JP3622443B2 - T-tree index construction method and apparatus, and storage medium storing T-tree index construction program - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a T-tree index construction method and apparatus and a storage medium storing a T-tree index construction program, and in particular, a memory resident database management system that realizes high-speed data access by making a database resident in a memory. The present invention relates to a T-tree index construction method and apparatus and a storage medium storing a T-tree index construction program.
[0002]
[Prior art]
FIG. 8 shows the structure of a conventional T-tree index. As shown in the figure, the conventional T-tree index 10 includes a node 111 that is a component of the T-tree index and an inter-node pointer 113 that connects between the nodes 111. Each node 111 includes an in-node element 112 that includes a key value 121 and a pointer 116 that points to the position of the record 202 in the table 20 corresponding to the key value 121. The number of in-node elements 112 is an identification value determined at the time of index creation.
[0003]
When searching for the record 202, the key value 121 that matches the search condition designated by the application is searched by using the inter-node pointer 113 and moving between the nodes 111.
Next, the pointer 116 between the records 202 paired with the key value 121 is referred to, and the target record 202 pointed to by the pointer 116 is accessed.
[0004]
FIG. 9 shows a node configuration in a conventional T-tree index.
From the viewpoint of index size, as shown in the figure, when the key value 121 is composed of two columns CLM1 and CLM2, and when the key value 121 is composed only of CLM1, the size of the node 111 is Is different. This is because the key value 121 is directly embedded in each node 111. The size of the node 111 changes depending on the number and length of the columns 201 constituting the key value 121. The size of the node 111 is T Determined for each tree index 10. When the key value 121 is not inserted in all the in-node elements 112, the area is managed as a vacant 115.
[0005]
FIG. 10 shows an increase in duplicate keys in a conventional T-tree index. This figure shows the handling of the duplicate key 133 in the T-tree index 10. KEY3 which is the duplicate key 133 is directly embedded in the node 111. At this time, KEY3 is further inserted, and the number of KEY3 is increased from 2 to 3, whereby the number of key values 121 handled by the T index 10 and the number of stages are increased. As described above, the number of keys 121 to be referred to at the time of search changes depending on the degree of duplication representing the number of records corresponding to one key value, and in some cases, the number of stages of the T-tree index 10 may change.
[0006]
FIG. 11 shows the insertion of key values in a conventional T-tree index. The figure shows an example of inserting the key value 121 into the T-tree index 10. If KEY6 is inserted into T-tree index 10 and the order of KEY5, KEY6, and KEY7 is maintained, KEY4 and KEY5 move, and after KEY6 is inserted, KEY6 is inserted. Is called. Thus, when the key value 121 is inserted into the T-tree index 10, the key value 121 may move.
[0007]
FIG. 12 shows the deletion of the key value in the conventional T-tree index. This figure shows the deletion of the key value 121 from the conventional T-tree index 10 as in the case of insertion. When KEY6 is deleted from the T-tree index 10, if the order of KEY5, KEY7, and KEY8 is maintained, KEY7 and KEY8 are moved. As described above, even when the key value 121 is deleted from the T-tree index 10, the key value 121 may move.
[0008]
FIG. 13 shows a conventional T-tree index backup. The figure shows a state in which a backup 30 of the T-tree index 10 is acquired. A backup 30 of the T-tree index 10 expanded on the memory is acquired on the disk. The contents are information on the hierarchical structure between the nodes of the T-tree index 10 and all information constituting the T-tree index 10 such as the key value 121.
[0009]
[Problems to be solved by the invention]
However, since the above conventional technique embeds key values (including duplicate keys) inside each node of the T-tree, there are the following problems.
1. Since the size of the node changes depending on the number and length of columns constituting the key, the construction of the T-tree index is affected by the key construction, and the construction of the T-tree index becomes complicated, which takes time.
[0010]
2. Since the number of elements in the node is fixed, the amount of memory used as the T-tree index increases when the size of the key value is large and there are many areas managed as empty.
3. Since the key value is moved during key insertion / deletion, the processing speed of record insertion / deletion is slow.
[0011]
4). As the degree of duplication of duplicate keys increases, the number of keys to be searched increases, and comparison with many key values must be performed, so that the processing speed at the time of search deteriorates.
5. Since the backup is acquired for all information included in the index, the time required for the backup is long.
[0012]
The present invention has been made in view of the above points. The T-tree index construction time is shortened, the memory usage is reduced, the processing speed at the time of key insertion / deletion is improved, the search speed at the time of key duplication is improved, and the backup time is reduced. It is an object of the present invention to provide a T-tree index construction method and apparatus capable of shortening and a storage medium storing a T-tree index construction program.
[0013]
[Means for Solving the Problems]
The present invention relates to a T-tree index construction method for constructing a T-tree index used in a database management system that makes a database resident in a memory and realizes high-speed data access.
The T-tree index is a T-tree body consisting of nodes having information on the hierarchical structure of the T-tree index, key value information for comparing conditions at the time of search, and duplicate key information corresponding to a plurality of records for one key. Divided into areas,
The node information of the T-tree index is constructed independently of the number of columns constituting the key, the length, and the key duplication,
Only the nodes constituting the T-tree body in the three divided areas are backed up on the disk.
[0014]
FIG. 1 is a principle configuration diagram of the present invention.
The present invention relates to a T-tree index construction apparatus for constructing a T-tree index used in a database management system that makes a database resident in a memory and realizes high-speed data access.
A T-tree main body storage area 100 for storing a T-tree main body composed of nodes having information on the hierarchical structure of the T-tree index.
A key value storage area 200 for storing key value information for performing condition comparison at the time of retrieval;
A duplicate key storage area 300 for storing duplicate key information corresponding to a plurality of records for one key;
And a management unit 400 that manages three areas: a T-tree main body storage area 100, a key value storage area 200, and a duplicate key storage area 300.
[0015]
In the T-tree main body storage area 100, the node information of the T-tree index is constructed independently of the number of columns constituting the key, the length, and key duplication.
The management unit 400 obtains a backup on the disk of only the nodes constituting the T-tree body among the three areas of the divided T-tree body storage area 100, the key value storage area 200, and the duplicate key storage area 300. Backup acquisition means 410 to be included.
[0016]
The node stored in the T-tree main body storage area 100 includes a pointer indicating the position of the key value stored in the key value storage area in the node element.
When one record corresponds to one key, the key value stored in the key value storage area 200 holds a pointer indicating the position of the record corresponding to each key value,
When a plurality of records correspond to one key, a pointer pointing to one element in the set of pointers in the duplicate key storage area is held.
[0017]
The duplicate key information stored in the duplicate key storage area 300 has a pointer that points to the position of the record corresponding to the duplicate key.
The management means 400 includes means for combining elements in the T-tree main body storage area, the key value storage area, and the duplicate key storage area with a pointer.
The present invention is a storage medium storing a T-tree index construction program for constructing a T-tree index used in a database management system that makes a database resident in a memory and realizes high-speed data access.
The T-tree index is a T-tree body consisting of nodes having information on the hierarchical structure of the T-tree index, key value information for comparing conditions at the time of search, and duplicate key information corresponding to a plurality of records for one key. It has a management process to manage by dividing into areas,
The management process is
The node information of the T-tree index is constructed independently of the number of columns that make up the key, the length, and the key duplication, and only the nodes that make up the T-tree itself are backed up on the disk in the three divided areas. Have a process to get.
[0018]
As described above, the present invention makes it possible to access data by combining, with a pointer, elements in each area where the T-tree index is divided into the following three areas and stored.
For the T-tree body, each node of the T-tree index is stored and information on the hierarchical structure between the nodes is held.
[0019]
The key value storage area stores and manages only the key value.
Further, only the duplicate key is managed for the duplicate key storage area.
For backup, only the T-tree itself is acquired.
Thereby, the following becomes possible compared with the conventional system.
1. Since the creation of a T-tree index node and the construction of a hierarchical structure between nodes can be established independently of the number and length of columns constituting the key, the processing speed of index construction is improved.
[0020]
2. Since the key value is not embedded in the node and the area for storing the key value can be minimized, the memory usage of the entire T-tree index can be reduced.
3. In the insertion / deletion of key values, the growth / decay of the T-tree index can be performed without moving the key values between nodes, and the processing speed at the time of insertion / deletion of key values is improved.
[0021]
4). When handling duplicate keys, the hierarchical structure of the T-tree index can be constructed independently of the degree of duplication of the duplicate keys, and the processing speed when searching for the T-tree index is improved.
5. The amount of backup data can be reduced, and the time required for backup acquisition can be reduced.
[0022]
DETAILED DESCRIPTION OF THE INVENTION
FIG. 2 shows the structure of the T-tree index of the present invention.
This figure shows the correlation of each region in the T-tree index 10 using the construction method of the present invention. The contents of each divided area are as follows.
1. T-tree body area 11:
This is an area for storing the nodes 111 necessary for constructing the T-tree index 10. The in-node element 112 of each node 111 holds a pointer 114 indicating the position of the key value 121 stored in the key value storage area 12.
[0023]
2. Key value storage area 12:
This is an area for storing a key value 121 used at the time of retrieval. When only one record corresponds to one key, each key value 121 holds a pointer 122 that points to the position of the record 202 corresponding to the key value 121. When a duplicate key corresponding to a plurality of records 202 for one key is handled, each key value 121 holds a pointer 123 that points to one element in a set 131 of pointers 132 in the duplicate key storage area 13. doing. The arrangement of the stored key values 121 is not limited.
[0024]
3. Duplicate key storage area 13:
This area is created only when dealing with duplicate keys. This is an area for managing a pointer 132 that points to the position of the record 202 corresponding to the duplicate key.
FIG. 3 shows the structure of the nodes and key values of the T-tree index of the present invention.
The figure illustrates the relationship between the column 201 constituting the key value 121 and the T-tree index 10. When the key value 121 is composed of two columns CLM1 and CLM2, and when the key value 121 is composed of only CLM1, the size of each node 111 of the T main body 11 is the same. Therefore, when a new node is created, when a plurality of indexes are set on one table, all the nodes that construct the index have the same size, and the design and construction of the T-tree index 10 is simplified.
[0025]
In addition, even if an empty space 115 in which the pointer 114 to the key value 121 is not stored in the element 112 of each node 111, the area of the key value 121 corresponding to the empty space 115 is not secured in the key value storage area 12. Therefore, the amount of memory used as the T-tree index 10 can be reduced.
[0026]
【Example】
Embodiments of the present invention will be described below with reference to the drawings.
The following embodiment will be described based on the configuration shown in FIGS.
FIG. 4 illustrates the increase of duplicate keys in one embodiment of the present invention. The figure illustrates the relationship between the duplication degree of the duplication key 133 and the T-tree main body 11. If the KEY3 duplication degree is increased from 2 to 3 by adding a record to the table, it indicates the position of the record inserted in the pointer set 131 in the duplicate key storage area 13 pointed to by the pointer 123 of the KEY3. A pointer 132 is added.
[0027]
The structure of the T-tree main body 11 and the key value storage area 12 does not change even when the importance increases. Therefore, it is possible to prevent a decrease in the processing speed of the key value 121 search due to the duplicate key, and the search processing speed for the key value 121 having no duplication is improved.
FIG. 5 illustrates the insertion of key values in one embodiment of the present invention. This figure illustrates a method for inserting a key value 121 into the T-tree index 10. An example in which KEY6 is inserted into the T-tree index 10 is shown below.
[0028]
1. KEY 6 is inserted into the key value storage area 12.
2. The position where the pointer 114 pointing to KEY 6 is inserted is found in the node 111 of the T-tree body 11. Since the keys are inserted in order of KEY5, KEY6, and KEY7, a pointer Pt6 indicating KEY6 is inserted between the pointers Pt5 and Pt7. At this time, since there is no area for inserting the pointer Pt6, the pointers are replaced for the pointers Pt4 and Pt5, and an area for inserting the pointer Pt6 is secured.
[0029]
3. A pointer Pt6 pointing to KEY6 is inserted.
FIG. 6 illustrates the deletion of the key value according to one embodiment of the present invention. The figure illustrates a method for deleting the key value 121 from the T-tree index 10. An example of deleting KEY6 from the T-tree index 10 is shown below.
1. A pointer 114Pt6 pointing to KEY6 is searched.
[0030]
2. KEY 6 is deleted from the key value storage area 12.
3. The pointer Pt6 is deleted from the node 111.
4). Since the area from which the pointer Pt6 has been deleted becomes empty 115 and the order cannot be maintained, the pointer 114 is replaced with respect to the pointer Pt7 and the pointer Pt8.
[0031]
FIG. 7 shows a backup of one embodiment of the present invention. The figure illustrates the backup 30. Of the three areas of the T-tree main body 11, the key value storage area 12, and the duplicate key storage area 13 constituting the T-tree index 10, only the T-tree main body 11 acquires the backup 30 on the disk. By using only the T-tree body 11 as the backup 30, the data amount of the backup 30 can be reduced, and the write time to the disk at the time of backup acquisition that affects the processing speed of normal operation can be shortened.
[0032]
When recovering from a failure, the backup 30 of the T-tree main body 11 is loaded from the disk onto the memory. Next, the key value 121 is retrieved with reference to the backup of the table 20 in the database, and is combined with each element 112 of each node 111 of the T-tree main body 11 by the pointer 114, whereby the key value storage area 12, duplicate key The storage area 13 is created and the record 202 can be accessed.
[0033]
In the above, the present invention provides a T-tree index, a T-tree body composed of nodes having information on the hierarchical structure of the T-tree index, key value information for performing condition comparison at the time of search, a plurality of records for one key Is divided into three areas of the corresponding duplicate key information, and the node information of the T-tree index is constructed independently of the number of columns constituting the key, the length, and the duplication of the key. Only the nodes that make up the T-tree itself are built as a program to obtain a backup on the disk and stored in a portable storage medium such as a disk device, floppy disk, or CD-ROM, and the database is stored in the memory. It can be used for general purposes by making it resident.
[0034]
The present invention is not limited to the above-described embodiments, and various modifications and applications can be made within the scope of the claims.
[0035]
【The invention's effect】
As described above, according to the present invention, in the T-tree index of the memory resident database management system, the T-tree body can be configured independently of the number and length of columns constituting the key value. The T-tree can be constructed at high speed because it can be handled only by changing the pointer without moving the key value between nodes.
[0036]
In addition, since the area for storing the key value can be minimized, the memory usage can be reduced. Even when a duplicate key is handled, since the key value is one, even data with a high degree of duplicate can be searched efficiently.
Furthermore, since only the T-tree body is used for backup, the time required for backup can be reduced.
[Brief description of the drawings]
FIG. 1 is a diagram for explaining the principle of the present invention.
FIG. 2 is a configuration diagram of a T-tree index according to the present invention.
FIG. 3 is a configuration diagram of nodes and key values of a T-tree index according to the present invention.
FIG. 4 is a diagram illustrating an increase in duplicate keys according to an embodiment of the present invention.
FIG. 5 is a diagram illustrating insertion of a key value according to an embodiment of the present invention.
FIG. 6 is a diagram illustrating deletion of a key value according to an embodiment of the present invention.
FIG. 7 is a diagram showing backup according to an embodiment of the present invention.
FIG. 8 is a configuration diagram of a conventional T-tree index.
FIG. 9 is a node configuration diagram in a conventional T-tree index.
FIG. 10 is a diagram showing an increase in duplicate keys in a conventional T-tree index.
FIG. 11 is a diagram illustrating insertion of a key value in a conventional T-tree index.
FIG. 12 is a diagram showing deletion of a key value in a conventional T-tree index.
FIG. 13 is a diagram showing backup of a conventional T-tree index.
[Explanation of symbols]
10 T-tree index 11 T-tree body 12 Key value storage area 13 Duplicate key storage area 20 Table 100 T-tree body storage area 111 Node 112 Intra-node element 113 Inter-node pointer 114 Intra-node element and inter-key value pointer 115 Free 116 In-node Element and inter-record pointer 121 Key value 122 Key value and inter-record pointer 123 Key value and duplicate key pointer inter-set pointer 131 Duplicate key pointer set element and inter-record pointer 132 Duplicate key pointer set inter-element and inter-record pointer 133 Duplicate key 200 Key value storage area 300 Duplicate key storage area 400 Management means 410 Backup acquisition means

Claims (9)

In a T-tree index construction method for constructing a T-tree index used in a database management system that makes a database resident in a memory and realizes high-speed data access,
The T-tree index is a T-tree body consisting of nodes having information on the hierarchical structure of the T-tree index, key value information for comparing conditions at the time of search, and duplicate key information corresponding to a plurality of records for one key. Divided into areas,
The node information of the T-tree index is constructed independently of the number of columns constituting the key, the length, and the key duplication,
A T-tree index construction method characterized in that a backup is acquired on a disk for only the nodes constituting the T-tree main body in the divided three areas. In a T-tree index construction apparatus for constructing a T-tree index used in a database management system that makes a database resident in a memory and realizes high-speed data access,
A T-tree main body storage area for storing a T-tree main body composed of nodes having information on the hierarchical structure of the T-tree index.
A key value storage area for storing key value information to be subjected to condition comparison at the time of search;
A duplicate key storage area for storing duplicate key information corresponding to a plurality of records for one key;
A T-tree index construction apparatus comprising: a management unit that manages three areas of the T-tree main body storage area, the key value storage area, and the duplicate key storage area. The T-tree main body storage area is
The T-tree index construction apparatus according to claim 2, wherein the node information of the T-tree index is constructed independently of the number of columns constituting the key, the length, and duplication of the key. The management means includes
A backup acquisition unit that acquires a backup on a disk of only the nodes constituting the T-tree main body among the three areas of the divided T-tree main body storage area, the key value storage area, and the duplicate key storage area. Item 4. The T-tree index construction device according to items 2 and 3. Nodes stored in the T-tree main body storage area are:
3. The T-tree index construction apparatus according to claim 2, wherein the in-node element includes a pointer indicating a position of a key value stored in the key value storage area. The key value stored in the key value storage area is:
When one record corresponds to one key, a pointer indicating the position of the record corresponding to each key value is held,
3. The T-tree index construction apparatus according to claim 2, wherein when a plurality of records correspond to one key, a pointer pointing to one element in the set of pointers in the duplicate key storage area is held. The duplicate key information stored in the duplicate key storage area is:
3. The T-tree index construction apparatus according to claim 2, further comprising a pointer that points to a position of a record corresponding to the duplicate key. The management means includes
8. The T-tree index construction device according to claim 2, 5, 6, and 7, comprising means for combining elements in the T-tree main body storage area, the key value storage area, and the duplicate key storage area with a pointer. A storage medium storing a T-tree index construction program for constructing a T-tree index used in a database management system that makes a database resident in a memory and realizes high-speed data access,
The T-tree index is a T-tree body consisting of nodes having information on the hierarchical structure of the T-tree index, key value information for comparing conditions at the time of search, and duplicate key information corresponding to a plurality of records for one key. It has a management process to manage by dividing into areas,
The management process is:
The node information of the T-tree index is constructed independently of the number of columns constituting the key, the length, and the duplication of the key, and only the nodes constituting the T-tree body in the three divided areas are placed on the disk. A storage medium storing a T-tree index construction program characterized by having a process for obtaining a backup.

Images