JPH02253449A - Simultaneous execution control system for data base - Google Patents

Abstract

PURPOSE:To improve the simultaneous feasibility of a data base and to realize acceleration by providing a time table to store the latest updating time of index information at a common space, and deciding whether the index information held at its own task is used or re-reading is performed from the data base by each task while referring to the table. CONSTITUTION:Access to a data file 1-2 is performed by using the index information in its own task when the time of the index information copied on the time table 6-2 in its own task is equal to or later than the time when copied on the time table 3 provided in the common space, and using the index information copied from an index file 1-1 when the time is older or no time exists. In such a manner, it is possible to make access to a data file 1-2 in recording unit at high speed by using the index information held in its own task by each task unless another task updates the index information.

Description

[Detailed Description of the Invention] (Summary) To solve the problem of deterioration in high-speed processing capacity or concurrency due to exclusive index when performing exclusive control on a record-by-record basis in a database management system having an index for high-speed search. , by setting the index content modification time in a common space within the system, and when the indexes in each process have the same time, access processing is performed using the index in the own process when the index is new, improving high-speed processing performance. It also realizes concurrency by localizing exclusive time for indexes.

[Industrial application field]

The present invention relates to a database concurrency control method that uses index files to access data files. ? When implementing exclusive control on a database record basis for Jl tasks, it is desired to implement exclusive control on indexes quickly and concurrently.

[Prior art and problems to be solved by the invention] In general, a database, as shown in FIG. 4, consists of a data file 1)-2 for storing actual data and an index file 1)-2 for high-speed data searching. It consists of 1. Data file 1)-2 is configured to perform exclusive control on a record-by-record basis in order to improve parallel processing performance. On the other hand, index file 1)-1 usually stores multiple record information for each access, and changes in response to record additions and deletions.
The following processing is performed.

(1 index file 1)-2 is locked from the time processing for one record starts to the time it ends.

(2) Each task has a copy of the index file 1) 1, and updates the contents within its own task when adding, updating, deleting, etc. At the end of the transaction, the database index file 1)-1 and the copy in the self-task are merged to create an index file ti.
-1.

This conventional method has the following problems.

(1) The invoking task needs index file 1)-1 only during the period when index file l-1 is locked.Since index file 1)-1 cannot be copied and used in the invoking task's buffer, each record operation Since it is necessary to read the relevant portion of the index file 1)-1 into the own buffer, there is a problem that the speed cannot be sufficiently increased by using the buffer.

(2) It is necessary for each task to secure an area to store a copy of the index file 1)-1, which increases the memory capacity in proportion to the number of tasks and increases the overhead when merging. There is a problem with putting it away.

The present invention provides a time table that stores the latest update time of index information in a common space, and each task refers to this time table and uses the index information held within its own task/index information reread from the database. By accessing data files using , the purpose is to localize exclusive control over index information to the period of operating one record, improve database concurrency, and speed up the use of buffers. .

[Means to solve problems]

Means for solving the problem will be explained with reference to FIG.

In FIG. 1, a database l is a data file l.
-2 and an index file l-1 for quickly searching records in this data file l-2.

The time table 3 is provided in a common space and stores the latest update time of the index information imported from the index file l-1 of the database l.

The time table 6-2 is a table that stores the times provided in each task at the time of copying or when the own task updates, etc., in association with index information.

[Effect]

As shown in FIG. 1, the present invention associates the time of copying from index file l-1 of database l or the latest update time with index information in a common space that can be referenced and updated from each task. A time table 3 to be stored is provided, and when each task accesses the database l, the index file 1-
If the time when the index information imported from 1 is copied or the time when the invoking task 6 updates it is equal to the time in the above time table 3, the index information in the invoking task 6 is used if it is new, while the index information in the invoking task 6 is used if it is old or does not exist. access the corresponding record in data file 1-2 using the copied index information, and update the time of the index information in time table 3 and the self-task to the current timer value when the index information is updated. That's what I do.

Therefore, the time of the index information held within the self-task is
By accessing the data file 1-2 using the index information within its own task if it is new, the time when it was copied to the time table 3 provided in the common space/the latest update time, other tasks can Unless the index information is updated, each task can access the data file 1-2 at high speed using the index information held within its own task.

〔Example〕

Next, the configuration and operation of one embodiment of the present invention will be explained in detail using FIGS. 1 to 3.

In FIG. 1, a database l is a data file l that performs exclusive control on a record basis, which is the minimum access unit.
-2, and an index file 1-1 for searching this data file 1-2 at high speed.

The time table 3 stores the time when the index information was copied from the index file l-1 of the database l or the latest update time in association with the index information. This time table 3 is provided in a common space that can be accessed from each task.

The lock manager 4 is for exclusive control of the index file 1-1 in file units and the data file 1-2 in record units. This exclusive control is performed by setting the file name of the index file l-1, the file name and record name of the data file l-2, which are subject to exclusive control, in the lock table 4, and also by setting the open flag to on. There is.

Timer 5 is a timer owned by the system and is used to set the time when index information is copied from index file l-1 to time table 3.6-3 or the latest update time. This time is, for example, the current time in μsec.

Tasks (database operation tasks) 6 are provided in the task-specific space and, for example, access database l in response to a database search request from a terminal and notify the terminal of the corresponding search results. It is something. This task 6 stores a buffer 6-1 that stores index information imported from index file 1-1 of database l and data imported from data file 1-2, and a buffer 6-1 that stores the index information imported from index file 1-1 and data imported from data file 1-2, and the time at which the index information was imported (copied time) and the buffer 6-1. It has a time table 6-2 that stores the latest time updated by task 6 in association with index information.

Next, the operation of the configuration shown in FIG. 1 will be explained in detail in accordance with the order shown in the flowchart of FIG.

In FIG. 2, ■ opens database l.

(2) It is determined whether the index file 1-1 is opened first in the system. If YES,
Step (2) acquires an area for the time table 3 in the common space, initializes the file identification value (file name) and time identification value (time), and executes step (2). In the case of NO, since the file has already been opened, execute ■ without executing ■.

■Starts a transaction.

(2) locks the Changhai file in the index file 1-1. This means that task 6 assigns a lock (exclusive control) to lock manager 4 by specifying the file name of index file l-1, and lock manager 4 assigns a lock to index file 4-1! This means that the corresponding file is locked by setting the file name and lock flag to ON.

In (2), it is determined whether or not file information (index information) exists in the buffer 6-1 within the self-task. YES
In the case of , compare the time of the corresponding file name in time table 6-2 with the time of the corresponding file name of time table 3 in ■, and if the former is equal to the latter and new (when YES), It is judged that the index information existing in the invoking task's buffer 6-1 is the latest one and executes ■.On the other hand, when the former is older than the latter (No), the invoking task's buffer 6-1 is executed. Judging that the index information existing in is old, execute [phase] to set the latest index information and current time, and then execute 0.

(2) reads file information from the index file 1-1 of the database 1 into the buffer 6-1 in the task 6;

0 sets the time of the corresponding file name in the time table 6-2 to the current time (for example, timer value T+). And ■
Execute.

(2) collects information on data file l-2.

@ locks records in data file 1-2.

This is an exclusive control (lock) that task 6 in Figure 1 specified the file name and record in data file l-2.
This means that the lock is performed by relying on the lock manager 4.

O operates (accesses) records in data file l-2.

[Phase] determines whether the index file 1-1 has been updated. If YES, update the index information with ■, set the times of time table 6-2 and time table 3 to the current time (for example, timer value T8) with [phase], and execute @. On the other hand, in the case of NO, since the index information is not updated, process (2) is executed.

■Unlocks (releases exclusive release) the index.

[Phase] determines whether the next record is valid or not. YES
In this case, repeat steps ◯ to ◯.

In the case of No, the series of transactions is completed at [phase], and the database l is closed at 0.

(2) It is determined whether the use of the index file l-1 within the system has ended. If YES, the corresponding index ↑R information is released from the time table 3 and the process ends. If NO, the process ends.

FIG. 3 shows an example of a database, where index file 1-1 contains index information i (for example, itame, Yamamoto, etc.) and data for high-speed searching of the corresponding record of data file l-2. Record number (
For example, record no.

2 etc.) are stored. Data file 1-2
The data is divided into a plurality of items and stored sequentially in units of records as shown in the figure.

Since a configuration is adopted in which data file 1-2 is accessed using index information copied from data file l-1, each task retains the index in its own task unless another task updates the index information. Using the information, data file l-2 can be accessed record by record at high speed. As a result, the lock period for the index file l-1 can be localized during the index operation for one record of the database l, and it is possible to improve concurrency and speed up by using buffers.

〔Effect of the invention〕

As explained above, according to the present invention, the time of the index information copied to the time lever 6-3 in the own task is the time when the index information was copied to the time table 3 provided in the common space/the latest update time. , use the index information in the self-task if it is new, or import it if it is old or does not exist.

[Brief explanation of drawings]

FIG. 1 is a configuration diagram of one embodiment of the present invention, FIG. 2 is a flowchart explaining the operation of the present invention, FIG. 3 is an example of a database, and FIG.
The figure shows an explanatory diagram of the prior art. In the figure, 1 is a database, l-1 is an index file, l-2 is a data file, 3 is a time table, and 4
4-1 is a lock manager, 4-1 is a mouth/two-piece group, 5 is a timer, 6 is a task, 6-1 is a buffer, and 62 is a time timer. Patent applicant: Pfn Co., Ltd.

Claims (1)

[Claims] A database concurrency control method in which a data file is accessed using an index file, comprising: a database (1) consisting of an index file (1-1) and a data file (1-2); A time table (3) provided in a common space that stores the latest time of the index information imported from the index file (1-1) in (1), and each task (6) within its own task (6). When the time when the index information imported into the above-mentioned index file (1-1) held in The data file (1-
2) and when index information is updated, etc., the time table (3) and the time of the relevant item in the own task are updated to the current timer value. control method.