JPH01173229A - Data base processor - Google Patents

Abstract

PURPOSE:To perform efficient data base processing by providing a hash bit array and accessing it by a search processing device and a sort processing device to reject data to which setting of hash values is necessary. CONSTITUTION:In the case of table coupling operation, a pertinent tuple group of a first table in a data base is extracted on a certain retrieval condition by a search processing device 3 and is sorted by a sort processing device 4, and the has value is set to a hash bit array 10. After a second table is retrieved in the same manner, rejection of unnecessary data by hashing and sorting are performed by the device 4. Unnecessary data of the first table are rejected thereafter if necessary. Since setting of hash values to the array 10 and rejection of unnecessary data are performed in both of devices 3 and 4 by these processings, the quantity of data transfer between each processing device and a CPU is considerably reduced and efficient data base processing is possible.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a database processing device, and more particularly to a database processing device suitable for speeding up database search and sorting processing.

[Conventional technology]

The main processing of a database machine is search processing and sort processing. In the search process, necessary data is retrieved from an external storage device (usually a disk device) in which the database is stored, according to the read speed of the disk. For this purpose, various products have been developed in which hardware for executing so-called on-the-fly processing is installed between the main unit (central processing unit) and an external storage device, thereby greatly reducing the processing on the main unit. Regarding sort processing, there is a method such as search hardware in which data is sorted using specialized hardware that follows the flow of data from disk, but in large-scale databases, the database is distributed across multiple disks, so As actual sorting hardware, hardware for sorting data collected in the memory of the main unit has been commercialized.

Note that database machines that process databases and basic processing techniques for databases are discussed, for example, in Software Oriented Architecture, December 25, 1985, pages 330 to 383.

[Problem that the invention seeks to solve]

In the conventional technology, consideration is not given to the balance between the combination of each elemental technology of the database machine and the performance.

For example, sorting hardware has a built-in type or a channel-connected type, but the former is hardware dedicated to the main unit and its usage is fixed to the model, and the latter uses channels, which increases channel overhead. There are performance issues such as.

In addition, in table join operations, there is a technique that uses a hash bit array to filter out tuples that have no possibility of being joined, but since sort 1- processing is essential in table join operations, a sort processing device can perform this function. In many cases, screening is performed using a hashish bit array. For example, when considering the hash of a sort processing device that is connected to the main unit via a channel, first the hash bit array is set using the data in the first table, and then the second table data is filtered out using the focus array of the previous setting. Resetting the Shuvitt array, and filtering out the first table using the Hatschvitt array, really 3
This requires multiple table transfers, resulting in a significant drop in processing speed.

In view of the above, an object of the present invention is to provide a database processing device that enables the construction of a database machine with excellent cost performance.

[Means for solving problems]

The above object is to provide a search processing device and a sort processing device in a database machine having a search processing device and a sorting processing device between a central processing unit and a storage device connected to the central processing unit through an input/output interface to store a database.
This is achieved by providing a hash bit array that can be accessed in common from the processing devices, and by referring to the hash value set at the time of search by the search processing device at the time of sort processing by the sort processing device.

[For production]

For example, in the case of a table join operation, a search processing device extracts the corresponding tuples of the first table in the database using a certain search condition, and a sorting processing device extracts the corresponding tuples from the first table in the database.
Then, the hash value is set in the hash bit array. Next, after searching the second table in the database in the same way, the sorting processing device performs hash screening and sorting. After that, if necessary, the first table is also screened. In these processes, by setting and filtering hash values into hash bit arrays in both the search processing unit and the sorting processing unit, the amount of data transferred between each processing unit and the central processing unit is greatly reduced. This enables efficient database processing.

〔Example〕

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

FIG. 1 is a block diagram of an embodiment of a database processing apparatus of the present invention. In Figure 1, the central processing unit (CP
U) 1 is connected to a search processing device 3 and a sorting processing device 4 via input/output channels 2, respectively. A disk device 5 as an external storage device is connected to the search processing device 3 and stores a database.

The search processing device 3 has an input buffer 6, a search section 7, a column extraction section 8, and an output buffer 9.

The sort processing device 4 includes an input/output buffer 11 and a key extraction unit 1
2, a sorting section 13, and an internal memory 14. Furthermore, the search processing device 3 and the sorting processing device 4 share a hash bit array 10 that stores hash values, and can be accessed in common from both.

Figure 2 shows the form of a table in a relational database. A is a primitive table and is composed of 1, 2°...N tuples and columns a, bTc,...X+MtZ. The database stored in the disk device 5 is in the form A. The search processing device 3 extracts from the A-shaped table a column to which a tuple matching the search target column is projected. The result is a table in the form B (search result table). For example, a certain value in column a is the search target, and tuples 1 to n that match this value are selected, and a, b are the projection targets.

...Y+Z is extracted and a table of the form B is created. The sort processing device 4 sorts this table B. When sorting table B, if we want to sort by the values of columns a,...2, the column values of a,...Z are called keys, and the column that becomes the key is extracted from each tuple. C table (sort key table)
It is. Sorting is performed according to this C table. Similarly, search processing and sorting processing are performed for other tables.

A table join operation is to combine two tables with the same key value to form one table. Therefore, tuples with inconsistent key values are not needed for table joins. The hash bit array 10 scatters the value of this key according to a certain hash function, addresses the hash bit array 10 with that value, and sets or refers to the hashing result to create a tuple of mismatched keys. It is to be removed. This filtering through hashing greatly reduces the number of tuples in the table, and the time required for subsequent table joins can be greatly reduced.

The operation shown in FIG. 1 will be described in detail below in the case of table join. First, the first primitive table (table A in Figure 1)
) is read from the disk device 5, temporarily buffered in the input buffer 6 of the search processing device 3, and the search unit 7 detects matching tuples. Match tuples are extracted from column extraction section 8.
The projection is performed by , and the result (table B in FIG. 1) is output to the output buffer 9. At the same time, the column extraction unit 8 extracts a part of the sort key from a part of the projection column, performs hashing, and sets it in the hash bit array 10. The data in the output buffer 9 is sent to the central processing unit 1 via the channel 2. The second source table to be joined is also searched by the search processing device 3 in the same way as the first source table, but the difference is that the hash bit array 10 is referred to based on the hashing result,
If the corresponding bit is not set, this tuple is filtered out, and if it is set, a new badge bit is set, and only the target tuple is sent to the central processing unit 1. This results in
The amount of data sent to the central processing unit 1 is significantly reduced,
Data transfer time is reduced.

The search results of the first table (table B) are transferred from the central processing unit 1 to the sorting processing unit 4 via the channel 2 for sorting processing. The data is temporarily buffered in the input/output buffer 11, and then sent to the sort key extraction unit 1.
After the key is extracted by 3, the hashed bit array 10 is referred to and the key has already been extracted by the search processing device 3.
Key screening is performed using the values reset by the second table in . The remaining keys are in the sort section 13
The data is sorted by the internal memory 14 (sorted on table C in FIG. 1) and transferred to the central processing unit 1 via the input/output buffer 11. The search results of the second table are similarly sorted by the sorting central processing unit 4 and returned to the central processing unit 1. The central processing unit 1 combines the sorted data of the first and second tables with the same key value to form one table.

〔Effect of the invention〕

As is clear from the above description, according to the present invention, in a database processing device that performs search and sort processing, unnecessary data can be filtered out using a hash during the search and before sorting, and the search and sort processing time is shortened. At the same time, the amount of data taken into the central processing unit is greatly reduced, enabling efficient database processing.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of an embodiment of a database processing apparatus of the present invention, and FIG. 2 is a diagram showing a table format of a relational database. DESCRIPTION OF SYMBOLS 1... Central processing unit, 2... Channel, 3... Search processing device, 4... Sorting processing device, 5... Database storage disk device, 10... Hash bit array. I-hetefu l'

Claims (1)

[Claims] (1) Database processing consisting of a central processing unit, a storage device that stores a database, and a search processing device that searches the database and a sorting device that sorts the database, which are placed between the central processing unit and the storage device. In the device, a hash bit array for hashing database data is provided in common to the search processing device and the sort processing device, and the hash bit arrays are mutually accessed by the search processing device and the sort processing device to set hash values. A database processing device characterized by filtering data.