JPH03231339A - Data base access system - Google Patents

Abstract

PURPOSE:To improve the access performance to a data base by deciding a buffer control processing system with an access request and performing the buffer control suited to the processing based on the store state of a table. CONSTITUTION:A store state setting part 25 checks the store states of tables 22-1 - 22-x contained in a data base file 20 before the access requests are received from an inquiry processing part 29 to the tables 22-1 - 22-x. The part 25 decides whether the key value of a key field has a forward state or a non- forward state and updates the store state entry of a directory part 28 to set the deciding information. Then the part 29 performs the buffer control in an FIFO system or an LRU (least recently used) system in response to a fact whether the accesses must be successively given to those tables of the file 20 or not. As a result, the access performance to a data base is improved.

Description

Detailed Description of the Invention [Objective of the Invention (Industrial Application Field) The present invention provides a database access method having a buffer management function for reading data in units of data blocks constituting a database and storing the data in a buffer. Regarding.

(Prior Art) Generally, when a database (stored in a disk device) is accessed and desired processing is executed, the stored data is called into a buffer and then executed.

A call to data stored in a database is performed using a page (also called a block, which is a unit of data transfer between disk and memory), which is a component of a database table.
It is done in units of. A buffer management method is used in which tables of this database are stored in a call buffer. The LRU method replaces the page in the buffer for which the longest time has passed since it was last accessed (the buffer with the least access frequency) with the new page to be read next, regardless of the table storage state on disk. ). At this time, if the content of the page on the buffer has been updated, it is written back to the disk, and if it has not been updated, it is left as is without being output. The LRU method is an effective method when pages on a disk are accessed in a somewhat random manner.

Here, the table storage state on disk refers to the relationship between the storage order of tuples in the pages that make up this table and the order of the values of key items corresponding to each tuple when the table is first stored in the database. It shows. That is, whether the logical order of the key values is the same as the physical order of the tuples containing the key values on disk (hereinafter this state will be referred to as the ordered state) or not (hereinafter this state will be referred to as the non-ordered state). ).

(Problem to be Solved by the Invention) The LRU method is an effective method when randomly accessing pages on a disk, but when pages in a table are accessed randomly, for example when calculating the total value of a field in a table. In the case of sequential access, if the storage state of the table is sequential, processing efficiency will decrease. In other words, even if the key values and tuples are stored in the same order, the LRU
This is because the buffers are managed using a method, so access (buffer management) using the table storage state (sequential state), that is, parallel processing cannot be performed.

The present invention has been made in view of the above-mentioned points, and an object of the present invention is to provide a database access method that improves database access performance by utilizing table storage states to perform buffer management suitable for processing. do.

[Structure of the Invention] (Means for Solving the Problems) The present invention provides a database access system having a buffer management function for reading data in units of data blocks constituting a database and storing the data in a buffer. make an access request to
an inquiry processing means for executing a desired process on the data stored in the buffer; and when an access request is made by the inquiry processing means, depending on whether the processing content is to sequentially access the database. The buffer management processing method (read-ahead FIFO)
and a buffer management means for reading data from the database in units of data blocks and storing the data in the buffer.

Further, a storage state determination means for determining whether the physical storage state of data in the data block is an order state or an out-of-order state, and a storage state for storing the storage state determined by the storage state determination means. and a state storage means for buffer management based on the storage state of the data block stored in the state storage means when the processing content by the inquiry processing means is to be sequentially accessed. The system is configured to determine the method.

(Operation) According to such a configuration, when the processing content of the inquiry is to perform sequential access, the access is performed using the look-ahead FIFO method as the buffer management processing method. For example, in the case of a query to find the total value, minimum, or maximum value of a certain field on all pages in a table (if all pages need to be accessed), buffer management using sequential access is possible. A look-ahead FIFO method is used that can shorten access time through parallel processing. Also, before query processing is executed, the storage state of the pages (data blocks) of each table in the database (sequential state, non-sequential state!
) in advance, if the content of the inquiry process requires sequential access (when it is necessary to access a specific page), sequential access is possible, and the read-ahead FIFO method is possible. It is determined whether it is possible to use By using the read-ahead FIFO method, when data is stored on a disk device, seek time is shortened due to sequential access, and throughput is significantly improved, especially in systems where the database is shared by many terminals (processes). .

(Example) Hereinafter, an example of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing the configuration of a database system to which the database access method according to the embodiment is applied. In the figure, 20 is a database file, which contains schema information 21° and a plurality of tables 22-1 to 22-1.
22-x. The schema information 21 includes table attribute information such as table name, storage location, size, attributes of constituent fields (data type, size, etc.), and number of fields. Tables 22-1 to 22-x are
It consists of multiple pages which are access units.

A page is a unit of access between the tables 22-1 to 22-x on the disk and the buffer 30 (described later), and is composed of a plurality of tuples. Furthermore, a tuple is made up of multiple fields.

The database file 20 is processed by the storage state setting section 25 and the buffer management section 26 .

The storage state setting section 25 checks the storage state of each table 22-1 to 22-x in the database file 20 and writes whether the table is in an order state or an out-of-order state as a storage state entry along with the table name in the directory section 28. It is for storing. The directory section 28 stores information indicating the table name of each table and the corresponding table storage state (order state, non-order state). The buffer management unit 26 receives data from the database file 20 from the inquiry processing unit 29.
Access to the tables 22-1 to 22-x within is requested, and buffer management is performed in response to this request. The sofa management unit 26 uses a sofa management method (LRU method, pre-read F
IFO method) and change tables 22-1 to 22-x.
is accessed and stored in the buffer 30 page by page. The inquiry processing unit 29 uses the schema information 21 of the database file 20 to create tables 22-1 to 22-
It requests access to x and executes the desired process.

It is assumed that the buffer 30 is provided with multiple storage areas (for multiple pages).

Next, the operation of this embodiment will be explained.

First, tables 22-1 to 22-1 by the inquiry processing unit 29
22-x, the storage status of the tables 22-1 to 22-x in the database file 20 is checked by the storage status setting unit 25 according to the procedure shown in the flowchart of FIG. Note that the table storage state refers to whether the logical order of key values is the same as the physical order of the tuples containing the key values on disk (order state);
Or not (non-order state).

The storage state setting unit 25 searches for all key values from the first tuple of the table 22-1, and sets the array (order state,
out-of-order state) (step AI). Figures 3 and 4
The figure shows an example of a table stored in a sequential state and a non-sequential state, respectively. When the storage state setting section 25 determines that the key value of the key field is in the order state as shown in FIG. 3 (step A2), it creates a new storage state management entry in the directory section 28 and enters the table name and Information indicating the forward state is set (step A3). Further, in step A2, if it is determined that the key value of the key field is in the ascending order state as shown in FIG. (Step A4).

Here, the storage status of all tables in the database file 20 is checked to determine whether status storage entries have been created for each table (step A5). That is, it determines what kind of table it is from the attribute information of the table written in the schema information 21, and determines whether state storage entries have been created for all tables. If a state storage entry for each table has not been created, the storage state of the next table is checked in the same manner as described above, and the result is set in the directory section 28. When state storage entries are created for all tables, the table storage state setting process ends. Note that as a criterion for determining whether or not the storage state of a table is in the order state, it is not necessary that the tuples be stored strictly in the order of key values, and a certain degree of deviation (dispersion) may be allowed.

Next, inquiry processing regarding the database stored in the database file 20 will be explained with reference to the flowchart shown in FIG. Inquiries to the database specified by the user are performed by the inquiry processing unit 29. The inquiry processing unit 29
In order to call the table to be processed from the database file 20 to the buffer 30 and perform processing, the buffer management unit 26 is requested to access the table. Upon receiving the request from the inquiry processing unit 29, the buffer management unit 26 determines whether the access request is for sequential access (step B1). For example, the determination is made based on whether the process requires accessing the table sequentially, such as when calculating the total value of a certain field or when performing sorting. Here, in the case of sequential access processing, the storage state entry with the table name of the table to be accessed is searched from the directory section 28 (Step B2), and the storage state of the table (sequential state, non-sequential state) is checked. The information shown is extracted (step B3). If the retrieved storage state entry indicates an ascending order state (step B4), the buffer management unit 26
is the LRU management method for the table.
The requested page is retrieved and stored in the buffer 30 (step B5). Note that step B1
If it is determined that the access request is not one for sequential access, buffer management is performed using the LRU method.

On the other hand, in step B4, if the storage state entry taken out from the directory section 28 indicates the storage state of the table in order, the read-ahead FIFO (first in, first out)
Perform bath/sofa management according to the method (step B6)
. The flowchart in FIG. 6 shows the processing procedure for bath sofa management using the look-ahead FIFO method. Note that the read-ahead FIFO method refers to sequentially rotating (buffer number 1 ,2゜...,n-1,n,
1, 2. ...) and 87 as shown in Figure 8.
Buffer i+ before starting to process page p on page 771
By issuing the input of the next page p11 on the first page,
This is a method in which the processing of page p and the input of page p+1 are performed in parallel.

First, the buffer management unit 26 determines whether the page of the table targeted for the access request already exists in the buffer 30 (step CI). If it does not exist (
Step C2), it is determined whether the contents of the buffer to be rewritten have been updated, and if so, they are written back to the disk (step CB). Whether the content has been updated or not is determined by, for example, providing an update flag on each side of the buffer 30 that is set to ON when the contents of the buffer are updated, and checking the state of this update flag. After the contents of the buffer are written back to the disk (if the contents have been updated), the buffer management unit 26 stores the contents of the requested page in the buffer targeted for rewriting (step C4).
).

In this way, when the request table exists (
(including the case where it is determined in step C2 that the requested page already exists in the buffer 30), it is determined whether the contents of the page stored in the next buffer in which the requested table is stored has been updated, and If so, write it back to the disk (step C5). Next, the page next to the requested page is read ahead using a buffering method and stored in the next buffer (step C6).

When the requested page is stored in a predetermined buffer, the processing unit 2 queries the address of the buffer where the requested page is stored.
9 (step C7). Inquiry processing unit 29
performs processing on the requested page stored in the buffer. In other words, by using the look-ahead FIFO method, the ninth
As shown in the figure, for example, the input of the page buffer is p,
If the flow is p+l, p+2,..., page processing on the buffer is performed in parallel with this page input.
1°p, p+1. It is executed as follows.

In this manner, when the content of the inquiry process is to sequentially access pages in a table, buffer management can be performed using the look-ahead FIFO method to shorten the access time. In particular, when an access request is made from the inquiry processing unit 29 to a specific database by multiple processes (for example, when there are a large number of terminals for manually issuing inquiry instructions), the read-ahead FIFO method is used for processes that can read ahead. Since access is performed, the access time for the entire system can be significantly shortened.

[Effects of the Invention] As described above, according to the present invention, when the processing content is sequential access, it is determined whether the storage state of the request table is in the sequential state, and if it is in the sequential state, pre-reading is performed. F
Bath and sofa management is performed using the IFO method. Therefore, processing time can be shortened since the storage state of the table is processed in parallel. Therefore, the database access performance of the entire system is improved.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the configuration of a database system to which a database access method according to an embodiment of the present invention is applied, FIG. 2 is a flowchart showing the procedure for setting information indicating the storage state of tables, and FIGS. FIG. 4 is a diagram showing an example of the storage state of a table, FIG. 5 is a flowchart showing the processing procedure for querying the database, FIG. 6 is a flowchart showing the processing procedure for buffer management using the look-ahead FIFO method, and FIGS. Figure 8 is a diagram for explaining the look-ahead FIFO method, and Figure 9 is a diagram to explain the look-ahead FIFO method.
FIG. 3 is a diagram for explaining the processing contents by the IFO method. 20... Database file, 21... Schema information, 22-1 to 22-x... Table, 25...
Storage state setting unit, 26...Buffer management unit, 28...
・Directory section, 29... Inquiry processing section, 30... Buffer

Claims (3)

[Claims] (1) In a database access method that has a buffer management function that reads data in units of data blocks constituting a database and stores it in a buffer, an access request is made to the database and a desired request is made for the data stored in the buffer. an inquiry processing means for executing a process, and determining a processing method for the buffer management according to whether or not the processing content requires sequential access to the database when an access request is made by the inquiry processing means; A database access method comprising: buffer management means for reading data from the database in units of data blocks and storing the data in the buffer. (2) The buffer management means uses a look-ahead FIFO method when the processing content of the inquiry processing means is to sequentially access the database, and uses an LRU method in other cases. A database access method according to claim 1. (3) storage state determining means for determining whether the physical storage state of data in the data block is in order or out of order; and storing the storage state determined by the storage state determining means. and a state storage means for storing the data block in the buffer based on the storage state of the data block stored in the state storage means when the processing content by the inquiry processing means is to perform sequential access. A database access method according to claim 1 or 2, characterized in that a management processing method is determined.