JP2509972B2 - Environment setting method - Google Patents

Description

DETAILED DESCRIPTION [Overview] Regarding an environment setting method that can set the number of IO buffers and block length of a data set to optimal values, environment settings such as the number of IO buffers and block length can be set reasonably for each job. For the purpose of enabling setting, in an information processing system that has an environment information storage unit that stores environment information at the time of execution of the job corresponding to the job, the number of IO buffers used when executing the job and the time during which the job is executed Statistical information acquisition unit that acquires statistical information including the actual IO count that has occurred and the actual value thereof, a statistical information file that stores the statistical information acquired by the statistical information acquisition unit, and the statistical information when the job is started. Based on the statistical information about the job recorded in the file and the environmental information about the job recorded in the environment information storage unit, An environment condition calculation unit that calculates the setting values of the number of IO buffers and block length to be assigned to the job, and an environment setting unit that sets the IO buffer and block length based on the calculated setting values Then, the statistical information on the running of the job is accumulated, and when the job is started, the environment is set based on the statistical information of the job and the environmental information of the job.

[Industrial applications]

The present invention relates to an environment setting method capable of setting the number of IO buffers and block length of a data set to optimum values.

[Conventional technology]

With the expansion of computer applications and the development of I / O devices with rich functions, I / O devices and storage media for data processing are treated as shared resources of the system independent of each processing program, and means for reading and writing are also provided. , The system needs to be managed collectively.
The programs that handle this are collectively called "data management" and include job management, task management, recovery management, etc.
It constitutes a control program and is included in the operating system (OS).

When performing some business using a computer system, reading and writing of data stored in a storage device such as a magnetic disk is frequently performed. Reading and writing of this data is performed via a memory area for temporarily storing data called an IO buffer.

That is, when reading data from a magnetic disk device, etc., the read data is temporarily stored in the IO buffer,
When data is transferred from the IO buffer to a predetermined memory area and written to a magnetic disk device or the like, the data is temporarily stored in the IO buffer and then transferred to the predetermined magnetic disk device.

Since such processing is performed under the management of the control program, the user of the computer system does not need to be particularly aware. However, it is necessary to properly set the number of IO buffers used by the business program to be executed and the block length described later.

If the number of IO buffers is too small, for example, if only one is set, the external storage device such as the magnetic disk device must be accessed each time reading and writing, which not only slows down the processing but also increases the computer system. Overall performance is reduced.

Further, there is an appropriate value for the amount of data processed by one read / write. Therefore, a predetermined number of records are collectively treated as one block as a processing unit for read / write. It is necessary to set this block length to an appropriate value.

Conventionally, the number of IO buffers and the block length are set manually, and the set values are used as environment values for VTOC.
It is recorded in the area that is composed of labels for controlling the dataset called (Volume catalog) and in the registry related to the dataset called Catalog, and the environment is set when the job is started based on these values. The Rukoto.

[Problems to be Solved by the Invention]

The above environment setting should be made in consideration of how to issue input / output instructions of the program, the structure of the data set, the number of data items, and the like.

However, under the present circumstances, there are separate program creators and environment configurators, etc., and there is no one who can grasp the requirements related to the operation of the entire system, so the number of IO buffers is set appropriately, and the block length also considers the number of processes. It is often set without. Therefore, the number of IO operations is increased more than necessary due to the improper number of buffers and block length, which often affects the performance.

An object of the present invention is to enable environment settings such as the number of IO buffers and block length to be set reasonably for each job.

[Means for solving the problem]

A statistical information acquisition unit 11, an environmental condition calculation unit 12, and an environmental setting unit 13 are provided in a data management unit 1 that performs basic control over data transfer between a standard input / output device and a main storage device. A statistical information file 14 for recording actual values such as the number of actual IOs generated during job execution is attached.

In the statistical information file 14, the job name and program name, the number of transactions, the data set classification, the access classification, the reference system actual IO count, the update macro issue count, the data set buffer count, etc. are recorded for job 5. is there.

When the job 5 is placed in the job space 3 on the main storage device and activated, the statistical information acquisition unit 11 is activated by the activation command.
Acquires the statistical information buffer 10 and prepares to acquire the statistical information while executing the started job.

On the other hand, the environmental condition calculation unit 12 reads the VTOC6, the environmental information of the data set to be used from the catalog 7, and various statistical information from the statistical information file 14, and calculates the appropriate values of the number of IO buffers 8 and the block length. To do. The environment setting unit 13 sets the IO buffer 8 in the job space 3 according to the calculated value, and also sets the DCB (data control block) 9
Set the block length to.

While the job 5'is running, the above-mentioned statistical information acquisition unit 11
Acquires the above various kinds of statistical information and writes them in the statistical information buffer 10. These statistical information are transferred and stored in the statistical information file 14 when the end command of the job 5'is issued.

[Work]

As described above, in the statistical information file, actual values such as the number of transactions that occurred during the running of the job, the number of reference system actual IOs, the number of update system macros issued, and the number of dataset buffers are recorded and accumulated for each job. There is.

In the present invention, the number of IO buffers and the block length are not arbitrarily determined, but the actual value is applied to a predetermined calculation formula when the job is started, and the number of IO buffers and the block length that are considered to be appropriate are calculated. Acquires the IO buffer according to the value and sets the block length. The actual performance of this job during running is also recorded in the statistical information file.

Since the above calculation formula is set so that the actual number of IOs can be reduced to a reasonable value, in the present invention in which the environment value is set based on the actual results for each job in this way,
The IO count can be set to a reasonable value, and the performance of the entire system improves.

〔Example〕

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

FIG. 2 is a diagram showing the operation of the data management unit 1 at the time of starting a job in this embodiment.

As shown in Fig. 1, DASD (Direct Access Volume)
When one job (job program) 5 is transferred from 4 to the job space 3 and activated as a job 5 '(see FIG. 2), the statistical information acquisition unit 11 first acquires the statistical information buffer 10 [second See figure).

Next, the environmental condition calculation unit 12 reads out various environmental information stored in each of the statistical information file 14, the catalog 7, and the VTOC 6. That is, from the statistical information file 14, with the job name or program name as a key,
Reads the corresponding data set classification, access classification, past transaction count, reference system actual IO count, update macro issue count, etc., and the IO buffer count set at that time. Also, from the catalog 7, the index level number, CI length, CA length, etc. of the data set to be used are shown. From VTOC6, the record length, block length of the above data set, and track length of the DASD storing the above data set are shown. Read-out (see FIG. 2) and then the number of IO buffers and block length to be set are calculated from the above-mentioned various read-out information (see FIG. 2).

Next, the environment setting unit 13 acquires the IO buffer 8 and sets the block length in the DCB 9 based on the calculated value [see FIG. 2].

In the above series of processing, the number of IO buffers and the block length are calculated as follows.

First, the number of IO buffers is set as an index buffer, which is a value obtained by adding 1 to the number of index levels in the case of forward processing, and 3 in the case of random processing.

The index level in the case of forward processing may be deepened by the division of the index key, so
In preparation for this, prepare one more buffer than the current index level.

Also, the records retrieved in random processing are random, so it is difficult to determine the required index. However, since the head of the index of the tree structure needs to be resident, a buffer for this and the next index is set. However, since the index may be divided in some cases, an extra one is provided in preparation for this, and there are three in total.

The number of dataset buffers depends on the virtual memory access method (VS
AM) forward processing, And, (CA length / CI length) …… Use the smaller one of the equation (2).

The numerator on the left side of the above equation (1) is the total number of reference system real IOs, and by dividing this by the number of transactions, the average number of IOs per transaction (however, this is an integer part, the fraction is Rounded down) is obtained. By adding 1 to this value, the number of buffers rounded up is calculated. The CA length is a value that indicates the amount of data that can be read and written at once in units of track length or cylinder length in VSAM (virtual memory access method), and CI length indicates the buffer length in VSAM. Therefore IO
The number of buffers cannot exceed the above CA length divided by CI length. Therefore, when the value of the expression (1) exceeds the value of the expression (2), the value is changed to the value and the value of the expression (2) is taken.

In the case of random processing in online processing, And This is a value obtained by rounding up the fractional number of macro updates issued per transaction. By preparing a buffer equivalent to the number of macros issued per transaction, the data required for processing one transaction Allows alignment on the buffer.

As described above, the number of buffers required for various processes is statistically averaged to reduce the actual number of IOs reasonably.

On the other hand, the block length is And [track length / record length] x record length The smaller one of the equations (5) is adopted.

The above formula (4) shows a value corresponding to the amount of data per transaction, and by setting the block length to this value, necessary data can be read at once.
However, since access to DASD cannot exceed one track at a time, when the value of equation (4) exceeds equation (5), the value of equation (5) is adopted.

The brackets used in the description of the above calculation method are Gauss symbols.

FIG. 3 is a diagram for explaining a method of acquiring statistical information in this embodiment.

After the job is started and the statistical information buffer 10 is acquired, the dataset is opened [see Figure 3] and READ.
Every time / REWRITE is performed, the statistical information acquisition unit 11 records various statistical data in the statistical information buffer 10. When the dataset is opened, the number of IO buffers, block length, access classification, and dataset classification of the dataset are recorded in the statistical information buffer [see FIG. 3].

That is, when the READ (see FIG. 3) of the data set is performed, the count value of the reference system IO number of the data set is recorded in the statistical information buffer 10 (see FIG. 3).

When REWRITE [see FIG. 3] is performed to update data, the number of times the update system macro is issued is counted and the count value is recorded in the statistical information buffer 10 [see FIG. 3].
To go.

Also, each time each transaction ends (see FIG. 3), the number of transactions generated is counted up and recorded in the statistical information buffer 10 (see FIG. 3).
To go.

The statistical information thus recorded in the statistical buffer 10 is
When the data set is closed (see Fig. 3), the statistics information buffer 10 is created using the job name and program name as keys.
Is transferred to the statistical information file 14 and stored as statistical information to be used when setting the number of IO buffers and the block length after that.

As described above, in this embodiment, a common statistical information file is created in the system, and the contents of the statistical information buffer 10 in which the results of running the job are recorded are written back to the statistical information file 14 at the end of the job.

Based on the statistical information accumulated in this way, it is possible to determine the required number of buffers and the appropriate block length for each job, so the number of IOs can be reduced to a reasonable value and the performance of the entire system can be reduced. Is improved.

Note that the above formulas for calculating the number of IO buffers and the block length are merely examples, and the present invention is not limited to these and may be arbitrarily set.

〔The invention's effect〕

As described above, according to the present invention, by optimizing the number of IO buffers and the block length, the number of IOs can be reduced and the performance of the entire system is improved.

[Brief description of drawings]

FIG. 1 is an explanatory view of the configuration of an embodiment of the present invention, FIG. 2 is an explanatory view of the environment setting processing of the above embodiment, and FIG. 3 is an explanatory drawing of statistical information acquisition processing of the above embodiment. In the figure, 1 is a data management unit, 2 and 4 are DASDs, 3
Is a job space, 5 is a job program, 5'is a job,
6 is VTOC, 7 is catalog, 8 is IO buffer, 9 is DCB, 1
Reference numeral 0 is a statistical information buffer, 11 is a statistical information acquisition unit, 12 is an environmental condition calculation unit, 13 is an environment setting unit, and 14 is a statistical information file.

Claims (1)

(57) [Claims] 1. An information processing system having an environment information storage unit for storing environment information at the time of execution of a job corresponding to a job, and the number of IO buffers used when executing the job and the actual IO generated during execution of the job. Statistical information acquisition unit (11) that acquires statistical information including the number of times and the related performance value
A statistical information file (14) for storing the statistical information acquired by the statistical information acquisition unit (11); statistical information about the job recorded in the statistical information file (14) when the job is started; An environmental condition calculation unit (12) that calculates setting values of the number of IO buffers and block lengths to be allocated to the job based on the environmental information about the job recorded in the environment information storage unit, and based on the calculated setting value. It has an IO buffer and an environment setting unit (13) that sets the block length, accumulates statistical information related to running of the job during job execution, and when the job is started, the statistical information of the job and the job The environment setting method is characterized in that the environment is set based on the environment information of.