3 to the execution state job queue 4.

  FIG. 10 is a diagram illustrating a hardware configuration of the computer system 1 according to the present embodiment. As shown in FIG. 10, the computer system 1 of this embodiment includes a job execution management unit 20, an executable state job queue 3, and an execution state job queue 4 on a memory, and stores in a memory, a magnetic disk device, or the like. Various processes are performed using a device or a communication device.

Next, the resource status table 14 will be described with reference to FIG.
FIG. 2 is a diagram showing an example of the resource status table 14 held by the job information 10 of this embodiment. The resource status table 14 has the purpose of recording the resource usage status in the job that holds it. The resource usage status items recorded in this embodiment are CPU usage accumulated time 14-01, job actual time 14-02, I / O write accumulated amount 14-03, and I / O read accumulated amount 14-04, respectively. There are four types, “time when CPU is assigned to the job”, “actual time elapsed since the job was submitted”, “cumulative amount when writing to I / O device”, “I / O” This means the cumulative amount when reading is performed on the device.

Next, an example of the condition table 12 and the compound conditional expression 13 will be described with reference to FIGS.
FIG. 3 is a diagram showing an example of the condition table 12 held by the job information 10 of this embodiment. The condition table 12 in FIG. 3 holds a plurality of conditional expressions 12-01, 12-02,..., 12-N for determining whether or not checkpoint collection is necessary.

  Each conditional expression in the condition table 12 stores the four items described in FIG. 2 and the corresponding threshold values as elements used for condition determination. For example, the conditional expression 12-01 with condition number 1 is stored. Indicates a condition of “true when the CPU usage cumulative time exceeds 1000 [seconds]”. The number of conditional expressions in the condition table 12 is not limited to four. There is a compound conditional expression 13 for the purpose of combining these conditional expressions, which is expressed as shown in FIG.

  FIG. 4 is a diagram illustrating an example of the compound conditional expression 13 held in the job information 10 according to the present embodiment. In FIG. 4, the numerical values in the compound conditional expression 13 mean that conditional expressions 12-01, 12-02,..., 12-N associated with the condition numbers in the condition table 12 are combined by an operator. The compound conditional expression 13 is composed of the logical sum (+) and logical product (x) of operators, the negation operator (￣), and parentheses.

  As an example, when a job that generates a large amount of data is executed, since the I / O write process is dominant in the job, the compound conditional expression 13− is related to the accumulated I / O write amount. By specifying 1, checkpoints can be collected at the checkpoint collection timing that matches the progress of the entire job.

  When the compound conditional expression 13-1 derived from FIGS. 3 and 4 is interpreted, “when the CPU usage cumulative time exceeds 1000 [seconds] and the CPU usage cumulative time does not exceed 10000 [seconds], or I / O It means “true” when the cumulative amount of writing exceeds 50000 [Byte].

  As another example, considering the case of executing a job that communicates via a network, a delay or waiting time occurs depending on the use status of the communication path in the job, so that the complex condition is related to the job execution time. When Expression 13-2 is designated, checkpoints can be collected not only at checkpoint collection points that match the accumulated CPU usage time but also at checkpoint collection timings at regular time intervals.

  When the compound conditional expression 13-2 derived from FIGS. 3 and 4 is interpreted, “when the CPU usage cumulative time exceeds 1000 [seconds] and the CPU usage cumulative time does not exceed 10000 [seconds], or the actual job time Is true when the value exceeds 50000 [seconds].

  Next, referring to FIG. 5, the job information 10 is updated by the CPU monitoring unit 23 and the I / O monitoring unit 24 in the course of a job transition from the executable state job queue 3 to the execution state job queue 4. Thereafter, the flow from the checkpoint collection execution unit 21 to the checkpoint collection execution after obtaining the checkpoint collection necessity determination will be described.

  FIG. 5 is a flowchart showing a processing procedure of checkpoint collection processing according to the present embodiment. The process of FIG. 5 starts from the point where the scheduler 22 of the job execution management unit 20 changes the job from the executable job queue 3 to the executable job queue 4 to the executable job (step 500).

  The I / O monitoring unit 24 monitors the I / O processing of the job that has transitioned to the executable state, acquires information on the write data amount and the read data amount of the job, and calculates the accumulated amount of the data. The job information 10 of the job is accessed, and the I / O write accumulation amount 14-03 and the I / O read accumulation amount 14-04 in the resource status table 14 are updated (step 501). Since the I / O monitoring unit 24 records the total amount of I / O processing data performed by the job, using the job ID 11 as an index, the I / O write accumulated amount 14-03 and I / O in the specific job information 10 The O reading accumulated amount 14-04 can be updated.

  Next, the CPU monitoring unit 23 obtains information on the CPU usage time and job execution time of the job that has transitioned to the executable state, calculates the accumulated amount of the data, and accesses the job information 10 of the job. The CPU usage accumulated time 14-01 and the job actual time 14-02 in the resource status table 14 are updated (step 502). Since the CPU monitoring unit 23 records the total CPU usage time used by the job, using the job ID 11 of the job information 10 as an index, the CPU usage accumulated time 14-01 in the specific job information 10 and the job actual time 14- 02 can be updated.

  Next, the checkpoint collection execution unit 21 reads out from the memory the conditional expression and compound conditional expression 13 in the condition table 12 and the system resource usage status information in the resource status table 14 to determine whether or not checkpoint collection is necessary. Determination is made (step 503).

  That is, the checkpoint collection execution unit 21 analyzes the condition table 12 and the compound conditional expression 13 held in the job information 10, and reads out the condition element of the condition number used in the compound conditional expression 13 from the condition table 12. After that, the current value of the resource status item corresponding to the condition element is read from the resource status table 14. Then, a logical expression for determining the current value with the conditional expression in the condition table 12 is generated and substituted into the composite conditional expression 13, and checkpoints are collected according to the determination result of the composite conditional expression 13 (step 504). ).

  After the checkpoint is collected, the resource status table 14 is all initialized to 0 so that the checkpoint collection condition is not unreasonably satisfied at the next condition determination (step 505).

  In general, in a computer system, when a job transitions to an execution state, CPU usage time is allocated from the job execution management unit 20, and the job is executed in units of the time. For this reason, in this embodiment, the job is not executed for a predetermined time or longer without updating the resource status table 14 for a specific job, and it is guaranteed that the resource status table 14 is updated at predetermined time intervals.

  Next, in the present embodiment, an example of a method for simultaneously inputting checkpoint collection conditions when a job is submitted will be described.

  FIG. 6 is a diagram illustrating an example of a command for inputting a job and checkpoint collection conditions according to the present embodiment. FIG. 6 shows that the user 5 can specify a condition by inputting a command 606 of a character string expression such as a set of conditional expressions 608 and a composite conditional expression 609 from the input device.

  The first argument of the command in FIG. 6 represents a set of conditional expressions 608 to be stored in the condition table 12. Each conditional expression 601 to 604 is delimited by “,”, and the resource status item and the threshold are further delimited by “:”. Condition numbers are assigned in order from the left of the character string of the conditional expression, such as 1, 2,... Further, the second argument of the command is a character string representing the compound conditional expression 609, and the third argument is the job program execution file name 610.

  At this time, the job program created by the user 5 on the premise that the condition element type in the condition table 12, the compound conditional expression and the description method of the conditional expression, and the job input command 607 are disclosed in advance in a description such as a manual. The designation is realized by giving the execution file name 610 as an argument of the job submission command.

  As described above, according to the computer system of the present embodiment, checkpoints are collected according to the usage status of system resources. Therefore, the conditions for determining the timing for collecting checkpoints are specified by the usage status of system resources. It is possible to reduce the user's burden of writing a checkpoint collection command in the job.

(Embodiment 2)
A computer system according to a second embodiment will be described below that accepts the conditions of a designated job again after the acceptance of the conditions for determining whether or not checkpoint collection is necessary and updates the conditions.

  In this embodiment, a condition for determining whether or not it is necessary to collect a checkpoint is already given to a job, and the user 5 can re-enter the condition even when the job is being executed.

  FIG. 7 is a diagram showing an example of the job information list display and display command of the present embodiment. When re-entering conditions, the user 5 must first be able to specify for which job the conditions are to be updated. Therefore, in the present embodiment, the user 5 is provided with a command command that can indicate to the output device the job information 10 that currently exists in the computer system 1, the job ID 11 that uniquely specifies the job information 10, and the checkpoint collection conditions that accompany it. .

  When this command is issued, as shown in an example of a job information list display example 701 in FIG. 7, the job execution management unit 20 according to the present embodiment includes information on the job ID 11 specifying the job information 10 and information on the condition table 12, The compound conditional expression 13 is read from the memory and displayed on the output device. Each row of the job information example 703 corresponds to the job information 10, and the job information display item 702 represents the description of the display items of the job information 10.

  In response to this result, the user 5 gives an argument to the condition change command 707 to change the checkpoint collection condition. The arguments are a set of conditional expressions 708, a compound conditional expression 709 to be overwritten, and a target job ID 711, respectively. At this time, the set of conditional expressions 708 includes conditional expressions 704 and 705. Further, as in the last line of the job information example 703, it is possible to execute the job without the checkpoint collection condition.

  By inputting the command 706 as described above, the checkpoint collection condition for a specific job can be changed.

  FIG. 8 is a flowchart showing a processing procedure of job information list display processing according to the present embodiment. The processing in FIG. 8 will be described by taking the operation when the command 606 in FIG. 6 is received as an example.

  First, information on the job ID 11, the condition table 12, the compound conditional expression 13, and the resource status table 14 is acquired from the job information 10 stored in the executable job queue 3 (step 801).

  Next, information on the job ID 11, the condition table 12, the compound conditional expression 13, and the resource status table 14 is acquired from the job information 10 stored in the execution state job queue 4 (step 802). Then, the data acquired in step 802 is displayed on the output device (step 803).

  FIG. 9 is a flowchart showing a processing procedure for updating the condition table 12 and the compound conditional expression 13 of the job information 10 according to this embodiment. The processing in FIG. 9 will be described as an example when the command 706 in FIG. 7 is received.

  First, referring to the job information 10 in the memory, it is determined whether or not there is a job having an ID specified by the user 5 (step 901). If the job does not exist, that fact is displayed. The process ends without doing anything (step 902).

  If there is a job having an ID specified by the user 5, the character string of the conditional expression specified by the user 5 is compared with a predetermined grammar rule to check whether the conditional expression is incomplete (step 903). If there is a defect in the conditional expression, that fact is displayed, and thereafter, nothing is done and the process ends (step 904).

  If there is no defect in the conditional expression, the specified content is stored and updated in the condition table 12 and the compound conditional expression 13 of the job information 10 in the memory for the specified job (step 905). Then, the job information 10 for the updated job is displayed as in the job information list display example 701 in FIG. 7 (step 906).

  As described above, according to the computer system of this embodiment, the conditions for determining whether or not checkpoints need to be collected are accepted again and the conditions are updated. Therefore, depending on the system resource usage status of the job being executed The checkpoint collection conditions can be changed.

It is a figure which shows schematic structure of the computer system of Embodiment 1. FIG. It is a figure which shows an example of the resource status table 14 which the job information 10 of Embodiment 1 hold | maintains. It is a figure which shows an example of the condition table 12 which the job information 10 of Embodiment 1 hold | maintains. 6 is a diagram illustrating an example of a compound conditional expression 13 held by job information 10 according to the first embodiment. FIG. 6 is a flowchart illustrating a processing procedure of checkpoint collection processing according to the first embodiment. 6 is a diagram illustrating an example of a command for inputting a job and checkpoint collection conditions according to the first embodiment. FIG. FIG. 10 is a diagram illustrating an example of a job information list display and a display command according to the second embodiment. 10 is a flowchart illustrating a processing procedure of job information list display processing according to the second embodiment. 10 is a flowchart illustrating a processing procedure for updating a condition table 12 and a compound conditional expression 13 of job information 10 according to the second embodiment. 2 is a diagram illustrating a hardware configuration of a computer system 1 according to the first embodiment. FIG.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 ... Computer system, 3 ... Executable state job queue, 4 ... Execution state job queue, 5 ... User, 10 ... Job information, 11 ... Job ID, 12 ... Condition table, 13 ... Compound conditional expression, 14 ... Resource status table 20 ... job execution management unit, 21 ... checkpoint collection execution unit, 22 ... scheduler, 23 ... CPU monitoring unit, 24 ... I / O monitoring unit, 14-01 ... accumulated CPU usage time, 14-02 ... job real time 14-03 ... I / O write accumulation amount, 14-04 ... I / O read accumulation amount, 12-01 to 12-N ... Conditional expression, 601 to 604 ... Conditional expression, 606 ... Command, 607 ... Job input command 608 ... A set of conditional expressions 609 ... Compound conditional expression 610 ... Job program execution file name 701 ... Job information list display example 702 ... Job information display Eyes, 703 ... job information example, 704 and 705 ... condition, 706 ... command 707 ... condition change command, a set of 708 ... condition, 709 ... complex conditional expression 711 ... job ID.

Claims (10)

In the checkpoint collection method for acquiring checkpoint information for restarting a job when a failure occurs during job execution,
Acquiring information indicating the usage status of system resources for each job and storing it in a storage device; conditions for determining whether or not to collect checkpoints according to the usage status of system resources for each job; Reading system resource usage information from the storage device and determining whether or not it is necessary to collect checkpoints; storing checkpoint information in the storage device based on the determination result; and collecting checkpoints; A checkpoint collection method characterized by comprising:   2. The compound conditional expression obtained by combining a plurality of conditional expressions for determining whether or not checkpoint collection is necessary is used as the condition for determining whether or not checkpoint collection is necessary. Checkpoint collection method.   The checkpoint collection method according to claim 1, wherein a condition for determining whether or not it is necessary to collect a checkpoint for the job is accepted when the job is submitted.   4. The method according to claim 1, wherein after completion of reception of the condition for determining whether or not the checkpoint is necessary, the condition of the designated job is received again and the condition is updated. Checkpoint collection method described in the section.   The checkpoint collection method according to any one of claims 1 to 4, wherein the accumulated usage time of the CPU is acquired as information indicating the usage status of the system resource.   6. The checkpoint collection method according to claim 1, wherein a cumulative amount of accessed input / output data is acquired as information indicating a use status of the system resource. In a computer system that acquires checkpoint information for restarting a job when a failure occurs during job execution,
A resource monitoring unit that acquires information indicating the use status of system resources for each job and stores the information in a storage device; a condition for determining whether or not checkpoint collection is necessary according to the use status of system resources for each job; Checkpoints that read the stored system resource usage information from the storage device, determine whether or not checkpoint collection is necessary, and store the checkpoint information in the storage device based on the determination result to collect the checkpoint A computer system comprising a collection execution unit.   8. The computer system according to claim 7, wherein after completion of acceptance of a condition for determining whether or not to collect the checkpoint, the condition of the designated job is accepted again and the condition is updated. In a program that causes a computer to execute a checkpoint collection method that acquires checkpoint information for restarting a job when a failure occurs during job execution.
Acquiring information indicating the usage status of system resources for each job and storing it in a storage device; conditions for determining whether or not to collect checkpoints according to the usage status of system resources for each job; Reading system resource usage information from the storage device and determining whether or not it is necessary to collect checkpoints; storing checkpoint information in the storage device based on the determination result; and collecting checkpoints; A program that causes a computer to execute.   10. The program according to claim 9, wherein after completion of acceptance of a condition for determining whether or not to collect the checkpoint, the condition of the designated job is accepted again and the condition is updated.

Images