JPH04117561A - Data base system - Google Patents

Abstract

PURPOSE:To improve the inquiring efficiency to a data base by processing only the partial assembly of the data base shown by a partial assembly list by a data base arithmetic unit when a command issuing means of a host CPU issues a command where the partial assembly list is set to the data base arithmetic unit. CONSTITUTION:A partial assembly list production means of a host CPU 10 has an access to an index file 60 with use of the key value designated with the selecting conditions and searches pointers to all records having the same value as the designated key value to produce a partial assembly list showing the position of the record group. This produced assembly list is set into a command 21 by a command issuing means of the CPU 10 and sent to a data base arithmetic unit 30. The unit 30 receives the command 21 and inputs only the partial assembly of a data base that is shown by the partial assembly list set in the command 21 to process the partial assembly. Thus the data base inquiry processing speed is increased.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Industrial Application Field) The present invention relates to a database system equipped with a database arithmetic unit for efficiently processing database queries.

(Prior Art) Conventionally, selection conditions are specified in query processing to a database. This selection condition is, for example, product number - “50
This is specified to select a target record using a specific value such as "01".In general, a method is used to efficiently perform query processing by adding an index to the field that is the selection condition. Such fields are called index keys (hereinafter simply referred to as keys).Keys include primary keys whose field values are unique, and secondary keys whose field values are unique. If you use a primary key, the value is unique, so there is only one record to process, and therefore processing can be done quickly.However, 2
Since the next key has duplicate values, it takes time to search for the record to be processed. Now, in order to quickly find the position of a target record with respect to a key value, an index file has traditionally been used. The index file stores as many index records as there are records on the database, each consisting of a key value and a pointer to a record on the file for each primary key and secondary key.

By the way, in a system that handles databases, host C
In order to reduce the load on the PU and process the database efficiently, a database processing unit is generally added that exclusively processes the database upon request from the host CPU. This database calculation device is
It is connected to the host CPU via a bus, and originally provided stable and high-speed processing performance that was not unique to keys. In other words, the database arithmetic device always inputs the entire database according to instructions from the host CPU in order to process the target record specified by the selection conditions. However, as mentioned above, using keys (index files) should be more efficient than inputting the entire database. In other words, the stable processing performance of the database processing unit meant that optimal high-speed processing could not be achieved.This inconvenience was caused by the host CPU giving the entire database to the database processing unit. was.

(Problems to be Solved by the Invention) As mentioned above, in the conventional database system, even though the key (index file) exists, the host CP
Since the entire database is given to the database arithmetic unit according to instructions from H, there is a problem in that the database access performance deteriorates.

This invention has been made in view of the above circumstances, and its purpose is to input only a subset of a database containing a target record group specified by selection conditions into a database processing device, thereby making it possible to utilize the database processing device. An object of the present invention is to provide a database system that can efficiently make inquiries to a database that has been updated.

[Structure of the Invention] (Means for Solving the Problems) This invention provides an index file and a record having this key value by referring to the index file from a key value of a selection condition specified in a query to a database. Subset list creation means for searching for a subset of a database consisting of a group and creating a subset list having positional information of the subset, and database processing in which the subset list created by the subset list creation means is set. A command issuing means for issuing commands from the host CPU to the database arithmetic unit, and a database arithmetic unit for exclusively processing database operations in response to commands from the host CPU are provided. The present invention is characterized in that when a command in which the above-mentioned subset list is set is issued to a computing device, only the subset of the database indicated by the subset list is input to the database computing device and processed. It is something to do.

(Function) According to the present invention, the subset list creation means accesses the index file using the key value specified in the selection condition, thereby accessing all the items having the same key value as the specified key value. For each record, find a pointer to that record and create a subset list indicating the position of this record group. When a subset list is created by the subset list creation means, the subset list is set in a command by the command issuing means of the host CPU, and is issued to the database arithmetic unit.

When the database arithmetic unit receives this command, it inputs and processes only the subset of the database (the records forming the subset) indicated in the subset list set in this command.

In this way, the present invention inputs and processes only a subset of the database that includes the target record group, so compared to the conventional system that inputs and processes the entire database, it is possible to use a database using a database calculation device. In addition to speeding up query processing, it is also possible to distribute the load on the host CPU.

(Embodiment) FIG. 1 is a block diagram of a database system according to an embodiment of the present invention. In the figure, 10 is a host CPU that forms the core of the system. This host CP
Ul0 is a subset of the database 50 (database subset) consisting of a group of records having this key value by referring to the index file BO, which will be described later, from the key value of the selection condition specified in the query to the database.
A subset list creation function that creates a list of database subsets 51 (not shown), and a command issuing function that issues commands 21 and the like for requesting the database processing unit 30 to perform database processing on the database subset 51 shown in this subset list. It has a function. 2o is the main memory where commands 21 etc. for database processing are stored; 30 is the host C
In response to command 21 etc. issued by PUIG,
A database arithmetic device 40 that performs a designated database arithmetic operation is a bus for interconnecting the host CPU 10, the main memory 20, the database arithmetic device 30, and a secondary storage device (not shown). Reference numeral 50 is a database located on a secondary storage device and stores records subject to database queries; 51 forms a part of the database 50; database subset to be processed,
60 is an index file located on the secondary storage device in which a group of index records is stored.

FIG. 2 specifically shows the correspondence between the database 50 and the index file 60. The database 50 shown in FIG. 2 consists of a group of records having a key field 52. The database 50 shown in FIG. This record group includes records 53 and 54 having a key field 52 with a key value of “8003°.The index file 60 also consists of an index record group.Each index record is stored in the database 50. The index file 80 consists of a key value 61 containing the value of the key field 52 of the record in question, and a pointer B2 pointing to the record.
The index record group stored in ``S 0
Key field 52 with the same key value 'SOO3' as 03'
A pointer to a record on database 50 with ``
It contains index records 68.64 consisting of "R4" and "R5°.

Next, the operation of one embodiment of the present invention will be explained using the database 50.
A command 21 for causing the database arithmetic unit 30 to process a record on the database 50 that has the value "5OO3" in the key field 52 in an inquiry to
The following will explain how to create the following with reference to the flowchart in FIG. In this embodiment, it is assumed that each index record in the index file 60 is arranged in ascending order by the key value 61 in the record (Note: the index file BO is a B-Tree etc., and the first It is generally not necessary to follow the index records sequentially.)

First, the host CPU 10 in the system of FIG. 1 initializes a subset list (step S1), and further sets the element number i of the list to 0 (step S2). Then, the host CPU 10 retrieves the selection condition "5003" specified in the query to the database 50 (Step 3)
, performs a search operation to search the index file 60 for an index record having the key value 61 that is equivalent to the selection condition "8003" (step S4). This search behavior results in
The first target index record 63 (see FIG. 2) with the key value 61 of "8003" is obtained from the index file 60.

When the host CPU 10 searches for the target index record B3, it moves the pointer "R4" (to the record 53 on the database 50) included in the index record 63 to the i-th element (designated by the element number i). Here i-
The first element Q) is set in the subset list (step S5). Next, host CPU10 uses element number i
is advanced one step (step S6), and the above step S4 is performed again.
performs the search operation. As a result, the host CPU 10 selects the next index record 84 from the index file 60 that has the key value B1 that is equivalent to the selection condition "5OO3" (see Figure 2).
get. When the host c p v io retrieves the index record B4, it retrieves the pointer "R5" (to the record 54 on the database 50) contained in the index record 64.
is the i-th element specified by element number i (here i
-1) in the subset list (step S5). Next, the host CPU 10 increments the element number i by one (step S6) and performs the search operation in step S4 again.The host CPU 10 performs a search operation on the index file 60 and has a key value B1 larger than the selection condition "8003". When an index record is found, in other words, after an index record with a key value 6■ which is the same value as the selection condition "8003" is obtained, an index record with a key value 61 which is a value other than that is detected, the index file 60 is Determine whether the search has ended. This termination judgment is based on the index file BO
This takes advantage of the fact that index records with the same key value (index record H, 84 in this case) are lined up next to each other as a property of By finishing the search of the index file 60 at this point, unnecessary search processing can be eliminated.

As described above, record 53 on database 5o
．． When a subset list with pointers “R4” and “R5°” pointing to 54 is created and the search in the index file 60 is completed, the host CPU ID is set to the created subset list and the element number i (here, 1-2)
and is set in the command 21 on the main memory 20 (step S7).

Based on the subset list and list size (element number) i set in this command 21, the database calculation device 3
Records 53, 54 on the database subset 51 to be processed by o, i.e. the database 5o as shown in FIG.
A database subset 51 consisting of is specified.

The host CPU 10 has a subset list and list size (
When element number) i is set in the command 21, the command 21 is transferred to the database arithmetic unit 30 via the bus 40. Note that if the list size (element number) i is zero, there is no index record in the index file 601 that matches the selection condition. in this case,
There is no need for the host CPU 10 to send the command 21 to the database arithmetic unit 30, and the result of processing the inquiry is zero.

When the command 21 in which the subset list and list size (element number i) are set is transferred (issued) from the host CPU 10, the database processing unit 80 executes the command 21 on the database 50 indicated by the subset list and list size i. Only the database subset 51 is processed. That is, the database processing device 30 transfers only the database subset 51 (records 5L 54 constituting the database subset 51) from the database 50 to the bus 4 according to the specification of the command 21.
0 and processes only this database subset 51.

In the above embodiment, the subset list is described as consisting of pointers of all index records having key values that match the selection condition value, but the present invention is not limited to this. For example, if you have a pair of consecutive pointers at the start and end, you won't need the pointers in between, so
The size of the subset list can be reduced. This effect becomes more noticeable when the number of records that make up a database subset is large. In addition, by doing this, when the database arithmetic unit 8o inputs the target record group, that is, the database subset (51) from the database 5o, it can input them all at once, and as a result, the access time of the secondary storage where the database 50 is placed can be reduced. You can also measure shortening.

C. Effects of the Invention] As detailed above, according to the present invention, by referring to the index file from the key value of the selection condition specified in the query to the database, the database consisting of the record group having this key value can be searched. By creating a subset list with location information of the subsets and issuing a command with this subset list set from the host CPU to the database processing device, the database processing device can select the target record specified by the selection condition. It is now possible to input and process only a subset of the database, including groups, from the database, making queries to the database using a database processing unit more efficient than with conventional systems that input the entire database. It can be carried out.

[Brief explanation of drawings]

FIG. 1 is a block configuration diagram of a database system according to an embodiment of the present invention, and FIG. 2 is a diagram specifically showing the correspondence between a database 50 and an index file BO applied in the embodiment. The figure is a flowchart for explaining the database subset determination procedure in the same embodiment, and FIG. 4 shows the database subset 51 determined in accordance with the flowchart of FIG. FIG. 10...Host CPLJ, 20...Main memory, 21.
...Command, 30...Database calculation device, 50
... Database, 51 ... Database subset, 52 ... Key field, 60 ... Index file, 61 ... Key value, 62 ... Pointer.

Claims (1)

[Scope of Claims] In a database system equipped with a database arithmetic unit that exclusively processes database arithmetic operations in response to commands for database processing issued from a host CPU, a key that is a value of a key field of a record on a database is provided. An index file in which an index record containing a pair of a value and a pointer specifying the position of the same record is stored for each record in the above database, and the above index file is stored from the key value of the selection condition specified in the query to the above database. A subset list creation means that searches for a subset of the database consisting of records having this key value by referring to the database, and creates a subset list having position information of the subset; command issuing means for issuing a command in which the created subset list is set from the host CPU to the database computing device; A database system characterized in that, when a command in which a list is set is received, processing is performed by inputting only a subset of a database indicated by a subset list in the command.