JP2013242674A - Database control device, b-tree index processing method, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce frequency of random key insertion operation in a B-tree index of each database control device while dispersing loads in a plurality of database control devices to an access into which new record groups are continuously inserted.SOLUTION: A database control device which stores data in a storage unit includes: prefix generation means for generating a prefix to be given to a key of a B-tree index stored in the storge unit of the data control device, by using a registration record and registrant information of the record as input, when registering the record in a database; index registration means for registering the registration record in the B-tree index by using a synthetic key in which the prefix is given to the key of the B-tree index; and index retrieval means for performing retrieval to the B-tree index by using the synthetic key as a retrieval key for input when retrieving the B-tree index.

Description

  The present invention relates to a technique for a method in which a database management device uses a Btree index in a distributed database device in which a distribution method is a hash distribution method among distributed database devices in which a plurality of database management devices are distributed. is there.

  In a conventional distributed database device, when a new record is inserted, a database management device that manages the record is first assigned (routed) from among a plurality of database management devices.

  Then, in the assigned database management device, when storing the inserted record, the record is inserted into the Btree index by using the key of the record designated for index construction.

  When the distribution method is a hash distribution method, a hash value of a specific column value of a record is used at the time of routing for load distribution. For example, in the case of a mail service or blog system, an example of distributing data for each user by using a hash value of a user ID is given.

  In addition, as a performance characteristic of Btree index in the database management device, random index key insertion operation causes random access in the disk device and the performance deteriorates, whereas sequential key insertion operation in the disk device It is known that the performance is higher because of sequential access. This is illustrated in FIG.

  The design policy of the distributed database based on the above-mentioned performance characteristics of the prior art is as follows.

  In order to improve the performance of the entire distributed database apparatus from the viewpoint of load distribution, the data is distributed based on the hash so that the records to be inserted into each database management apparatus are evenly allocated.

  In each database management device, in order to improve the performance of Btree, the index key of the insertion record is designed so that the sequential key insertion operation increases.

  However, in the distributed database device using the hash distribution method, it is difficult to satisfy both of the above two points. The reason is that, in general, the hash input key is used as the Btree key during distribution, so when using a key in which records are evenly distributed among multiple database management devices, the Btree of each database management device This is because a random index key insertion operation occurs.

  For example, in the case of a mail service or blog system, when routing is performed to a database management device distributed by the hash value using the user ID as a key, and the user ID is used as a key in the Btree of each database device, multiple different uses As a result of the record insertion operation by the user, there is a problem that random access frequently occurs in the disk device and the performance of the Btree index deteriorates.

  On the other hand, when using a key that allows a sequential index key insertion operation to be performed in the Btree of each database management device, multiple records are continuously assigned to the same database management device in multiple database management devices. The problem that the load is not distributed arises.

Modern B-Tree Techniques, Goetz Graefe (Author), Publisher: Now Publishers (2011/08), ISBN-10: 1601984820, ISBN-13: 978-1601984821

  The present invention has been made in view of the above points, and for the access in which a new record group is continuously inserted, the load on a plurality of database management devices is distributed and at the same time the Btree index of each database management device. It is an object of the present invention to provide a database management apparatus in a distributed database apparatus using a hash distribution method, which realizes high-speed record insertion by reducing the number of random key insertion operations.

In order to solve the above problems, the present invention provides a database management apparatus for storing data in a storage device,
When registering a record in a database, a prefix generation means for generating a prefix to be given to a key of a Btree index stored in a storage device of the database management device, using a registration record and registrant information of the record as input,
Index registration means for registering a registration record in a Btree index using a composite key with the prefix added to the key of the Btree index;
When a Btree index is searched, the database management apparatus includes index search means for searching the Btree index using a synthetic key as a search key.

The present invention is also a database management device for storing data in a storage device,
When registering a record in the database, the registration generation and the registrant information of the record are input, and a prefix generation means for registration is generated that generates a prefix to be added to the key of the Btree index stored in the storage device of the database management device When,
Index registration means for registering a registration record in a Btree index using a composite key in which a prefix generated by the prefix generation means for registration is added to the key of the Btree index;
When searching a Btree index, a search key and a search target registrant information as input, and a prefix generation means for search that generates a prefix,
You may comprise as a database management apparatus which has an index search means which searches with respect to a Btree index using the synthetic | combination key which provided the prefix produced | generated by the said prefix generation means for the said search key to the said search key.

  The prefix generation means is configured to generate a prefix so that a composite key obtained by adding a prefix in a plurality of registered records is continuous in a Btree index.

  The prefix generation means may be configured to cluster keys registered in an index in a record group to be registered, assign an ID to each cluster, and generate the obtained cluster ID as a prefix.

  In addition, the present invention may be configured as a Btree index processing method executed by the database management apparatus, and may further be configured as a program for causing a computer to function as each unit in the database management apparatus.

  According to the present invention, for accesses in which new record groups are continuously inserted, the load on a plurality of database management devices is distributed, and at the same time, the number of random key insertion operations in the Btree index of each database management device is reduced. By doing so, it is possible to provide a database management apparatus in a distributed database apparatus of the hash distribution system that realizes high-speed record insertion.

It is a block diagram of the database management apparatus 10 in embodiment of this invention. It is a figure explaining the control flow of record registration operation in an embodiment of the invention. It is a figure explaining the control flow of the record search operation in embodiment of this invention. It is a figure which shows the processing flow of the prefix production | generation means 21 and 31. FIG. It is a figure which shows the 1st specific example in tweet data management. It is a figure which shows the 2nd specific example in tweet data management. It is a graph which shows the performance characteristic of a Btree index.

  Embodiments of the present invention will be described below with reference to the drawings. The embodiment described below is only an example, and the embodiment to which the present invention is applied is not limited to the following embodiment.

(Outline of the embodiment)
First, an outline of an embodiment of the present invention will be described. In the present embodiment, in the Btree index of the database management device, a value common to a plurality of records to be inserted into the index is calculated, and a composite key in which the value is added as a prefix to the index key value is created, and the composite key is Use to index records.

  With this configuration, it is possible to reduce the number of random key insertion operations compared to the conventional Btree index of a database management apparatus for operations in which record groups are continuously inserted.

  When searching for a Btree, the search may be performed by entering a composite key as a search key, or by using the search key and registrant information to be searched as input and obtaining a prefix by means of prefix generation. It is also possible to realize a simple index search.

  In the present embodiment, the prefix generation means generates a prefix so that a composite key obtained by adding a prefix in a plurality of registered records is continuous in a Btree index.

  When a plurality of records are continuously registered in the Btree index, the number of random key insertion operations can be reduced by registering the index using a composite key with the prefix added to the key.

  In addition, when searching for a Btree index, the prefix generation means derives a prefix by using a search key and registrant information to be searched as input, and uses a composite key in which the prefix is added to the search key. Can be accurately searched.

  Therefore, for accesses to which a large number of record groups are inserted, the load on a plurality of database management devices is distributed, and at the same time, the number of random key insertion operations in the Btree index of each database management device is reduced, thereby achieving high speed Record insertion can be realized.

(Details of the embodiment)
Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

<Device configuration>
FIG. 1 shows a functional configuration diagram of the database management apparatus 10 according to the present embodiment. As shown in FIG. 1, the database management apparatus 10 in the present embodiment includes a record registration unit 11, a record search unit 12, a record management unit 13, and a prefix generation rule management unit 14. The database management apparatus 10 of this embodiment is a single database management apparatus that constitutes a distributed database apparatus (system) in which a plurality of database management apparatuses are distributed and arranged.

  The data operator inputs a registration command or a search command to the database management device 10 via the distributed database management device. In FIG. 1, the description of the distributed database management apparatus is omitted.

  The record registration unit 11 is a functional unit that processes a registration command from the data operator. The record search unit 12 is a functional unit that processes a search command from the data operator. The record management unit 13 includes a database for storing record groups in the main memory and the hard disk device, and a Btree index constructed for the record groups. Since a Btree is constructed by specifying a key of a record group, a plurality of Btree indexes may be constructed with different keys for the same record group.

  The prefix generation rule management unit 14 holds and manages a rule that defines how a prefix is generated for each Btree index.

  The database management apparatus 10 according to the present embodiment can be realized by causing a computer including a CPU, a memory, a hard disk, and the like to execute a program corresponding to the processing of each functional unit. The program may be stored in a storage medium such as a portable memory and distributed, installed on the computer, or downloaded from a server on a network and installed on the computer.

<Detailed configuration and operation of the device>
Hereinafter, the configuration and operation of the database management apparatus 10 according to the present embodiment will be described in more detail.

  With reference to FIG. 2, the control flow of the record registration operation in the database management apparatus 10 of this Embodiment is demonstrated. As shown in FIG. 2, the record registration unit 11 includes a prefix generation unit 21 and an index registration unit 22.

  First, the data operator inputs a record registration operation to the database management apparatus 10 via the client program 20 (step 21).

  The registration command includes a registration record and registrant information that registered the record. The registration command input to the database management apparatus 10 is input to the prefix generation unit 21 and the index registration unit 22 of the record registration unit 11 (step 22).

  The prefix generation unit 21 specifies the Btree index related to the main registration operation using the input registration record information, and acquires the prefix generation rule regarding the corresponding Btree index from the prefix generation rule management unit 14 (step 23). 24). The prefix generation rule management unit 14 holds a prefix generation rule for inputting a registration record or registrant information of a record and outputting a prefix for each Btree.

  The prefix generation unit 21 generates and outputs a prefix by applying the registration record and the registrant information of the record to the obtained prefix generation rule (step 25).

  The index registration means 22 derives a composite key from the input registration record and prefix, and registers the registration record in the index and database stored in the record management unit 13 using this composite key (step 26).

  Next, with reference to FIG. 3, the control flow of the record search operation in the database management apparatus 10 of this embodiment will be described. As shown in FIG. 3, the record search unit 12 includes a prefix generation unit 31 and an index search unit 32. Note that the record search unit 12 may use the prefix generation unit 21 of the record registration unit 11 instead of the record search unit 12 including the prefix generation unit 31.

  A data operator inputs a record search command to the database management apparatus 10 via the client program 20 (step 31). The search command includes a search key and search target registrant information.

  The search command input to the database management apparatus 10 is input to the prefix generation unit 31 and the index search unit 32 of the record search unit 12 (step 32). The prefix generation unit 31 specifies the Btree index related to this search command using the input search key information, and acquires the prefix generation rule related to the corresponding Btree index from the prefix generation rule management unit 14 (step 33). 34). The prefix generation rule management unit 14 also holds a prefix generation rule that inputs a search key and search target registrant information and outputs a prefix for each Btree.

  The prefix generation unit 31 generates and outputs a prefix by applying a search key and search target registrant information to the obtained prefix generation rule (step 35).

  The index search means 32 derives a composite key from the input search key and prefix, and issues a search command to the index stored in the record management unit 13 using this composite key (step 36). A search result record is acquired (step 37).

  The search result record obtained as a result is returned to the data operator via the client program 20 (step 38).

  Instead of providing the prefix generation means 31, a composite key may be input as a search key, and the index search means 32 may perform a search using this composite key.

  Next, the processing flow of the prefix generation means 21 and 31 will be described with reference to FIG.

  In step 41, the prefix generation means 21 corresponding to the record registration specifies the Btree index related to the main registration operation using the input registration record information. Then, a prefix generation rule regarding the corresponding Btree index is acquired from the prefix generation rule management unit 14.

  In step 42, the prefix generation means 21 applies the prefix generation rule specified in step 41 using the registration record and the registrant information of the record, and outputs the prefix.

  In step 41, the prefix generation means 31 corresponding to the record search specifies the Btree index related to the main search operation using the search key information. Then, a prefix generation rule regarding the corresponding Btree index is acquired from the prefix generation rule management unit.

  In step 42, the prefix generation means 31 applies the prefix generation rule specified in step 41 using the search key and search target registrant information, and outputs a prefix.

(Concrete example)
Hereinafter, specific examples will be described with reference to the drawings.

FIG. 5 is a diagram showing a first specific example in tweet (registered trademark) data management (“tweet” described in the following description is a registered trademark). As shown in FIG. 5, the database schema in this example is
(Tweet ID, tweet creator ID, timezone, timestamp, tweet content, {hashtag},…)
The tweet ID is the primary key. One Btree index is built, and its key is a multi-column index of (tweet creator ID, tweet ID).

In this example, the timezone to which the tweet creator belongs is used as a prefix. The prefix generation rule at this time is
-When registering a record, the time zone is extracted from the registrant information of the record.-When searching for a record, the time zone is extracted from the registrant information to be searched.

As a result, the composite key obtained by adding the prefix to the index key is
(timezone + Tweet creator ID, Tweet ID)
It becomes. Here, "+" is an operation that concatenates the timezone character value and the tweet creator ID value. A record is registered in the index when registering a record using the synthetic key thus obtained.

On the other hand, when searching for a record, a record search is executed using an index using a composite key. For example, the following two search commands Search command 1: Search key = (tweet creator ID)
Search command 2: Search key = (tweet creator ID, tweet ID)
Search command 1: Composite key = (timezone + Tweet creator ID)
Search instruction 2: Composite key = (timezone + Tweet creator ID, Tweet ID)
And record search is performed using the index.

  FIG. 6 is a diagram showing a second specific example in tweet data management. In the second specific example, the difference from FIG. 5 is the location of the prefix. In this example, as a prefix, users are clustered in advance to extract communities to which a plurality of users belong, and IDs are assigned to the communities. By adding this community ID to the registrant information of the record or the registrant information to be searched, the community ID is used as a prefix as in the time zone of FIG.

(Summary of the embodiment, effects)
As described above, by registering a record in the Btree index using a composite key with a prefix added, for example, in tweet data management, sequential keys of records registered from multiple users belonging to the same timezone such as Tokyo are registered. It becomes an insert operation. As a result, the load in the plurality of database management devices is distributed, and at the same time, the record registration operation for the index becomes a sequential access in the disk device, and as a result, high-speed record registration can be realized.

  The present invention is not limited to the above-described embodiments, and various modifications and applications are possible within the scope of the claims.

DESCRIPTION OF SYMBOLS 10 Database management apparatus 11 Record registration part 12 Record search part 13 Record management part 14 Prefix generation rule management parts 21 and 31 Prefix generation means 22 Index registration means 32 Index search means

Claims (7)

A database management device for storing data in a storage device,
When registering a record in a database, a prefix generation means for generating a prefix to be given to a key of a Btree index stored in a storage device of the database management device, using a registration record and registrant information of the record as input,
Index registration means for registering a registration record in a Btree index using a composite key with the prefix added to the key of the Btree index;
When searching a Btree index, an index search means for performing a search on the Btree index using a synthetic key as a search key,
A database management device. A database management device for storing data in a storage device,
When registering a record in the database, the registration generation and the registrant information of the record are input, and a prefix generation means for registration is generated that generates a prefix to be added to the key of the Btree index stored in the storage device of the database management device When,
Index registration means for registering a registration record in a Btree index using a composite key in which a prefix generated by the prefix generation means for registration is added to the key of the Btree index;
When searching a Btree index, a search key and a search target registrant information as input, and a prefix generation means for search that generates a prefix,
Index search means for performing a search on a Btree index using a composite key with a prefix generated by the search prefix generation means added to the search key;
A database management device. The database management device according to claim 1, wherein the prefix generation unit generates a prefix so that a composite key obtained by adding a prefix in a plurality of registered records is continuous in a Btree index. The prefix generation unit clusters keys registered in an index in a record group to be registered, assigns an ID to each cluster, and generates the obtained cluster ID as a prefix. The database management device described in 1. A Btree index processing method executed by a database management device that stores data in a storage device,
When registering a record in the database, a prefix generation step for generating a prefix to be given to a key of a Btree index stored in the storage device of the database management device, using the registration record and the registrant information of the record as input,
An index registration step of registering a registration record in the Btree index using a composite key with the prefix added to the key of the Btree index;
When searching a Btree index, an index search step for performing a search on the Btree index using a synthetic key as a search key,
Btree index processing method. A Btree index processing method executed by a database management device that stores data in a storage device,
When registering a record in a database, a registration prefix and a registrant information of the record are input, and a prefix generation step at the time of generating a prefix to be given to a key of a Btree index stored in the storage device of the database management device; ,
An index registration step of registering a registration record in the Btree index using a composite key in which the prefix generated by the prefix generation step at the time of registration is added to the key of the Btree index;
When searching a Btree index, using a search key and search target registrant information as input, a prefix generation step at the time of generating a prefix,
An index search step for performing a search for a Btree index using a composite key to which the prefix generated by the prefix generation step at the time of search is added to the search key;
Btree index processing method.   The program for functioning a computer as each means in the database management apparatus described in any one of Claims 1 thru | or 4.

Images