JPS61145661A - Assignment management system of main storage area - Google Patents

Abstract

PURPOSE:To reduce number of input and output in case of join operation especially and to improve the through-put of retrieval processing in a data base system by managing read and storage of a data to a main storage area. CONSTITUTION:When the main storage area is required to read a data, the main storage area represented by an NBEP pointer 4 is used and the main storage areas are rearranged so as to by pointed out by an NBSP pointer 3. In reading an index of a combined column of merge join, the main storage pointed out by the pointer 4 is used and the main storage areas are rearranged so as to be pointed out by a PBSP pointer 1. Then the areas are rearranged to be pointed out by the pointer 4 at the end. That is, the area from the storage area pointed out by the pointer 1 to the storage area pointed out by the PBEP pointer is used as a priority data main storage area. Thus, the through-put of retrieval processing is improved and the number of input/output times for join operation is reduced especially.

Description

[Detailed Description of the Invention] [Field of Application of the Invention] The present invention relates to a main storage area allocation management method % expression % used for inserting data from an auxiliary storage device to a main storage device in a relational database system. [Background of the Invention] As a means of improving the processing speed of database systems,
There is a method to reduce the number of inputs and outputs by retaining frequently referenced data in the main memory among the data stored on the auxiliary memory. On the other hand, 1. Increasing the speed of index calculation in relational databases is an important technical issue, and an example is Japanese Patent Application Laid-Open No. 58-5051. Merge input is one of the processing methods for the input operation.

Merge-in performs a merge-in operation by sequentially comparing indexes added to join columns or join columns arranged in order of size. occurs. Therefore, by retaining indexes and data on the main storage device for a certain period of time, the number of inputs and outputs required for join operations can be reduced, and join operations can be made faster.

[Purpose of the invention]

The present invention improves the throughput of search processing in a relational database system. In particular, the purpose of this invention is to reduce the number of inputs and outputs when performing an input operation using the merge method.

[Summary of the invention]

In database systems, allocation management of main storage areas for data input/output is performed using the so-called LRU method. That is, when reading data into the main storage area, the storage area that holds the oldest referenced data is allocated and used. The method of the present invention generally ensures that frequently referenced data is retained in the main storage area.

However, in processing KD such as merge join, data references are performed intensively in a specific order.

[Embodiments of the invention]

An embodiment of the present invention will be described below with reference to FIG. The main storage area for reading data is divided into n@, each holding a part of the data on the auxiliary storage device. The divided main storage areas are connected to each other by pointers and arranged in the order in which they are used. Generally, when a main storage area is required to read data according to a request from a user program, the main storage area indicated by the NBBP pointer (i.e., the oldest used main storage area) is allocated and used. When reading the index of a join column of a merge twin that rearranges the storage area so that it is pointed to by the NBSF pointer, the read index is referred to frequently from the beginning of the storage area to the end. Therefore, in this case, N
The main storage area pointed to by the BgP pointer is allocated and used, and the main storage area is reordered to be pointed to by the PBSP pointer.

In addition, when the use of the index data is finished, NBε
Rearrange so that it is pointed to by the P pointer. That is,
The storage area starting from the storage area pointed to by P88P to the storage area pointed to by pBgp is a storage area for reading and holding data that is sequentially and intensively referenced. Before reading data, it is checked whether the data is held in the main storage area, and if it is, the content of the reading storage area is not read, or if necessary, the contents of the reading storage area are Processing necessary for managing the main storage area for normal input/output, such as writing out the data to the auxiliary storage device, is also performed in this embodiment. According to this embodiment, when performing a merge twin, the index data is sequentially read only once, and a significant reduction in the number of inputs and outputs is expected.

〔Effect of the invention〕

According to the present invention, regarding the processing of data that is sequential and whose start and end of reference can be detected, as typified by the input processing of a relational database system, the data is read into the main storage area by utilizing its characteristics. By managing the data and its retention, the number of inputs and outputs of the data can be minimized, which has the effect of speeding up various database operation commands, including diversion operations.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing the relationship between the state of a data input/output main storage area and the pointer that manages it in an embodiment of the present invention. 1... Data start pointer, 2...
Final pointer of data, 5... Start pointer of normal data, 4... Final pointer of normal data.

Claims (1)

[Claims] 1. In a relational database system that has means for storing a database based on a relational data model on an auxiliary storage device, and means for analyzing a database search request and reading the data on the database into the main storage device to generate a search result. When allocating the main storage area for reading data, there is a means to manage the storage area for reading and holding data that is sequentially referenced, and to detect the start and end of references to the storage area. 1. A main storage allocation management method, characterized in that the storage area is preferentially allocated to reading and holding the data during the reference period.