JPH03118645A - Exclusive control method for data base - Google Patents

Abstract

PURPOSE:To attain a consistent transaction process by registering previously the data identifiers into a data base information table and using these identifiers to register data into an exclusive control information table to perform simultaneously the addition of data to the table and the retrieval of the table. CONSTITUTION:The record identifiers R1 - R4 correspond to the records 1 - 4 respectively and an identifier R5 is prepared for the records to be added to a table A. A transaction executing part 13 obtains a record identifier out of a data base information table 14 to use the identifier for process of data when a transaction is inputted from a terminal 17. The inquiries are given to an exclusive control part 11 to decide whether the reference, the update and the addition are possible or not to secure the consistency of the transaction. The part 11 searches both tables 12 and 14 based on the transaction identifier, a table identifier, a record identifier, and the process type received from the part 13. At the same time, the part 11 performs the deciding operations based on a flow and informs the part 13 of these deciding results. Thus it is possible to attain a data base process that has high performance for simultaneous execution of transactions.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention relates to a database processing device, and in particular, to a database processing device that is used for operations that frequently use the same database in multiple transactions, and in particular for operations that frequently involve adding data to the database. This invention relates to an exclusive control method that increases the concurrency of transactions in business operations where transaction consistency must be guaranteed.

[Conventional technology]

In a database processing device, exclusive control technology guarantees the consistency of all transactions while simultaneously processing the same database with multiple transactions.
It is an important technology for increasing transaction concurrency and realizing database processing with high transaction processing capacity.

FIG. 4 shows a schematic configuration of an example of a conventional database processing device, in which 31 is an exclusive control section, 32 is an exclusive control information table, 33 is a transaction execution section, 34 is a database information table, and 35 is an address conversion table. , 36 is a database storage unit, and 37 is a terminal.

A conventional exclusive control method will be explained using FIG. 4. In the following description, referring to data is a process of reading data from the database storage unit 36, and updating data is a process of rewriting the contents of data in the database storage unit 36 according to update conditions. Furthermore, table search is a process of referring to the data that constitutes a table and selecting data that satisfies certain conditions, and data addition is a process of writing data to the database storage unit 36 so that it becomes part of the table. It is processing.

Referring to FIG. 4, a method (2) for ensuring transaction consistency using read locks and write locks when multiple transactions simultaneously refer to and update the same database will be described. The transaction execution unit 33
When a transaction is input from the terminal 37, data to be referenced or updated by the transaction is determined based on the database information table 34, and reference is made to ensure consistency of the transaction.

The exclusive control unit 31 is inquired as to whether updating is possible.

In the exclusive control information table 32, when data is referenced by a transaction, information indicating that the transaction controller has read-locked the data is registered, and when data is updated by a transaction, the transaction registers information indicating that the data is write-locked. The exclusive control unit 31 follows the rules that transactions other than the transaction that has read-locked the data cannot write-lock the data, and the rules that transactions other than the transaction that has write-locked the data cannot read-lock or write-lock the data. , while searching the exclusive control information table 32, determine whether read lock and write lock are possible in response to the inquiry from the transaction execution unit 33;
A notification is returned to the transaction execution unit 33 that if locking is possible, execution is possible, and if locking is not possible, execution is not possible. When the notification from the exclusive control unit 31 is executable, the transaction execution unit 33 obtains an address in the data database storage unit 36 using the address conversion table 35, references or updates the data, and executes the notification from the exclusive control unit 31. When the notification cannot be executed, the data reference or update is made to wait for execution, or the transaction is aborted. By using the above method ■, data that is being referenced by one transaction will not be updated by another transaction, and data that is being updated by one transaction will not be referenced or updated by another transaction, and multiple transactions can be processed simultaneously. transaction consistency can be guaranteed. For guaranteeing transaction consistency using locks, see the document ``Database
Principles of Systems” Chapter 11 (J.

It is explained in detail in D.

However, in method I, when a transaction adds data to a table, the data is not locked by another transaction, so the exclusive control unit 31 determines that the data addition can definitely be executed. Consider a situation where a transaction that searches the same table multiple times and another transaction that adds data to that table are being processed at the same time.

At this time, the table search results and the table search results that were performed before the data was added and the data The problem arises that the search results for the same table performed after the table is added are different.

An example of the occurrence of this phenomenon is shown in FIG. In FIG. 5, transaction A searches the table 41 at times a and d under the same conditions, and transaction B adds data 42 to the table 41 at times already present. Transaction B has already write-locked data 42,
Reference and update of data 42 by other transactions is prohibited, but when transaction B ends at time C, the write lock on data 42 is released, so data 42 can be referenced by transaction A at time d. Therefore, in transaction A, the results of the search at time d may include data that was not obtained by the search at time a.

A method to prevent such problems is to provide data (terminus identification data) in advance for identifying the end of a table, read-lock the end identification data when searching the table, and add data to the table. A method has been proposed in which the terminal identification data is write-locked, and the lock management is performed for the terminal identification data in the same manner as method (2) (Reference Ber.
Stein, P.

A., Hadzilacos, V.,
Goodman, N., “Concurrenc.
y Control and Recovery in
Database Systems” (ADDISO
N-WESLEY PUBLESHENG COMPA
NY, 1987)).

According to the above method (■), in the table being searched by a certain transaction, the end identification data is read-locked, so even if another transaction attempts to add data, the end identification data cannot be write-locked. , data cannot be added until the read lock on the terminal identification data is released, that is, until the transaction searching the table ends. Therefore, even if one table is searched multiple times in one transaction, different searches will occur. You will never get results.

[Problem to be solved by the invention]

However, in the above method (2), it is not possible to read and write lock the terminal identification data at the same time, so a transaction that searches for the same table in a relational database and a transaction that adds data to the table end up waiting. Therefore, there was a problem in that these two transactions could not be processed simultaneously.

The present invention has been made to solve the above problems.

An object of the present invention is to provide a method that simultaneously processes searches and data additions to the same table in a relational database and guarantees consistency of search results performed multiple times in one transaction.

The above and other objects and novel features of the present invention will become apparent from the description of this specification and the accompanying drawings.

[Means to solve the problem]

To achieve the above object, the present invention provides an input/output device that outputs data, a transaction execution unit that executes transaction processing on data forming tables of a relational database, and a an exclusive control unit that guarantees transaction consistency; a database storage unit that stores data forming tables of the relational database; and database information for registering data identifiers for identifying data forming tables of the relational database. a table, an address conversion table that registers the storage address in the database storage section of data constituting the table of the relational database corresponding to the data identifier; a read lock indicating that a transaction is referencing the data; and an exclusive control information table that registers exclusive control information consisting of a write lock indicating that the transaction is updating, and an additional lock indicating that the transaction is adding data, the database processing device comprising: 1 assigned to the data when adding data to the table of
One data identifier is registered in the database information table in advance, and when data is added to the exclusive control information table, the exclusive control unit locks the addition of the record using the data identifier registered in advance. At the same time as registering in the exclusive control information table, a new data identifier for the next additional data is registered in the database information table, and the read lock information for the new data identifier is the same as the data identifier for which the additional lock was registered. The exclusive control information is registered in the exclusive control information table so that it is the same as the data, the transaction execution unit is permitted to add the data to the exclusive control information table, and the data that has been additionally locked by another transaction is registered. At the time of reference, by registering a read lock for the data in the exclusive control information table and instructing the transaction execution unit to omit reference to the data, The most important feature is that it guarantees transaction consistency while simultaneously adding data to the table and searching the table.

[Effect]

According to the above-described means, a data identifier to be assigned to data to be added to a table of a relational database is registered in the database information table in advance. When the exclusive control unit is requested by the transaction execution unit to determine the consistency guarantee for data added to a table, the exclusive control unit registers an additional lock for the data in the exclusive control information table using the data identifier, and when the data is added. Notify the transaction execution unit that it is possible. Furthermore, the exclusive control unit registers in the database information table a new data identifier to be assigned to data added to the table, and determines that the read lock for the data identifier used to register the additional lock is the same as the read lock for the data identifier. Register exclusive control information in the exclusive control information table so that

Furthermore, when the exclusive control unit is requested by the transaction execution unit to determine the consistency guarantee regarding the reference of additionally locked data, the exclusive control unit registers the read lock of the data in the exclusive control information table, and the transaction execution unit registers the read lock of the data. Notify to omit data reference.

That is, in the present invention, it is possible to simultaneously perform a read lock and an add lock on a data identifier corresponding to the terminal identification data of the prior art, thereby allowing another transaction to add a record to the table being searched by a certain transaction. When a table search finds data that is additionally locked by another transaction,
By omitting the reference to this data, the consistency of the transaction that searches the table is not compromised due to records in the adding process, and consistent transaction processing can be performed while searching the table and adding data to the table at the same time. It is considered to be achievable.

[Embodiments of the invention]

Hereinafter, one embodiment of the present invention will be specifically described using the drawings.

FIG. 1 is a block diagram showing a schematic configuration of an embodiment of a database processing device to which the present invention is applied.

In FIG. 1, components 11 to 17 correspond to components 31 to 37 in FIG. 3, respectively, and 11 is an exclusive control section;
2 is an exclusive control information table, 13 is a transaction execution unit,
14 is a database information table, 15 is an address conversion table, 1
6 is a database storage unit, and 17 is a terminal.

In the embodiment shown in FIG. 1, records are selected as the size of data to be referenced, updated, and added, and records 1 to 3 are
Transaction T1 searches table A consisting of , then transaction T2 adds record 4 to table A,
Furthermore, a state in which transaction T1 has searched table A again is shown.

The record identifiers corresponding to records 1 to 4 are R1-R4
, and the record identifier prepared in advance for the record to be added to table A is R5.

When a transaction is input from the terminal 17, the transaction execution unit 13 executes the database information table 14.
Find the record identifier of the record to be used for processing from
The exclusive control unit 11 is inquired as to whether references, updates, and additions that guarantee transaction consistency are possible. The exclusive control unit 11 uses the transaction identifier, table identifier, record identifier, and processing type (reference/update/additional commit release/lock release due to abort) received from the transaction execution unit 13.
While searching the exclusive control information table 12 and the database information table 14, determination is made according to the flowcharts in FIGS. 2(a) to 2(d), and the results are notified to the transaction execution unit 13.

The exclusive control information table 12 includes a lock record list 18 that stores record identifiers of records locked by transactions, and a lock transaction list 19 that stores transaction identifiers and lock modes of transactions that lock records. The lock record list 18 is composed of a record identifier field that stores a record identifier, and a list pointer field that stores a pointer pointing to the corresponding lock transaction list 19. The lock transaction list 19 includes a transaction identifier field that stores a transaction identifier of a transaction that is locking a record, and a lock mode field that indicates the type of read lock, write lock, or additional lock.

The database information table 14 is composed of a table identifier field for storing a table identifier and a constituent record list field for storing record identifiers of records forming the table. The record identifier at the beginning of the configuration record list corresponds to the record identifier assigned to the record added to the table.

FIG. 2(a) shows the flow of processing when the exclusive control unit 11 is requested by the transaction execution unit 13 to determine consistency guarantee regarding record reference.

In the present invention, an additional lock and a read lock can be simultaneously applied to an additional record, and when the record addition is completed, if the record is read-locked, the additional lock is changed to a write lock. The referenced data may be read-locked and write-locked at the same time. If the transaction requesting the reference does not have a read lock on the record, and the record is write-locked by a transaction other than the transaction requesting the reference, the record cannot be read-locked. , the exclusive control unit 11 notifies the transaction execution unit 13 that the record cannot be referenced and the transaction identifier of the transaction that has write-locked the record. If the transaction requesting the reference has a read lock on the record, and the record is write-locked by a transaction other than the transaction requesting the reference, the previous reference to the record is omitted by the transaction. Therefore, at this time, the exclusive control unit 11 notifies the transaction execution unit 13 to omit reference to the record. If the record is additionally locked by a transaction other than the transaction requesting the reference, the record is not yet valid as a constituent record of the table, so at this time the exclusive control unit 11 adds the relevant information to the exclusive control information table 12. A read lock for the record is registered and the transaction execution unit 138 is notified to omit reference to the record. If the record identifier of the record is prepared to be assigned to an additional record in cases other than the above three cases, the entity of the record does not exist in the database storage unit 16, so in this case, the exclusive control unit 11 A read lock for the record is registered in the control information table 12, and the transaction execution unit 13 is notified to omit reference to the record. Except for the above four cases, the record has not been subjected to any processing that would impair the consistency of the transaction. At this time, the exclusive control unit 11 registers a read lock for the record in the exclusive control information table 12, and The transaction execution unit 13 is notified that the record can be referenced.

FIG. 2(b) shows the flow of processing when the exclusive control unit 11 is requested by the transaction execution unit 13 to determine whether consistency is guaranteed for updating a record. If the record is locked by a transaction other than the transaction requesting the update, the record cannot be write-locked, so at this time, the exclusive control unit 11
The transaction execution unit 13 is notified that the record cannot be updated and the transaction identifier of the transaction that is locking the record. In cases other than the above, the record has not been subjected to any processing that would impair transaction consistency, so at this time the exclusive control unit 11 registers a write lock for the record in the exclusive control information table 12, and the record is The transaction execution unit 13 is notified that the update is possible. However, if the transaction requesting the update has read-locked the record, the registration is performed by rewriting the lock mode of the exclusive control information to write lock, and if the transaction has additionally locked the record, This registration will not be performed.

FIG. 2(C) shows the flow of processing when the exclusive control unit 11 is requested by the transaction execution unit 13 to determine consistency guarantee regarding record addition.

The exclusive control unit 11 obtains from the database information table 14 a record identifier to be assigned to an additional record registered in advance in the table to which the record is added, and registers an additional lock for the record in the exclusive control information table 12. Furthermore, a new record identifier to be assigned to the record to be added to the table is registered in the database information table 14, and read lock information for the data identifier of the additionally locked data is copied as information for the newly registered record identifier for exclusive control. Register in information table 12. Thereafter, the database execution unit 13 is notified that the record can be added.

FIG. 2(d) shows the flow of processing when the exclusive control unit 11 receives a request from the transaction execution unit 13 to release a lock associated with a commit or to release a lock associated with an abort.

The exclusive control unit 11 deletes all exclusive control information related to the transaction requesting lock release from the exclusive control information table 12. However, when releasing a lock associated with a commit, if another transaction is simultaneously holding a read lock on a record that has been additionally locked by this transaction, the lock mode of the exclusive control information is changed from additional lock to write lock, Guarantees transactional consistency by preventing additional records from being referenced or updated until the read-locked transaction is completed. The write lock is released simultaneously when all read locks for the record are released. Furthermore, when a lock is released due to an abort, the record identifier of the record additionally locked by the transaction is deleted from the database information table 14, and read locks of the record by other transactions are also released at the same time. Further, the exclusive control unit 11 notifies the transaction execution unit 13 of the record identifier of the record for which the lock has been released.

When the transaction execution unit 13 receives a notification from the exclusive control unit 11 that the record can be referenced or updated, the transaction execution unit 13 obtains the address in the database storage unit 16 of the record corresponding to the record identifier from the address conversion table 15, and refers to or updates the record. Perform updates. When a notification that addition is possible is obtained from the exclusive control unit 11, an empty address in the database storage unit 16 is registered in the address conversion table 16 in association with the record identifier of the added record, and the record is added using the address. conduct. When receiving a notification from the exclusive control unit 11 that it cannot be referenced or updated, it is determined based on the transaction identifier notified at the same time and the list of transactions waiting to be executed at the time of the notification. Determine whether a deadlock will occur by placing a transaction determined to be unupdatable in a waiting state. When it is determined that no deadlock will occur,
The transaction is set to be pending and added to the pending processing list. When it is determined that a putlock has occurred, the transaction is aborted and the exclusive control unit 1
1 to request unlocking due to abort.

In the present invention, the size of the data to be referenced, updated, or added is not limited to the record used in the above explanation, but may be any size such as a page.

FIG. 3 is an explanatory diagram showing an example of transaction processing using the present invention. In FIG. 3, a read lock described by a dotted arrow indicates that reference to the data pointed to by the dotted arrow is omitted. As shown in FIG. 3, in this embodiment, table 51 at time gpl of transaction C
In the search for Table 51, the reference to data 52 added by transaction D at time f is omitted, and the same data as in the search for table 51 at time e of transaction C is referenced, based on the processing flow in FIG. 2(a). be done. Therefore, all the search results in Table 51 at time 8 pgtl of transaction C are the same.

As is clear from the above explanation, according to this embodiment, even if another transaction adds data to the table that is being searched by a certain transaction, the transaction that is searching for one table will not be affected by the added data. Since processing can proceed without any processing, it is possible to guarantee transaction consistency while simultaneously processing data addition to a table and table retrieval, making it possible to realize a database processing device with high transaction concurrency. .

The present invention has been specifically explained above based on examples, but
It goes without saying that the present invention is not limited to the embodiments described above, and can be modified in various ways without departing from the spirit thereof.

〔Effect of the invention〕

As described above, according to the present invention, it is possible to guarantee transaction consistency while simultaneously processing data addition to a table and table search, thereby realizing a database processing device with high transaction concurrency. can.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing a schematic configuration of an embodiment of a database processing device using the present invention, and FIG.
) is a diagram showing the processing flow of the exclusive control unit in the embodiment of FIG. 1, and FIG. 3 shows an embodiment in which adding data to a table and searching the table are simultaneously processed using the present invention. Fig. 4 is a block diagram showing a schematic configuration of an example of a general database processing device, and Fig. 5 is a block diagram showing a schematic configuration of an example of a general database processing device. FIG. In the figure, 11.31...exclusive control section, 12.32...
Exclusive control information table, 13.33...Transaction execution unit, 14゜34...Database information table, 15.3
5...Address conversion table, 16.36...Database storage unit, 17.37...Terminal, 18...Lock record list, 19...Lock transaction list, 41.51...Table, 42.52...Data. Figure 2 (C)

Claims (1)

[Claims] (1) An input/output device that inputs and outputs data, a transaction execution unit that executes transaction processing on the data that makes up the tables of the relational database, and an exclusive control unit that guarantees transaction consistency when multiple transactions are executed simultaneously. a database storage unit that stores data that constitutes the tables of the relational database; a database information table that registers data identifiers for identifying data that constitutes the tables of the relational database; and the relationships that correspond to the data identifiers. An address conversion table that registers the storage address in the database storage section of data that makes up a database table, a read lock that indicates that a transaction is referencing data, and a write lock that indicates that a transaction is updating data. and an exclusive control information table that registers exclusive control information consisting of an additional lock indicating that a transaction is adding data, the database processing device comprising: One data identifier assigned to the data is registered in advance in the database information table, and when data is added to the table, the exclusive control unit locks the addition of the record using the previously registered data identifier. At the same time, a new data identifier for the next additional data is registered in the database information table, and the read lock information for the new data identifier is the data for which the additional lock was registered. Register exclusive control information in the exclusive control information table so that it is the same as the identifier information, permit the transaction execution unit to add the data to the table, and when referencing data that has been additionally locked by another transaction. , by registering a read lock for the data in the exclusive control information table and instructing the transaction execution unit to omit reference to the data, 1. A method for exclusive control of a database, which guarantees transaction consistency while simultaneously adding a table and searching the table.