KR102013839B1 - Method and System for Managing Database, and Tree Structure for Database - Google Patents

Abstract

According to the present invention, in configuring an index of a database, an index page is stored by storing a lower limit value and an upper limit value of a key value of a plurality of records included in one page as a delimiter, and deleting duplicate key values from the plurality of records using the same. A database management method and system that saves storage space and thereby improves database performance.

Description

Method and System for Managing Database, and Tree Structure for Database

The present invention relates to a database management method, a system and a database tree structure, and more particularly, to a database management method, a system and a database tree structure using an index compression method.

Database Management System (DBMS) is a system for managing a database that stores a large amount of data. It is an indispensable part of the current era when a large amount of information is constantly created. It is recognized as an element.

In such a DBMS, all data is stored in a database in the form of a table, where a table is a basic structure for storing data in a database, and one table is composed of one or more records. Here, the record means one row of the table. In addition, each record is composed of one or more columns. A column means a domain having a name representing a table item of a real world constituting a table, and is also called an attribute or a field.

When a specific query is input from the outside, the DBMS performs a function of selecting, inserting, updating, and deleting data in the database according to the input query. In this case, the query describes what kind of request for data stored in a table of a database, that is, what kind of manipulation of the data is desired, and is expressed using a language such as SQL (Structured Query Language).

On the other hand, as the amount of data increases, the DBMS generally has an index. Here, an index means a data structure that speeds up a search for a table in a database field. Such an index is a data structure composed of {key value, pointer} pairs to quickly access a data record (tuple). Has

The background art described above is technical information possessed by the inventors for the derivation of the present invention or acquired during the derivation process of the present invention, and is not necessarily a publicly known technique disclosed to the general public before the application of the present invention.

In an embodiment of the present invention, in configuring an index of a database, a lower limit value and an upper limit value of a key value of a plurality of records included in a page are stored as a delimiter, and the duplicated portion of the key is deleted from the plurality of records by using the identifier. Its purpose is to provide a database management method, system, and database tree structure that saves storage space where index pages are stored and thereby improves database performance.

In addition, an embodiment of the present invention enables to set whether or not to compress in run-time, so that the database operation by adjusting not to perform compression when the load of insert / delete in a specific region becomes high Its purpose is to provide an improved database management method, system and database tree structure.

In addition, an embodiment of the present invention eliminates the need to additionally record metadata about secondary compression schemes and ranges, and includes meta information in records that are compressed as the number of records stored on a page increases. Its purpose is to provide a database management method, system, and database tree structure that can maximize compression efficiency compared to

In addition, an embodiment of the present invention easily checks an error of an index structure by performing validation of leaf nodes using LFK and UFK of each page whenever traversing a tree structure of an index. It aims to provide a database management method, system and database tree structure that can be done.

According to an embodiment of the present invention, a lower limit value of key values of a plurality of records included in each page is stored as a lower fence key (LFK), or an upper limit value of a key value of a record is stored as an upper fence key (UFK); Extracting a common area among key values of a plurality of records constituting the page as a prefix; And storing a part of the plurality of records except for a portion corresponding to the prefix in key values of the plurality of records.

In the present embodiment, the prefix may be stored in a lower fence key (LFK) or an upper fence key (UFK).

In the present embodiment, in the records other than the record in which the lower fence key (LFK) or the upper fence key (UFK) is stored, key values excluding the prefix among the original key values of the respective records may be stored.

In the present embodiment, the record may be a multi-column type record including a plurality of key values.

In the present embodiment, among the plurality of key values, a key value having a same value as a record forming one page may be extracted as the prefix.

In the present embodiment, each page may be a leaf node of a B tree or a B + tree structure.

In the present embodiment, the database management method further includes the step of restoring an original key value of a record included in the page, and the step of restoring the original key value includes whether the LFK and the UFK exist on the page. Is confirmed; If the LFK and the UFK exist on the page, comparing the LFK and the UFK and extracting a common prefix thereof; And restoring an original key value by combining the extracted prefix with a key value of a corresponding record.

In the present embodiment, when the LFK and the UFK do not exist in the corresponding page, the key value stored in each record may be an original key value.

In the present embodiment, the database management method further includes the step of adding a new record to the page or changing an existing record, wherein the step of adding or changing the record comprises: the LFK and UFK on the page. Checking whether there is a; If the LFK and the UFK exist on the page, comparing the LFK and the UFK and extracting a common prefix thereof; And adding or changing the remaining key values from which the prefix is excluded from the record to be added or changed to the record.

In the present embodiment, when the LFK and the UFK do not exist in the page, the prefix is not extracted, and a record may be added or changed in the page.

Another embodiment of the present invention provides a method for managing a database of a B tree or a B + tree structure, comprising: generating an index of a B tree or a B + tree structure; Restoring a predetermined record from the index; And adding or changing a predetermined record to the index, wherein the creating of the index includes: at least one end of one or more leaf nodes, wherein a lower limit of a key value belonging to the leaf node is LFK (lower fence key) A method of managing a database includes a step of storing an upper fence key (UFK) in which an upper limit value of a key value belonging to or corresponding to a leaf node is stored.

Another embodiment of the present invention includes a query analysis unit for receiving and analyzing a query that is defined together with the retrieval request of the record included in the specific table and the update request for at least one column included in the record; An execution plan generation unit generating an execution plan for executing the analyzed query statement; An execution plan execution unit that executes the execution plan by performing the withdrawal of the record and updating the at least one column according to the execution plan; And create an index for a specific table, and store the lower limit of key values of a plurality of records belonging to each page of the index as a lower fence key (LFK), or store the upper limit value of a key value of a record as an UFK (upper fence key). Disclosed is a database management system including an index management unit including an index generation unit.

In the present embodiment, the index generator may extract a common area among key values of a plurality of records constituting each page as a prefix.

In the present exemplary embodiment, the index generator may store, in the index, remaining portions of the key values of the plurality of records constituting the pages except for the portion corresponding to the prefix, which is a common area.

In the present embodiment, the prefix may be stored only in a lower fence key (LFK) or an upper fence key (UFK).

In the present embodiment, only the key values excluding the prefix among the original key values of each record may be stored in a record other than the record in which the lower fence key (LFK) or the upper fence key (UFK) is stored among the records.

In the present embodiment, the index management unit may further include a record restoration unit for restoring an original key value from a record included in each page of the index.

In the present embodiment, the record restoration unit checks whether the LFK and UFK exist in the page, and extracts the prefix by comparing and calculating the LFK and UFK when the LFK and UFK exist in the page. The original key value may be restored by combining the prefix and the value of the corresponding record.

In an embodiment, the index manager may further include a record updater that adds a new record to each page of the index or changes an existing record.

In the present embodiment, the record update unit checks whether the LFK and the UFK exist in the page, and if the LFK and the UFK exist in the page, compares the LFK and the UFK to extract the prefix, and adds the result. Alternatively, the remaining data excluding the prefix from the record to be changed may be added or changed as a record on the corresponding page.

Another embodiment of the present invention provides a database tree structure of a B tree or B + tree structure, comprising: a root node located at the top of the tree structure and storing one or more distinct key values; And at least one leaf node having a lower limit value of a key value belonging to the leaf node at least at one end as a lower fence key (LFK), or an upper limit value of a key value belonging to the leaf node having an upper fence key (UFK) stored therein. Initiate the database tree structure.

In the present embodiment, a record existing between a lower fence key (LFK) and an upper fence key (UFK) in the leaf node may store only the remaining key values except for the common area in the key value of the record.

In the present embodiment, the distinguishing key value stored in the root node may be an LFK of one side and an UFK of the other side of leaf nodes adjacent to each other.

In the present embodiment, the LFK may not be stored at the leftmost leaf node of the plurality of leaf nodes, and the UFK may not be stored at the rightmost leaf node of the plurality of leaf nodes.

Other aspects, features, and advantages other than those described above will become apparent from the following drawings, claims, and detailed description of the invention.

According to the present invention, it is possible to save a storage space for storing the index page, thereby improving the performance of the database.

In addition, an embodiment of the present invention enables to set whether or not to compress in run-time, so that the database operation by adjusting not to perform compression when the load of insert / delete in a specific region becomes high The efficiency can be improved.

In addition, an embodiment of the present invention eliminates the need to additionally record metadata about a secondary compression scheme and range, and thus, the compression efficiency may be maximized as the number of records stored in a page increases. .

In addition, an embodiment of the present invention easily checks an error of an index structure by performing validation of leaf nodes using LFK and UFK of each page whenever traversing a tree structure of an index. You can get the effect.

1 is a schematic block diagram of a database management system according to an embodiment of the present invention.
2A is a flowchart illustrating an index generation step of a database management method according to an embodiment of the present invention.
2B is a flowchart illustrating a record restoration step of a database management method according to an embodiment of the present invention.
2C is a flowchart illustrating a record addition / change step of the database management method according to an embodiment of the present invention.
3 is a diagram exemplarily illustrating a structure of a B-tree index to which a database management method and system of the present invention are applied.
4 is a diagram showing an arrangement of key values of a general B tree page.
5 is a diagram illustrating an arrangement of key values of a B-tree page according to the database management method and system of the present invention.
6A and 6B illustrate a process of dividing a page in a B-tree index to which the method and system for managing a database of the present invention are applied.

DETAILED DESCRIPTION OF THE INVENTION The following detailed description of the invention refers to the accompanying drawings that show, by way of illustration, specific embodiments in which the invention may be practiced. These embodiments are described in sufficient detail to enable those skilled in the art to practice the invention. It should be understood that the various embodiments of the present invention are different but need not be mutually exclusive. For example, certain shapes, structures, and characteristics described herein may be implemented with changes from one embodiment to another without departing from the spirit and scope of the invention. In addition, it is to be understood that the location or arrangement of individual components within each embodiment may be changed without departing from the spirit and scope of the invention. Accordingly, the following detailed description is not to be taken in a limiting sense, and the scope of the present invention should be taken as encompassing the scope of the claims of the claims and all equivalents thereto. Like reference numerals in the drawings indicate the same or similar elements throughout the several aspects.

Hereinafter, various embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those skilled in the art can easily implement the present invention.

1 is a schematic block diagram of a database management system according to an embodiment of the present invention. 1, the database management system 100 according to an embodiment of the present invention is configured as follows.

First, the database 110 stores various data in a table format. As described above, each table is composed of one or more records, and each record is composed of one or more columns. For example, in the case of a database in which posts for a predetermined bulletin board are stored, a table means a set of posts, a record means each post, and a column means an area in which a post identifier, a creator of a post, and a number of views of a post are stored. . Although a plurality of databases 110 are illustrated in the drawings, the spirit of the present invention is not limited thereto, and the number of databases 110 and the number of databases 110 may vary depending on the configuration of the database management system 100, the amount of data to be stored, and the purpose thereof. The configuration may vary.

The database management system 100 is connected to the database 110 to perform a function of integrally managing the database 110 such as updating or deleting data recorded in the database 110 or adding data to the database 110. As a large, it includes a query statement analysis unit 120, execution plan generation unit 130, execution plan execution unit 140. In addition, the database management system 100 may further include an entry manager 150 and an index manager 160.

The query analysis unit 120 receives a query for processing data stored in the database 110 from various external servers (not shown) or administrator terminals (not shown) interworking with the database management system 100. Analyze the received query. The query analysis unit 120 may include a query reception unit and a parser, and may further include a validation unit.

The execution plan generation unit 130 generates an execution plan for retrieving the requested record and updating the columns included in the record based on the pastry determined to be valid by the validity validation unit of the query statement analysis unit 120. Stored in the memory 170 to be described later. Here, the execution plan means a data structure including a method of retrieving a record from a specific table, a result record list, and whether to perform an increment operation on a column requested for updating.

In one embodiment, the execution plan generator 130 may select one of a sequential scan method and an index scan method as a method of fetching a requested record from a specific table. Here, the sequential scan method means a method of fetching a record having an identifier of the corresponding record while sequentially scanning the records included in a specific table, and the index scan method is an index for each record identifier, such an index By scanning only means how to fetch that record. The index of such a DBMS will be described in detail later.

The execution plan execution unit 140 fetches the corresponding record from a specific table according to the execution plan generated by the execution plan generation unit 130, and records the column value recorded in the column on the record corresponding to the physical position of the requested column for update. Update the column value by performing an increment operation. Specifically, the execution plan execution unit 140 processes the execution plan generated by generating a transaction for executing the execution plan generated by the execution plan generation unit 130 during the transaction. Here, a transaction constitutes one logical unit of work and is defined using one or more SQL statements. The use of these transactions ensures data consistency and data concurrency.

The database management system 100 generates or deletes an entry consisting of an identifier of a record and an identifier of a column to be updated, and matches the generated entry with a column value corresponding to a column identifier included in the entry to the memory 170. The apparatus may further include an entry manager 150 to store the data, and the column 170 of the column requested to be updated may be matched with an entry generated in the entry manager 150 in the memory 170.

The database management system 100 may further include an index manager 160 that generates or deletes an index and stores the generated index in the memory 170. The index manager 160 may include an index generator 161, a record restorer 163, and a record updater 165.

The index generator 161 generates an index for the specific database 110. At this time, the lower limit of the key value belonging to each page of the index is stored as a lower fence key (LFK), and the upper limit of the key value is an upper fence key (UFK). Serves as a store. In addition, the index generator 161 extracts a common area among key values of a plurality of records constituting each page as a prefix. In addition, the index generator 161 deletes a portion corresponding to a prefix from key values of a plurality of records constituting each page and stores the index.

The record restoration unit 163 restores the original key value from the record included in each page of the index. In detail, the record restoring unit 163 checks whether LFK and UFK exist on the page, and extracts the prefix by comparing and calculating the LFK and UFK when the LFK and UFK exist on the page. Restores the original key value by combining the values of.

The record updating unit 165 adds a new record to each page of the index or changes an existing record. In detail, the record update unit 165 checks whether the LFK and UFK exist in the page, extracts the prefix by comparing the LFK and UFK when the LFK and UFK exist in the page, and in the record to be added or changed. It adds or changes the rest of the data without the prefix as a record on the page.

Such an index generation process of the index generator 161, a record restoration process of the record restorer 163, and a record update process of the record updater 165 will be described in detail with reference to FIG. 2.

Hereinafter, a method of managing a database using a separator-based index compression method will be described in detail.

2A, 2B and 2C are flowcharts illustrating a database management method according to an embodiment of the present invention. 3 is a diagram illustrating a structure of a B-tree index to which the database management method and system of the present invention are applied. FIG. 4 is a diagram illustrating an arrangement of key values of a general B-tree page. It is a figure which shows the arrangement of the key value of B-tree page according to the database management method and system of this invention.

2 to 5, the method for managing a database according to an embodiment of the present invention includes creating an index (S100), restoring a record from the index (S200), and adding / changing a record from the index. It includes a step (step S300).

This will be described in more detail as follows.

An index is a data structure that speeds up operations on a table in the database field. Indexes can be created using a single column or multiple coulmn indexes in a table, providing the basis for efficient ordering operations in relation to record access as well as fast search operations. do. The disk space required to store the index is usually less than the disk space required to store the table. This is because an index usually has only key-fields and no other details of the table.

The B tree (or B + tree) is a tree data structure widely used in databases and file systems to construct such indexes. It is an associative data structure for quickly retrieving records with specific keys. The B tree (or B + tree) has a page structure in order to reduce the number of I / Os because data is recorded on a large disk with slow access. In one page, a key and the position (Object ID or Record ID) of the actual record containing the key as an attribute are recorded in the order of the keys. That is, {Key-OID} combine to form each index record (hereinafter, referred to as a record).

Due to the above characteristics, the key values of a record in a page can have considerable similarity between adjacent values. For example, if a company's mail system specifies a group number, a mail folder number, and a mail serial number as an index key for distinguishing each mail (that is, a multi-column index), an employee has 100,000 total mails and one If you have 10,000 mails in a mail folder, 10,000 indexes have the same number and the same folder number, and 100,000 records have the same number. If a page of a typical B tree contains 1,000 entries in the conventional way, about 10 pages will store duplicate number and mail folder number each time. Therefore, there is a problem in that unnecessary memory waste occurs because the same value is repeatedly stored several times.

In order to solve such a problem, the database management method and system using the Separator-Based Index Compression Method according to the present invention, a page delimiter to the prefix (prefix) stored among the key values of one page In other words, it is stored in a lower fence key (LFK) and an upper fence key (UFK), thereby saving memory space.

Referring back to FIG. 2, in the step of generating an index in the database management method according to an embodiment of the present invention (step S100), a lower limit value of a key value belonging to one page is stored as a lower fence key (LFK), or The upper limit value of a key value belonging to one page is stored as an upper fence key (UFK) (step S110), and the common area among key values of a plurality of records constituting one page is extracted as a prefix (step S120). And a step in which a prefix, which is a common area, is deleted from key values of a plurality of records constituting one page, and then stored (step S130).

If this is explained easily, it is as follows.

Referring to FIG. 3 showing the structure of the B-tree index to which the present invention is applied, in the index used for quickly retrieving record data from a database, the structure of the B-tree index used in the present invention is a leaf indicating actual record data. It consists of nodes and leaf nodes. The root node is one node at the top of the intermediate nodes. In FIG. 3, there are four leaf nodes and one intermediate node, and one intermediate node becomes the root node. At this time, each leaf node constitutes each page. That is, the four leaf nodes shown in FIG. 3 constitute four pages.

Here, the root node of the B-tree index shown in FIG. 3 has three partition key values P1, P2, and P3. The first separator key value P1 becomes a separator for separating page 1 and page 2, the second separator key value P2 becomes a separator for separating pages 2 and page 3, and the third separator key value P3 is page 3 It is a delimiter between page 4 and.

As described above, the database management method according to an exemplary embodiment of the present invention stores the lower limit of the key value belonging to each page as a lower fence key (LFK), and also stores the upper limit of the key value belonging to each page as an UFK (upper fence key). It is characterized by storing as. In this case, the lower fence key (LFK) is stored at the left end of each page when viewed in the drawing to define the lower limit of the page. The upper fence key (UFK) is stored at the right end of each page when viewed in the drawing. Is to define the upper limit of.

The database management method according to an embodiment of the present invention extracts a common area as a prefix among key values of a plurality of records existing between a lower fence key (LFK) and an upper fence key (UFK) on each page. In other records except the lower fence key (LFK) and the upper fence key (UFK), the prefix is deleted from the key value, and only the remaining key value is saved, thereby saving memory space by preventing duplicate data storage.

In other words, the first division key value P1 becomes UFK (UFK1) on page 1 and LFK (LFK2) on page 2. Similarly, the second division key value P2 becomes UFK (UFK2) on page 2 and LFK (LFK3) on page 3. Similarly, the third division key value P3 becomes UFK (UFK3) on page 3 and LFK (LFK4) on page 4. At this time, the leftmost leaf node (page 1) located on the far left of the leaf nodes cannot set the lower fence key (LFK), and therefore, the prefix extraction cannot be performed on the page. Likewise, you can't set the upper fence key (UFK) on the far right leaf node (page 4), so you can't extract the prefix from that page.

To illustrate this in more detail, FIG. 4 shows an arrangement of key values of a general B tree page and FIG. 5 shows an arrangement of key values of a B tree page according to a database management method and system of the present invention. In FIG. 4 and FIG. 5, it is assumed that there are only 10 records for explanation.

Referring to FIG. 4, the same value is duplicated and stored ten times in the ten records constituting one page with the number KR10000 and the mail folder number FD0001. On the contrary, referring to FIG. 5, according to the database management method of the present invention, the number (KR10000) and the mail folder number (FD0001), which are duplicated and stored among the key values of one page, set the lower limit value of the corresponding page. It is stored only in UFK (Upper Fence Key) which sets the Fence Key) and the upper limit of the page. Only the key value with the prefix deleted is stored in the general record between LFK and UFK. That is, as shown in FIG. 5, all of the key number KR10000 and the mail folder number FD0001 which are common to ten records are deleted, and only the mail serial number which is a unique key value is stored in each record.

Hereinafter, a description will be given of the step of restoring a record in the database management method according to an embodiment of the present invention. 2 and 5 again, the step of restoring a record in the method of managing a database according to an embodiment of the present invention (step S200) may include checking whether LFK and UFK exist on a corresponding page (step S210). ), If LFK and UFK exist on the page, extracting the prefix by comparing LFK and UFK (step S220) and restoring the original key value by combining the prefix and the value of the corresponding record (step S230). do. This will be described in more detail as follows.

First, find out which page the page to which the key value to restore belongs belongs by binary search, and then check whether the LFK and UFK exist in the page. If either LFK or UFK does not exist on the page, the page is not compressed (i.e., duplicate data deletion using a prefix), so the key value stored in each record is the original key value (S240). step). On the other hand, if both LFK and UFK exist on the page, the page is a page on which data compression is performed using a prefix, and thus a predetermined restoration routine is executed. That is, LFK, which is the lower limit of the page, and UFK, which is the upper limit of the page, are compared to extract a common area, that is, a prefix, between the LFK and the UFK. In the example of FIG. 5, KR10000: FD0001, which is a common region of LFK and UFK, will be extracted as a prefix. The original key value is restored by combining the extracted prefix and the key value stored in each record. That is, the original key value which restored SN10001 is KR10000: FD0001: SN10001.

Hereinafter, a step of adding or changing a record in a database management method according to an embodiment of the present invention will be described. 2 and 5 again, in the step of adding or changing a record in the method of managing a database according to an embodiment of the present invention (step S300), a step of checking whether LFK and UFK exist in a corresponding page ( Step S310), if LFK and UFK exist on the page, extracting the prefix by comparing LFK and UFK (step S320), and the remaining data excluding the prefix from the record to be added or changed are added to the page as a record. Or change (step S330). This will be described in more detail as follows.

First, find out which page the key value to which you want to add or change belongs by binary search, and then check whether LFK and UFK exist in the page. If either LFK or UFK does not exist on the page, the page has not been compressed (i.e. duplicate data deletion using a prefix), so the value of the record can be added or changed immediately without further processing. (Step S340). On the other hand, if both LFK and UFK exist on the page, since the page is a page on which data compression is performed using a prefix, a predetermined decomposition routine is executed. That is, LFK, which is the lower limit of the page, and UFK, which is the upper limit of the page, are compared to extract a common area, that is, a prefix, between the LFK and the UFK. In the example of FIG. 5, KR10000: FD0001, which is a common region of LFK and UFK, will be extracted as a prefix. Next, only key values other than the prefix in the key value of the record to be added or changed are added or changed at the corresponding position on the page. For example, if the added key value is KR10000: FD0001: SN13000, the prefix is KR10000: FD0001, so that only SN13000 is added to the page without the prefix.

Hereinafter, a process of splitting or merging pages in a database management method according to an embodiment of the present invention will be described.

6A and 6B illustrate a process of dividing a page in a B-tree index to which the method and system for managing a database of the present invention are applied.

The root node of the B-tree index shown in FIG. 6A initially has two distinct key values P1 and P2. The first division key value P1 is a delimiter for dividing page 1 and page 2, and the second division key value P2 is a delimiter for dividing page 2 and page 3. Then, the first division key value P1 becomes UFK (UFK1) on page 1 and LFK (LFK2) on page 2. Similarly, the second division key value P2 becomes UFK (UFK2) on page 2 and LFK (LFK3) on page 3.

In this state, it is assumed that there is no more storage space on page 2, so that page 2 should be divided based on S, which is an arbitrary value between the first division key value P1 and the second division key value P2.

In this case, page 2 stores the reconstructed S as a new UFK (UFK2) as shown in FIG. 6B. Then, the data is compressed again using the existing LFK (LFK2) and the newly generated UFK (UFK2). That is, after extracting the prefix by comparing and calculating LFK2 and UFK2, the remaining records except for LFK2 and UFK2 are deleted from the key value and only the remaining key value is stored.

On the other hand, newly created page 3 of FIG. 6B stores S as a new LFK (LFK3). Then, UFK (UFK2) of page 2 of FIG. 6A is stored as UFK (UFK3) of page 3 of FIG. 6B. Then, data is compressed again using LFK3 and UFK3. In other words, after extracting the prefix by comparing and calculating LFK3 and UFK3, the remaining records except the LFK3 and UFK3 are deleted from the key value and only the remaining key value is stored.

Finally, in page 3 of FIG. 6B, S is inserted as a new division key value between the first division key value P1 and the second division key value P2 of the parent node (here, the root node).

On the other hand, although not shown in the drawing, if the page is not compressed (i.e., a page in which duplicate data deletion is not performed using the prefix), when the page is divided based on the new separator key value S, a new separator key value is added to the existing page. S) is set to UFK, and data can be newly compressed using the existing LFK and the newly generated UFK. On the other hand, in the newly created page, the new separator key value S is set to LFK, but since UFK is not set, data compression will not be performed.

Also, although not shown in the drawing, when two neighboring pages are of a mergeable size (that is, the sum of the number of records stored in two pages is less than or equal to the maximum number of records that can be stored in one page). In this case, it may be possible to merge the two pages. In this case, if both pages are compressed pages (that is, pages in which duplicate data is deleted using a prefix), compression is newly performed after merging two pages. On the other hand, if any one of the two pages is not a compressed page, merging is not performed.

Table 1 shows the total number of index pages when a general database management method and system are applied, and the total number of index pages when a database management method and system is applied according to an embodiment of the present invention. This is a comparison table. Table 2 shows average key lengths when a general database management method and system are applied and average key values when a database management method and system according to an embodiment of the present invention are applied. This is a comparison table.

Index name existing
Total number of index pages When applying the present invention
Index page
% Reduction Primary key (PK) 4,120,635 21 Index 1 5,740,191 23 Index 2 6,056,621 18 Index 3 6,849,043 19 Index 4 5,793,081 15

Index name existing
Average key value length (bytes) When applying the present invention
Average key value length
% Reduction Primary key (PK) 18 39 Index 1 30 33 Index 2 32 28 Index 3 38 29 Index 4 30 27

As shown in Table 1 and Table 2, when the database management method and system according to an embodiment of the present invention is applied, the total number of index pages is reduced by about 19% compared with the conventional database management method and system. The average height was reduced by about 31%. That is, the present invention can save the storage space, it can be seen that the performance of the database can be improved accordingly.

The database management method according to an exemplary embodiment of the present invention has a fence key (LFK and LFK) at both ends of a page using a key value used as a separator of the page within a page of the B tree as a {virtual key-OID} record. UFK), respectively, can quickly perform a decomposition operation of combining the original key value from the LFK and the UFK and deleting the prefix from the original key value.

On the other hand, if there is no LFK or UFK on that page, you can store records in the traditional way, so that compressed pages (that is, pages whose prefixes have been deleted from each record) and uncompressed pages (that is, each record has a prefix) Pages not deleted) may be mixed. In addition, compressed and uncompressed records may be mixed in a page in which LFK and UFK exist. Therefore, it is possible to dynamically set whether or not to compress in real time (run-time) regardless of the current state of the B-tree. In addition, by using such a feature, when an insert / delete load is increased in a specific region, compression may be adjusted so as to improve efficiency of database operation.

On the other hand, since it is determined whether the compression using the prefix using LFK and UFK, it is not necessary to additionally record metadata about the secondary compression scheme and range. Therefore, as the number of records stored in the page increases, the compression efficiency is maximized as compared with the conventional method of including meta information in the compressed record.

According to the present invention, since only the key value excluding the prefix is stored, more records can be stored in one page of the B tree. This saves storage space and increases the performance of the database by increasing the possibility of being included in the buffer cache of main memory. Furthermore, the database management method according to an embodiment of the present invention is simpler in partitioning and restoring than other compression methods, and is advantageous in terms of backward compatibility because there is no change in storage structure, and only one flag indicating LFK and UFK is added. It has the advantage of being structurally simple because there is no additional recorded metadata.

Furthermore, using the database management method of the present invention, each time the traversal of the tree structure of the index can be performed, the validity check of the leaf node can be performed by using the LFK and UFK of each page. You can easily check the error of. In addition, since the connection relationship between the leaf nodes can be determined by the LFK and the UFK, the link can be removed from the leaf nodes, thereby improving the performance and efficiency of the structure modification operation (SMO).

The above-described database management method may be implemented in the form of a program that can be executed using various computer means. In this case, the program for performing the database management method may be a hard disk, a CD-ROM, a DVD, a ROM, or a RAM. Or a computer-readable recording medium such as a flash memory.

In the present specification, the present invention has been described with reference to limited embodiments, but various embodiments are possible within the scope of the present invention. In addition, although not described, equivalent means will also be referred to as incorporated in the present invention. Therefore, the true scope of the present invention will be defined by the claims below.

100: database management system
110: database
120: query analysis unit
130: execution plan generation unit
140: execution plan execution unit
150: entry management unit
160: index management unit
161: index generator
163: record restore unit
165: record update unit

Claims (15)

Storing a lower limit of key values of a plurality of records included in each page as a lower fence key (LFK), or storing an upper limit of a key value of a record as an upper fence key (UFK);
Extracting a common area among key values of a plurality of records constituting the page as a prefix; And
And storing portions other than portions corresponding to the prefix in key values of the plurality of records.
The LFK and the UFK are original key values including the prefix,

Adding new records to the page or changing existing records;
The step of adding or changing the record,
Checking whether the LFK and the UFK exist on a corresponding page;
If the LFK and the UFK exist on the page, comparing the LFK and the UFK and extracting a common prefix thereof; And
And adding or changing the remaining key values excluding the prefix from the record to be added or changed to the record. delete The method of claim 1,
And a key value excluding the prefix among the original key values of each record is stored in a record other than a record in which a lower fence key (LFK) or an upper fence key (UFK) is stored among the plurality of records. The method of claim 1,
And the record is a multi-column type record including a plurality of key values. delete delete delete The method of claim 1,
If the LFK and UFK do not exist on the page,
And the record is added or changed in the corresponding page without the prefix being extracted. A recording medium having recorded thereon a program for performing the method according to any one of claims 1, 3, 4 and 8. A query analysis unit configured to receive and analyze a query statement including a drawing request of a record included in a specific table and an update request for at least one column included in the record;
An execution plan generator for generating an execution plan for executing the analyzed query statement;
An execution plan execution unit that executes the execution plan by performing withdrawal of the record and updating the at least one column according to the execution plan; And
Create an index for a specific table, and store the lower limit of the key value of a plurality of records belonging to each page of the index as a lower fence key (LFK), or store the upper limit of a key value of a record as an UFK (upper fence key). And an index manager including an index generator for extracting a common area as a prefix among key values of a plurality of records constituting each page.
The LFK and the UFK are original key values including the prefix,
The index management unit,
The apparatus may further include a record updating unit for adding a new record to each page of the index or changing an existing record.
The record update unit,
Check whether the LFK and UFK exist in the page, and if the LFK and UFK exist in the page, compare the LFK and UFK to extract the prefix, and remove the prefix from the record to be added or changed. A database management system for adding or changing data as records on a corresponding page. The method of claim 10,
The index generating unit stores a remaining portion of the key value of the plurality of records constituting each page in the index except for a portion corresponding to the prefix, which is a common area. The method of claim 10,
The prefix is stored only in the lower fence key (LFK) or upper fence key (UFK). The method of claim 12,
And in the records other than the record in which the lower fence key (LFK) or the upper fence key (UFK) is stored among the plurality of records, only key values excluding the prefix among the original key values of each record are stored. delete delete

Images