JPS61235957A - Data control system - Google Patents

Abstract

PURPOSE:To decrease the pre-processing time for data input/output by storing read information to a buffer pool of a data page while being rewritten statistically at each execution of a program, using it and forecasting the next data page to read the page to the buffer tool. CONSTITUTION:A data page utilizing frequency Ci is stored in a statistical information area 21 to a data Di. That is, the number of times of a corresponding data page read and used in the buffer pool so far at present is used, the number of times is counted by a statistical information collection program 9 at each occasion and revised and stored in a corresponding area 21a as the newest count data. A next page forecast program 8 compares number of times of a statistical information area 21a of each data page 20, forecasts the statistical data page corresponding to the largest number of times as the data requested next by the user program 6 and the result is read to the buffer pool 10.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a data control method, and more particularly to a data control method for reading data on a secondary storage device into a buffer pool of a main storage device as needed.

Conventional technology Conventionally, such data page control methods read the necessary data page into the buffer pool after the target data page requested by the program is actually needed, or store unnecessary pages in secondary storage. It is designed to output to the device.

The buffer control method in this case is FIFO (Fi
rst 1nfirst out), L I F
This is done using O (Last in first out) or a method specific to the program.

In the conventional page input/output control method described above, the system control method is always FIF even if the user is doing routine work.
O, LIFO or a method specific to that program is executed, and if it is suitable for the business, page control can be performed efficiently, but if it is not suitable, page control cannot be performed efficiently. Furthermore, since the user has to start inputting the next page to be read from the moment the desired page is needed, it is not possible to read the next required page in advance.

An object of the present invention is to statistically predict the next data page that an execution program will request, thereby reducing the preprocessing time for data input/output. The purpose is to provide a data control method.

■1 and 1 The data control method according to the present invention is a data control method for reading data on a secondary storage device into a buffer pool of a main storage device as needed. A read statistics information area corresponding to each data page is added to the data page, and information indicating the manner in which the corresponding data page is read into the buffer pool is stored in this information area while being statistically rewritten each time the program is executed. This information is used to predict the next data page that the program will request, and this predicted data page is read into the buffer pool.

Statistical information indicating the manner of reading into the buffer pool includes, for example, usage frequency information on how often the data page corresponding to the information area is read into the buffer pool, and information on how often the data page corresponding to the information area is used. This includes transition probability information that moves to the data page.

1 Yutaka 1 Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram of an embodiment of the present invention, in which a main storage device 1 is provided with a buffer pool 10, and this buffer pool 10 contains each database stored in a secondary storage device 2. Data 20 is read under the control of page input program 3. Output control from the buffer pool 10 to the secondary storage device 2 is performed by a page output program 4.

Each data page 20 of the data pages has a statistical information area 21 added thereto corresponding to the data page, and includes, for example, the frequency of use of that page, the frequency of use in chronological order, and whether the page is used again after being used once. Information such as the frequency or probability of use and the probability of transition from the currently used page to the next page to be used are stored.

The user program 6 specifies the required data in the database and reads it into the buffer pool 10 of the main storage device 1. Furthermore, the statistical information collection program 9 has the function of updating at least one of the above pieces of information in the statistical information area 21 corresponding to each data page 20 every time the program is executed.

The next page prediction program 8 refers to the statistical information in the corresponding information area 21 of the data page 20 and predicts the data page that the user program 6 will request next, in parallel with the user program 6. Secondary storage device 2
It has a function of instructing to read the data from the buffer pool 10 into the buffer pool 10.

The failed page cancellation program 7 detects when the next page predicted by the next page prediction program 8 is not actually requested by the user program 6, and updates the corresponding information in the statistical information 21. be. These user programs 6, statistical information collection program 92, next page prediction program 8, and failed page cancellation program 7 are controlled under the management of the input/output buffer pool management program 5.

In this configuration, the general flow is as follows. In response to a record input request requested by the user program 6, the input/output buffer pool management program 5 finds the target page of the database in the secondary storage device 2, then notifies the page input program 3 to that effect, and transfers it to the main storage device. 1 Enter the page above. Then, the attendance buffer pool management program 5, which has been notified of the page input by the page input program, notifies the user program 6 that the record has been read.

After processing the record, the user program 6 requests the attendance buffer pool management program 5 for the next record. This process is repeated until there are no more user requests.

In this series of processes, in the present invention, the statistical information area 21 provided in the page 20 of the database
The statistical information collection program 9 is used to accumulate statistical information, and the next page prediction program 8 predicts the next page to be processed based on this information.

FIG. 2 is a diagram illustrating an example of information stored in the statistical information area, and the statistical information area 21 for data Qi stores the usage frequency Ci of that data page. In other words, it is the number of times the corresponding data page has been read into the buffer pool 10 and used, and this number of uses is counted each time in the statistical information collection program 9 and updated and stored in the corresponding area 21a as the latest count data. It is something that will be done.

The next page prediction program 8 compares the number of uses of the statistical information area 21a of each data page 20 and predicts the data page corresponding to the largest number of uses as the data that the user program 6 will request next. , this is read into the buffer pool 10.

In addition to statistics on the total number of uses to date, this data on the number of uses can include, for example, data that constantly counts and updates the latest number of uses between now and a certain period of time ago. It is also possible to use series unit usage frequency data), and in this case, prediction accuracy is considered to be higher than in the case of total usage frequency data.

FIG. 3 is a diagram illustrating another example of the information stored in the statistical information area, in which the area 21 for data Di is divided into an area 21b for storing the data page to be used next to this data page, and an area 21b for storing the data page to be used next to this data page. It consists of an area 2IC that stores the transition probability to the data page to be used. In this example, each area 21b and 2IC can store two pieces of information, and the two pieces of information with the highest transition probabilities are stored.

For example, it is shown that for data D1, the probability of transition to data D2 is the highest at 80%, and the probability of transition to data D5 is the next largest at 20%. Therefore, when the data D1 is read into the buffer boolean, the next page prediction program 8 refers to the statistical information areas 21b and 21G of this data D1 and selects D2 as the next data page that the user program 6 will request.
is predicted and read into the buffer pool.

In this case, in the statistical information area 21 for each data page,
Although the transition probabilities of two data pages are stored, and two data with higher probabilities are stored, more data may be stored.

In addition to the above-mentioned information such as the illuminance used and the next page transition probability, information such as the frequency or probability that a page that has been used once will be used again may be used. Further, the next page prediction may be made by using all of the above information and taking all of them into account. In this way, failures in the prediction result will be reduced.

1.1 As explained above, when you have several routine business programs in the field of data processing of databases that are diversified for each user, by using the input/output control method using statistical information according to the present invention. The following effects can be obtained. In other words, the more frequently the program is used, the more accurate the prediction of the next desired input page becomes, and since the input/output program operates in parallel with the user program, the preprocessing time for input/output can be reduced to zero. I can do it. In addition, since unnecessary pages are accurately found, the number of inputs and outputs to and from main memory and secondary memory is reduced, further increasing execution speed.

[Brief explanation of drawings]

FIG. 1 is a block diagram of an embodiment of the present invention, FIG. 2 and FIG.
The figure is a diagram showing an example of a statistical information area of a data page for explaining the operation of the embodiment of the present invention. Explanation of symbols of main parts 1...Main storage device 2...Secondary storage device

Claims (3)

[Claims] (1) A data control method for reading data on a secondary storage device into a buffer pool of a main storage device as necessary, which adds a reading statistical information area corresponding to each data page of the data, and Information indicating how the data pages corresponding to each information area are read into the buffer pool is stored while being statistically rewritten each time the program is executed, and this information is used to store the data pages requested by the next program. A data control system characterized in that a data page is predicted to be a blank data page and the predicted data page is read into the buffer pool. (2) The data control method according to claim 1, wherein the information is usage frequency information on how the data page corresponding to the information area is read into the buffer pool and used. (3) The data control method according to claim 1, wherein the information is transition probability information for moving from a data page corresponding to the information area to another data page.