JP2885907B2 - Online processing time analysis method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention relates to an online task processing time analysis method, and mainly relates to a response / time generated in an online system.
In order to elucidate the cause of abnormalities such as reduced turnaround,
The present invention relates to an online task processing time analysis method that enables actual measurement and analysis of necessary data such as task names and program names that have caused an increase in CPU processing time in an environment where online service has been continued.

[Conventional technology]

Conventionally, as a method of actually measuring and analyzing desired data in an environment in which an online service is continued, an online program correction method is known. This involves embedding a trap routine in each online program that temporarily collects information such as the start / end time of the program, and collecting and analyzing the information. When the analysis by the online program is completed, the online program is restored. It is a method to return to.

[Problem to be Solved by the Invention] In this online program correction method, replacement or return of the program is required for embedding or removing the information collection routine in the online program, and the online service must be interrupted each time. is there. Also, since embedding of information collection routines cannot be targeted at every part of the online program,
It is necessary to aim at the parts that are likely to be a problem due to experience many times, the time required to find the cause and operation are large, and with the embedding and removal of many information collection routines, online bugs due to the inclusion of program bugs There is also a problem that the quality tends to deteriorate. Furthermore, if the trap routine is left unreturned, information collection overhead will always occur during online operation, which will cause a decrease in the performance of the online system.

An object of the present invention is to solve the above-described problems of the conventional method, and it is possible to continue the operation of an online service by eliminating the need to modify an online program, and to provide a program for each program that runs in an online real environment. The system automatically collects specific information such as the actual running time of the car while minimizing the deterioration of online performance,
It is intended to deal with performance bottlenecks early.

[Means for solving the problem]

In order to achieve the above object, the present invention provides an online processing computer system in which control transfer between programs is performed by issuing / returning an SVC (Supervisor call) or issuing a RUN / STOP macro. And a task ID list table holding a task ID list to be measured and an SVC table holding an information collection routine address in addition to the SVC processing routine address for each SVC.
Is issued, the status flag indicates that performance is being measured, and the task ID list table indicates the task ID.
Exists, it is assumed that the SVC table is referred to, an information collection processing program is started based on the information collection routine address, and after the information collection processing, the SVC processing program is started based on the SVC processing routine address. And Then, as the information collection processing program, at least a program name, a task ID, a start event and a start time in the case of RUN, and an end event and an end time in the case of STOP corresponding to the SVC of the RUN / STOP macro are collected. An information collection processing program that is an item, an information collection processing program that has a task ID, an end event, and an end time as collection items corresponding to the SVC of the SRTN macro; and
For other SVCs, an SVC name, a task ID, a start event and an information collection processing program having a start time as a collection item are provided.From the information collected by the information collection program group, a net processing time of a RUN start program, It is characterized in that the net processing time of the SVC routine is obtained.

(Operation)

According to the present invention, when an SVC is issued from an online task, if the computer is measuring the online processing performance and the task ID to be measured is registered, the performance evaluation is automatically performed prior to the original SVC processing. Then, the process proceeds to the original SVC process, and otherwise performs the original SVC process directly.

In the information collection process, at least for the SVC of the RUN / STOP macro, the program name, task ID, RUN / STOP start /
Collect end event, start / end time, and
For the SVC, the task ID, the end event, and the end time are collected, and for the other SVCs, the SVC name, the task ID, the start event, and the start time are collected. This collected information record is categorized by task, and the RUN program or S
The net processing time of the VC routine is corrected to data that can be calculated, and these net processing times are obtained.

This enables the raw data (CPU
The program name, CPU execution time, etc. that caused the increase in processing can be obtained without modifying the online program, and the influence on online performance can be minimized.

〔Example〕

Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

FIG. 1 is an overall functional block diagram of an embodiment of the system of the present invention. Reference numeral 10 denotes an online task, and a unique task ID is set for each task. 11 is the SVC analysis processing unit, 12
Is an information collection routine activation monitoring unit, 13 is an SVC routine activation unit, 14 is a pre-editing processing unit (1) for classifying records by task, 15 is a pre-editing processing unit (2) for net processing time calculation, and 16 is an editing unit It is a processing unit. These processing units constitute part of the processing functions of the CPU in terms of hardware.

17 is an information collection processing program, and 18 is an SVC processing program. The information collection processing program 17 is prepared for each SVC, the details of which will be described later. SVC processing program 18
There are many in the system, some of which are resident on main storage, others are placed on external main storage (non-resident) and loaded into main storage as needed.

21 is a status flag indicating whether or not online processing performance is being measured, and 22 is a task holding a list of task IDs to be measured.
The ID list table 23 is an SVC table for managing resident / non-resident, SVC processing routine address, and the like for each SVC. These may use either a register group of the CPU or a main storage device. Reference numeral 31 denotes a buffer memory for temporarily storing collected information during online processing, and 32 denotes an information collection file for filing the collected information. Reference numeral 33 denotes an edit file (1) for storing task-specific classification results, reference numeral 34 denotes an edit file (2) for storing net processing time corrected results, and reference numeral 35 denotes an edit processing result output.

In FIG. 1, the part related to information collection in the online real environment is above the dashed-dotted line, and the part related to editing and analysis of the collected information is below. Hereinafter, each process will be described.

FIG. 2 shows a status flag 21, a task ID list table 22,
FIG. 4 is a detailed diagram of an SVC table 23. (A) is the status flag 21
In accordance with an instruction from an operator or an automatic operation program or the like, ON / OFF can be switched to the online performance measurement start time / start time, respectively. That is, "ON" indicates that performance is being measured (performance measurement is permitted), and "OFF" indicates that performance is not being measured (performance measurement is inhibited). (B) is a measurement target task ID list table 22, in which IDs (A, R, and the like) of the measurement target tasks specified by the operator, the automatic operation program, and the like are stored. Unregistered places are "NULL". (C) is an SVC table 23, in which the type of SVC is represented by an SVC number, and
A resident / non-resident bit, an SVC routine address indicating a pointer of the corresponding SVC processing program, and an information collection routine address indicating a pointer of the corresponding information collection processing program are registered for each number.

When an SVC is issued from online task 10, the SVC
The analysis processing unit 11 analyzes the SVC, and passes the task ID set for the task and the type (SVC number) of the SVC to the information collection routine activation monitoring unit 12. The information collection routine activation monitoring unit 12 refers to the status flag 21 and the task ID list table 22, determines that the status flag 21 is NO (performance measurement allowed), and
If the task ID is stored in the task ID list table 22, the information collection routine address is searched from the SVC table 23 using the SVC number as a key, and the information collection processing program 17 is started. Then, after the information collection processing by the information collection processing program 17, the control is passed to the SVC routine activation unit 13. On the other hand, when the status flag 21 is OFF (performance measurement suppression is disabled), or when the task ID is not stored in the task ID list table 22 even if the status flag 21 is ON, the information collection routine activation monitoring unit 12 collects information. Processing program
The control is immediately passed to the SVC routine activating unit 13 without activating the 17. The SVC routine starting unit 13 searches the SVC table 23 for an SVC routine address using the SVC number as a key, and starts the SVC processing program 18. SVC processing program
18 executes the SVC process (if the SVC is nonresident, execute it after loading the SVC), and returns to the SVC issuing source.

Next, a specific example will be described with reference to FIGS.

FIG. 3 shows the case of online normal operation, and the state flag 21 is OFF. In this case, when an SVC (SVCα) is issued from the online task (task ID = A) (), an SVC routine address from the SVC table 23 is obtained (1000 in this example), and The processing program is started (), the original processing of the SVCα is executed (), and when the processing ends, the process returns to the source of the SVCα ().

FIG. 4 shows a case where the online performance is being measured, and the status flag 21 is ON. Here, the task ID list table 22
, ID = A is stored as the task ID to be measured. From task ID = A online task to SVCα
Is issued (), before the original SVCα processing, first, an information collection routine address is obtained from the SVC table 23 (5000 in this example), and the information collection processing program 17 of the SVCα is started (), The information necessary for the performance evaluation of the SVCα is collected (). The collected information is stored in the buffer memory 31 one after another, and then read out to the information collection file 32. When the information collection is completed (), the SVC routine address is obtained from the SVC table 23 (100 in this example).
0), starts the processing program of the SVCα (),
The original process of α is executed (), and the process returns to the SVCα issue destination (). In addition, like task ID = B,
When an SVC is issued from an online task that is not stored in the task ID list table 22, the status flag 21 is set to O.
Even if N, the original processing of the SVC is directly executed.

Next, the collected information will be described in detail. In the present invention, the information collection processing program 17 includes:
(SRTN-SVC) (2) SVC that starts a RUN program (RUN-SVC) (3) SVC that ends a RUN program (STOP-SVC) (4) Four types of general SVCs other than the above (SVCα, SVCβ, etc.) Are prepared for each of the SVCs, and at least the information shown in Table 1 is collected each time these SVCs run once.

That is, in each information collection processing program (routine No.2, 3) corresponding to SVC of RUN / STOP macro, program name, task ID, start event and start time for RUN, and end for STOP The event and end time are the minimum collection items. In the information collection processing program (routine No. 1) corresponding to the SVC of the STRN macro, the task ID, the end event, and the end time are the minimum collection items. further,
For other SVCs (routine No. 4), the SVC name, task ID, start event, and start time are the minimum collection items.

Collected Information Editing Process As described above, the information collected by each information collection processing program 17 is stored as a record in the information collection file 32 via the buffer memory 31. The pre-editing processing unit (1) 14 classifies the collected records of the information collection file 32 into files (record groups) for each task ID, and accumulates them in the edited file (1) 33. In this case, in the file of each task ID, the collected records are simply arranged in chronological order.

The pre-editing processing unit (2) 15 corrects the record of the file of each task ID in the editing file (1) 33 as follows so that the editing processing unit 16 can calculate the net processing time. . And each task after correction
Edit file (record group) for each ID file (2) 34
To accumulate.

(1) The SVC routine end information record of SRTN forming a pair with RUN is changed to RUN program start information, and the RUN program start information record is deleted (see FIG. 5 (a)).

(2) Delete the SVC routine end information record of the SRTN paired with STOP (see FIG. 5 (b)).

(3) In the remaining SVC routine end information record, a corresponding SVC name is obtained and added thereto (see FIG. 5 (c)).

The processing procedure in the pre-editing processing unit (2) 15 will be described below.

Set the read position at the beginning of the file with a certain task ID.

The SVC routine end information record is searched for in the forward direction, and if there is an SVC routine end information record, its position is recorded and the record is read in the reverse direction. If there are no more records to read, the process is terminated.

When the RUN program start information record is read, after storing the program name and event information from it, delete the record, return to the position of the SVC routine end information record, and overwrite this record with this storage information Then, the processing is continued from the procedure of the above.

When reading the RUN program end information record,
Return to the position of the SVC routine end information record, delete the record, and continue the processing from the procedure.

When reading a record that is an SVC routine start information record and does not have “*” (indicating the completion of association with SRTN) “Done” in Table 1, “Done” is recorded in that record, and the SVC name is recorded from that record. Read. Further, the process returns to the SVC routine end information record position, and the SVC routine
After inserting the first name, the process is continued from the step.

In other cases, the record is skipped, the next record is read in the reverse direction, and the process is continued from the above procedure.

The edit processing unit 16 calculates the net processing time of the RUN startup program (RUN PM) and the net processing time of the SVC routine based on each file (corrected record group for each task) of the edited file (2) 34. , And an edit result output 35 is obtained. RU
N The processing time of the startup program is
It is obtained as the difference between the time of the start information record of the program and the time of the latest end information record following the record. Similarly, the processing information of the SVC routine is the SV of the same SVC name.
It is obtained as the difference between the time of the C routine start information record and the time of the immediately following end information record following the record.

FIG. 6 shows a specific example of the processing. In the figure, is the program
PM-A measured task runs, the execution routine of SVCarufa ₁ this task issued, the SVCarufa ₁ execution routine is further issued by SVCarufa ₂ execution routine, the information collection processing program started by a trigger to issue the SVC The group is the recorded contents of the information records collected by the information collection processing program group, respectively, is the pre-editing process 1 for classifying the collected information record group by task and the processing result (task classification result file group). Further processing results
The pre-editing process 2 that corrects the net processing time of the RUN program or SVC routine to a form that can be calculated, and the processing result (net time corrected file group) are an example of an editing / analysis report as an editing result output.

FIG. 6 shows the information collection processing program group.
As the collected information record of task ID = A, PM-
Records from the RUN start SVC of A to the end of the STOP-SVC of the same PM are collected. In this, since records of other tasks are mixed, the records are classified according to the tasks in the pre-editing process 1 and are filed as shown in FIG.
Further, in the pre-editing processing 2, the file is corrected so that the net processing time of the RUN program or the SVC routine can be calculated, and the resulting file group is obtained. By performing various editing processes based on this, as shown in (1), analysis information of the program nest time and PM / SVC list information requiring a long time can be obtained.

In the example of FIG. 6, each information collection processing routine is executed by the OS.
It is also possible to collect information other than this example by configuring a program with the same authority as above and adding necessary functions.

〔The invention's effect〕

As described above, according to the present invention, specific online raw data (a program name causing an increase in CPU processing, CPU execution time, etc.) capable of elucidating a direct cause of an online performance bottleneck can be obtained without modifying the online program. As a result, the required operation for investigating the cause of the performance bottleneck related to the task processing time is greatly reduced,
The early solution is achieved.

In addition, collection of performance analysis information is limited to specified tasks,
By performing the process only for the designated period, the influence on the online performance can be minimized. The effect on online performance can be further reduced by performing editing and analysis after information collection using only an offline system.

In the future, with the advancement of online system operations, hardware and software tend to become larger and more complex, and in such a situation, it is becoming increasingly important to provide a means for quickly and easily investigating system performance problems. It is. The present invention is positioned as a new method corresponding to this.

[Brief description of the drawings]

FIG. 1 is an overall functional block diagram of an embodiment of the system of the present invention, FIG. 2 is a detailed diagram of a status flag, a task ID list table and an SVC table, and FIG. FIG. 4 is a diagram for explaining processing in the case of online performance evaluation information collection, FIG. 5 is a diagram for explaining correction processing of the classification result for each task, and FIG. 6 is a diagram showing a specific processing example. 11: SVC analysis processing unit, 12: Information collection routine activation monitoring unit, 13: SVC routine activation unit, 14, 15: Pre-editing processing unit, 16: Editing processing unit, 17: Information collection processing program , 18: SVC processing program, 21: Status flag, 22: Task ID list table, 23: SVC table, 32: Information collection file, 33, 34: Edit file.

Continuation of front page (56) References JP-A-61-58055 (JP, A) JP-A-1-48133 (JP, A) JP-A-4-76640 (JP, A) (58) Fields investigated (Int) .Cl. ⁶ , DB name) G06F 11/28-11/34

Claims (1)

(57) [Claims] A control transfer between programs is performed by SVC (Super-vi
An online processing computer system that issues / returns a sor call) or issues a RUN / STOP macro. A task that holds a status flag indicating whether online processing performance is being measured and a task ID list to be measured It has an ID list table, an SVC table that holds an information collection routine address in addition to the SVC processing routine address for each SVC, and when an SVC is issued from an online task, the status flag indicates that performance is being measured, and If the task ID is present in the task ID list table, the SVC table is referred to, an information collection processing program is started based on the information collection routine address, and after the information collection processing is completed, the SVC processing is performed based on the SVC processing routine address. In the computer system that starts up the processing program of And RUN / ST
The program name, task ID,
In addition, the task I corresponds to the information collection processing program that collects the start event and start time in the case of RUN, the end event and the end time in the case of STOP, and the SVC of the SRTN macro.
D, Information collection processing program that uses end event and end time as collection items, and SVC for other SVC
An information collection processing program that has a name, a task ID, a start event, and a start time as collection items.
An online task processing time analysis method, wherein a net processing time of an N startup program and a net processing time of an SVC routine are obtained.