JP4623318B2 - Write-once database management system, unnecessary area repair processing method, and unnecessary area repair program - Google Patents

Description

  The present invention relates to a write-once database management system, an unnecessary area repair processing method, and an unnecessary area repair program, and more particularly to a write-once database management system and an unnecessary area repair processing method for efficiently repairing an unnecessary area of a table accompanying database update. And an unnecessary area repair program.

  Conventionally, in a database management system for managing a database, it has been necessary to consider the concurrency control of transactions.

  This transaction refers to a series of processing procedures for processing data in a database.

  Specifically, when processing data using a computer, this is a logical unit that ensures data processing so that the entire system will not fail even if a hardware or software failure occurs. Point to.

  Here, in a conventional database management system with write-once storage, when updating a database, multi-version concurrent control (Multi Version Concurrency Control) that retains both the data before update and the data after update Is used.

  In the multi-version simultaneous effective control using this write-once storage, the acquisition of the inquiry lock and the acquisition of the write lock do not compete with each other, so that the rollback and recovery implementation becomes very simple.

  However, this multi-version simultaneous execution control has a problem that the entire old data becomes unnecessary every time deletion or update occurs, and the old data remains, so that the storage area for storing the data is consumed greatly. .

  Even in the case of existing database management systems that are not write-once type, unnecessary areas are generated by rewriting data at the time of deletion or update. Although the unnecessary area is not large, an unnecessary area similarly occurs.

  In such a case, there is a function of reusing unnecessary areas by dividing and storing records in the database, but a state occurs in which one record is inserted into a plurality of blocks, Insertion / update may take time.

  Because a database is a collection of records, a record is like a card. This is because the record includes pieces of information related to each other called fields, so that the information included in one record is divided into a plurality of blocks.

  Therefore, the write-once database management system automatically fills unnecessary areas periodically for each page that is a recording unit of the database, and stores the page address of the usable area and the usable area size to store the data. Processing for reuse is performed.

  Furthermore, in a page where unnecessary areas are packed, unnecessary area repair processing (hereinafter referred to as unnecessary area repair) is performed depending on the ratio of the number of updates of data to the table (hereinafter, this may be referred to as table data). A method of automatically performing processing is also performed (see Non-Patent Document 1).

  The table referred to here is an object for storing data and means a collection of records sharing the same field. The table is managed by record (row) and field (column) objects.

  However, in such a conventional technique, the unnecessary area repair process scans all pages in which unnecessary areas have occurred in the table, thereby increasing the processing load on the disk and the like. Therefore, there is a problem that a high processing load is given to a storage device of a database having a storage or the like, which adversely affects other processes.

  In addition, since the unnecessary area repair processing is performed when the number of updated rows (that is, the number of records) in the table exceeds a certain ratio with respect to the entire table, the unnecessary area is repaired. Unnecessary area repair processing was performed regardless of the situation.

  Therefore, when performing the unnecessary area repair process, there is a problem that the unnecessary area repair process is performed even when the area repair rate is low.

  Specifically, for example, when variable-length data is defined, the size of each table data is different, so the number of updated rows is not necessarily proportional to the unnecessary area size. For this reason, there is a possibility that unnecessary area repair processing may be performed when the number of updated rows is large even if the number of unnecessary areas is small as a whole.

  In addition, even when an unnecessary area of the page is biased, there is a problem that processing for packing data occurs for a page with a low area repair rate, which is very inefficient.

Accordingly, Patent Documents 1 to 3 describe handling of unnecessary area data in such a database management system.
Tatsuo Ishii, “PostgreSQL Complete Strategy Guide Revised 5th Edition”, Technical Review Publishing Co., Ltd. April 1, 2006, pp. 437-446 JP 2000-137639 A JP 2001-075832 A JP 2004-303368 A

  However, even if the ones described in Patent Documents 1 and 3 are applied to a conventional write-once database management system, when a large amount of unnecessary area is accumulated in the database and free space is secured again, There is still a problem that the unnecessary area repair processing becomes complicated and the processing capacity is lowered due to a considerable burden on the storage disk, which adversely affects other processing.

  The present invention has been made in view of the above problems, and provides a write-once database management system, an unnecessary area repair processing method, and an unnecessary area repair program that efficiently perform unnecessary area repair processing accompanying data update. Objective.

  The write-once database management system according to the present invention is a write-once database management system for managing a database that stores data. In the write-once database management system, unnecessary areas generated in the database by updating or deleting the data are recorded in units of recording of the database. Management means that manages while updating every page, and calculates the size of the unnecessary area from the unnecessary area managed by the management means for each page, and the size of the unnecessary area and the unnecessary in the database An unnecessary area storage unit that stores the position of each area for each page; and when the unnecessary area occurs in the page, the size of the unnecessary area stored in the unnecessary area storage unit, and the size of the page in the page Whether to release data belonging to the unnecessary area based on the ratio of the unnecessary area An unnecessary area ratio determining means for determining the unnecessary area ratio determining means when the unnecessary area ratio determining means exceeds a preset threshold based on the size of the unnecessary area and the ratio of the unnecessary area on the page Is further provided with unnecessary area release processing means for releasing data belonging to the unnecessary area of the page from the position of the unnecessary area.

  Further, in the write-once database management system according to the present invention, the unnecessary area release processing unit is the oldest involved as the transaction identifier that is an execution log of data belonging to the unnecessary area of the page and the oldest execution log of the system. A tuple data validity judging means for comparing the transaction and determining whether or not the data belonging to the unnecessary area is deleteable tuple data; and the tuple data validity judging means determines that the transaction identifier is the oldest transaction. If it is older than the identifier, the data belonging to the unnecessary area is determined to be the deleteable tuple data, and the tuple data deletion reservation means for performing deletion reservation for the deleteable tuple data, and the tuple data deletion When there is tuple data that can be deleted by the reservation means When there is the index data assigned to the removable tuple data by the index data judging means for judging the presence or absence of the index data assigned to the removable tuple data, and the index judging means Index data deleting means for deleting the index data, wherein the index determining means is when the index data is deleted by the index data deleting means or when the index data is not in the deleteable tuple data Alternatively, the deleteable tuple data may be released to secure a free area.

  Further, the write-once database management system according to the present invention is a tuple in which the management means updates the data stored in the database by updating the page, and newly appends to the database by the update. Tuple length calculating means for calculating a record of data as an additional tuple data length, and the additional tuple is secured in the free space secured by calculating the additional tuple data length by the tuple length calculating means. When the tuple data record having the data length can be inserted, the tuple data having the calculated tuple data length is inserted into the empty area, and the page of the inserted tuple data is updated. If it is not possible to insert into the free space, the tuple is added to the last page of the database. The tuple data of the data length can be inserted into the last page, and the tuple data of the calculated tuple data length cannot be inserted into the empty area, and the last page of the database If the tuple data having the tuple data length cannot be inserted, a new page may be secured in the database and the tuple data having the tuple data length may be additionally written.

  In the write-once database management system according to the present invention, the management unit manages the unnecessary area generated in the database while totaling each data table, and the table total management unit includes: Table unnecessary area storage means for storing the unnecessary area for each table while updating the size, and the table total management means includes the size of the unnecessary area for each table and the unnecessary area in the table. And a start threshold value for starting the unnecessary area release processing means, or an end threshold value for ending the unnecessary area release processing means, based on the ratio of the areas, and the unnecessary area depending on the start threshold value or the end threshold value. The release processing means may be controlled.

  In the write-once database management system according to the present invention, the unnecessary area ratio determining unit may generate the unnecessary area when the unnecessary area is generated on the page and exceeds the preset threshold. The system further comprises unnecessary area release processing reservation means for performing reservation for implementing the release processing means, and the unnecessary area release processing reservation means controls the unnecessary area release processing means based on the processing load of the database. May be.

  According to the present invention, the unnecessary area generated in the database is managed for each page which is a recording unit of the database, and is set in advance based on the size of the generated unnecessary area and the ratio of the unnecessary area in the page. When the threshold is exceeded, an unnecessary area release process can be performed on the page that exceeds the threshold. Therefore, the unnecessary area can be repaired by limiting the size of the unnecessary area to a page that exceeds the threshold value. Thus, the write-once database management system that efficiently performs the unnecessary area repair processing accompanying the data update is unnecessary. An area repair processing method and an unnecessary area repair program can be realized.

Next, the configuration of the embodiment of the present invention will be described in detail with reference to the drawings.
(First embodiment)
(1) Write-once database management system according to first embodiment (1-1) Configuration of write-once database management system Referring to FIG. 1, a write-once database management system according to the first embodiment includes a database client 1 And a database management system 2 to which the database client 1 is connected, and a storage device 3 for storing the database.

  The database management system 2 includes a query management unit 21, a transaction management unit 22, and a storage management unit 23.

  The query management unit 21 includes a query analysis unit 211 and an execution plan creation unit 212.

  The query analysis unit 211 analyzes the query passed from the database client and passes the result to the execution plan creation unit 212.

  The execution plan creation unit 212 creates an execution plan in the transaction management unit 22 and the storage management unit 23.

  The transaction management unit 22 includes a transaction control management unit 221.

  The transaction control management unit 221 determines whether the tuple is valid or invalid from the tuple data included in the tuple management unit 233 of the storage management unit 23 and the commit status information of the transaction information storage unit 33. Judgment is made.

  Here, a tuple is a function that combines several values into one and treats them as if they were one value, and corresponds to a row in the table. The tuple data refers to data assigned by the tuple function.

  The storage management unit 23 includes an index management unit 231, a page management unit 232 (corresponding to management means in the present invention), a tuple management unit 233 (corresponding to management means in the present invention), and a free capacity management unit. 234 (corresponding to the unnecessary area ratio determining means in the present invention) and a statistical information management unit 235.

  The index management unit 231 receives a tuple image, which will be described later, from the tuple management unit 233, and creates and deletes index data. Further, the index management unit 231 manages the created index data by inputting / outputting the index data to / from an index data storage unit 31 described later.

  The page management unit 232 receives a tuple image, which will be described later, from the tuple management unit 233 and creates a page. The page data (hereinafter referred to as page data) is used as table data, and a table data storage unit, which will be described later. I / O to 32. The page management unit 232 also performs unnecessary area repair processing and the like.

  The tuple management unit 233 generates tuple images and acquires tuple data from the page data received from the page management unit 232 based on information received from the query management unit 21 and the transaction management unit 22. Further, input / output is performed to / from the transaction information storage unit 33 of transaction information described later from the tuple generation and extracted information.

  Here, the tuple image corresponds to table data corresponding to page data generated by the tuple data and the tuple function.

  The free space management unit 234 inputs / outputs page information including an unnecessary area generated by update or deletion and size information of the unnecessary area to a page unnecessary area information storage unit 35 described later.

  The statistical information management unit 235 manages the entire database information such as table definition, table data size, page size, page unnecessary area repair processing start threshold, page address and size with free capacity, oldest transaction identifier, etc. To do. Further, input / output to / from the database statistical information storage unit 34 is performed.

  The storage device 3 includes an index data storage unit 31, a table data storage unit 32, a transaction information storage unit 33, a database statistical information storage unit 34, and a page unnecessary area information storage unit 35 (in the unnecessary area storage means in the present invention). Applicable).

  The index data storage unit 31 stores index data when an index is created in the table.

  The table data storage unit 32 stores database table data in units called pages.

  The transaction information storage unit 33 stores state information for each transaction by a transaction identifier for uniquely indicating each transaction (in the figure, this may be referred to as a transaction ID).

  The database statistical information storage unit 34 is the oldest committed transaction identifier (this is called the oldest transaction identifier), table definition, table data size, page size, A page unnecessary area repair process start threshold, a page address having a free space, a size, and the like are stored.

The page unnecessary area information storage unit 35 stores the total unnecessary area size for each page generated when the table data stored in the table data storage unit 32 is updated or deleted.
(1-2) Unnecessary Area Repair Processing of Write-Once Database Management System Next, referring to the block diagram of FIG. 1 and the flowcharts of FIGS. 2 and 3, the write-once database management system according to the first embodiment The unnecessary area repair process will be described in detail.

  In the first embodiment, the operation of the target database is performed only by deleting unnecessary areas and updating queries when unnecessary areas are generated by multi-version concurrent control (Multi Version Concurrency Control). The explanation is limited.

  First, the database management system 2 (FIG. 1) receives an update query from the database client 1 (FIG. 1).

  Next, when the query management unit 21 receives the query, the query analysis unit 211 analyzes the query.

  Then, the execution plan creation unit 212 creates an execution plan and sends the information (that is, the created execution plan) to the storage management unit 23 (step A1 in FIG. 2).

  The storage management unit 23 receives a transaction identifier (referred to as transaction ID in FIG. 2) by the transaction control management unit 221 of the transaction management unit 22 in order to perform multi-version simultaneous execution control of transactions (FIG. 2 describes this as a transaction ID). Step A2).

  Here, the operation changes depending on whether the previously received query is update or deletion (step A3 in FIG. 2).

  If the previously received query is an update, a tuple image is created by the tuple management unit 233 (FIG. 1), and the table data storage unit 32 uses the tuple image as table data by the page management unit 232 (FIG. 1). Write to.

  Here, when the created tuple image has an index, the index management unit 231 writes the index of the created tuple image in the index data storage unit 31 (step A4 in FIG. 2).

  On the other hand, when the query received earlier is deletion or when the update data has been written, the deletion flag and the transaction identifier are written in the tuple image created by the tuple management unit 233 (FIG. 2). In step A5), the tuple image is written as table data in the table data storage unit 32 by the page management unit 232 (step A6 in FIG. 2).

  As a result, the current transaction identifier is deleted.

  In step A5, the free space management unit 234 stores the page address of the tuple image and the size of the tuple image for which the deletion flag has been set, and in step A6, the deleted page address and the size of the tuple image data. To the page unnecessary area information storage unit 35 in which the total unnecessary area size for each page is stored.

  The tuple management unit 233 writes the committed information in the transaction identifier in the transaction information storage unit 33 in order to record the completion of the transaction. In addition, the tuple management unit 233 writes the update information when the oldest transaction identifier stored in the database statistical information storage unit 34 can be updated via the statistical information management unit 235 (FIG. 1) (see FIG. 1). Step A7).

  Thereafter, the free space management unit 234 acquires the page size and the page unnecessary area repair processing start threshold from the statistical information management unit 235, and whether or not the unnecessary area size ratio of the page updated in step A4 exceeds the threshold. Or, it investigates for every page from the information memorize | stored in the page unnecessary area | region information storage part 35 (step A8-A9 of FIG. 2).

  Here, if it is an updated page (step A8), the ratio of the page unnecessary area in the updated page is determined (step A9), and if it exceeds the threshold ratio, the page will be described later. Unnecessary area release processing is performed (step B in FIG. 2).

  If the unnecessary area releasing process in step B is completed or if the threshold ratio is not exceeded in step A9, the process returns to step A8 to perform the process for the next page.

As described above, when the confirmation of the ratio of the page unnecessary area is completed for all the updated pages, the processing of the query received initially is terminated.
(1-3) Unnecessary area release processing of write-once database management system (step B)
Next, the unnecessary area releasing process (step B in FIG. 2) will be described with reference to the flowchart in FIG.

  First, the page management unit 232 (FIG. 1) and the tuple management unit 233 (FIG. 1) inspect the tuple area from the first for the page for releasing the unnecessary area acquired from the table data storage unit 32 (FIG. 1). (Step B1 in FIG. 3).

  Here, when there is an unexamined tuple in the page, the oldest transaction identifier is referred to through the database statistical information storage unit 34.

  Then, the transaction control management unit 221 compares the tuple transaction identifier with the oldest transaction identifier, and if it is determined that the tuple transaction identifier is older than the oldest transaction identifier, the transaction control management unit 221 can delete the tuple (FIG. 3). Step B2).

  If deletion is possible, the storage management unit 23 stores the tuple position information in the deletion reservation list (step B3 in FIG. 3).

  In this way, the steps B1 to B3 in FIG. 3 are repeated, and all the tuples stored in the page are inspected.

  When the inspection is completed for all the tuples, it is confirmed whether or not the tuple deletion information is stored in the deletion reservation list confirmed in step B3 in FIG. 3 (step B4 in FIG. 3).

  Here, if there is a deletion target tuple in the deletion reservation list, the presence / absence of an index in the table is checked (step B5 in FIG. 3).

  If there is an index in the deletion target tuple, the index management unit 231 deletes the index information (that is, corresponding to the index data) of the deletion target tuple included in the page (step B6 in FIG. 3). .

  Thereby, when there is no index information or when deletion of all the index information is completed, the page management unit 232 secures a free area of the page (step B7 in FIG. 3).

  The free space management unit 234 stores the free space and page number secured in the database statistical information storage unit 34 via the statistical information management unit 235, and updates the free space and region information (step B8 in FIG. 3).

  Here, the securing of the free space (step B7 in FIG. 3) will be described with reference to FIG. 4 as a specific example.

  Referring to FIG. 4, the page 4 before free space release includes page management information 41 and tuple data 42.

  The tuple data 42 includes valid tuple data 421 and 424, and deleteable tuple data 422, 423, and 425.

  In the page management information 41, validity information of the tuple data 421 to 425 is stored. Then, the page 5 after securing the free area from the page 4 before release becomes the page 5 after free area release.

  The page management unit 232 arranges the valid tuple data 421 and 424 in order from the back, invalidates the tuple data of the erasable tuple data 422 and 423 and 425 managed as erasable in the page management information 41, The management information 51 is rewritten.

  That is, the valid tuple data 421 is equal to the valid tuple data 521 and the valid tuple data 424 is equal to the valid tuple data 522.

As a result, only valid tuple data remains, and a free area can be secured appropriately.
(1-4) Reserved Area Additional Recording Process of Write-Once Database Management System Next, the reserved area additional recording process using the reserved empty area will be described with reference to the flowchart shown in FIG.

  When additional data is generated by inserting and updating table data in the table data storage unit 32, which is a database, the storage management unit 23 operates the tuple data to be added one by one (step C1 in FIG. 5).

  First, the tuple management unit 233 calculates the tuple length of the tuple data to be added (step C2 in FIG. 5).

  Then, the free capacity management unit 234 checks whether there is a page whose tuple length can be stored in the reuse area (step C3 in FIG. 5).

  Here, if there is a page that can be reused, the page management unit 232 inserts additional tuple data into the target page of the table data storage unit 32 (step C4 in FIG. 5).

  Thereafter, the free space management unit 234 updates the page size information used for the reuse area information (step C5 in FIG. 5).

  On the other hand, if there is no page that can be stored in the reuse area, the page management unit 232 checks whether or not the tuple data to be added can be inserted into the final page (step C6 in FIG. 5).

  If the additional target tuple data can be inserted, the additional target tuple data is inserted into the last page (step C7 in FIG. 5).

  On the other hand, if the tuple data to be added cannot be inserted on the last page, a new page is secured and the tuple data to be added is inserted (step C8 in FIG. 5).

The above steps C1 to C8 are repeated, and when all the tuple data has been inserted, the process ends.
(1-5) Effect of Write-once Database Management System In this embodiment, the free space management unit 234 stores the unnecessary area generated by the update or deletion in the page unnecessary area information storage unit 35 for each page. Therefore, the unnecessary area for each page is managed and stored by updating or deleting the table data.

  Thereby, automatic unnecessary area repair and reuse for each page can be efficiently performed. Also, depending on the operation status, it is possible to make it difficult to generate unnecessary capacity by reducing the threshold for starting page unnecessary area repair processing, or by increasing the threshold for starting page unnecessary area repair processing, It can also be limited to highly efficient pages.

  As described above, according to the present embodiment, the size of the unnecessary area is managed for each page that is a recording unit of the database, so that the ratio of the unnecessary area is determined for the page to be subjected to the unnecessary area repair. Since it is possible to limit the number of pages to a predetermined threshold or more, it is possible to efficiently perform unnecessary area repair processing.

Furthermore, according to the present embodiment, unnecessary area release processing is not performed on all pages where unnecessary areas have occurred, and each time unnecessary areas are generated by deletion or update, the threshold of the ratio of unnecessary areas Since the unnecessary area repair process can be performed only on pages exceeding the limit, the unnecessary area repair process can be distributed.
(Second Embodiment)
(2) Write-once database management system according to second embodiment (2-1) Configuration of write-once database management system Next, a write-once database management system according to the second embodiment will be described.

  Referring to FIG. 6, in the write-once database management system according to the second embodiment, in addition to the configuration in the first embodiment, the storage device 6 further includes a table unnecessary area information storage unit 61.

  The table unnecessary area information storage unit 61 records the total of the unnecessary area sizes stored in the page unnecessary area information storage unit 35 for each table.

  The database statistical information storage unit 34 stores a table unnecessary area repair processing start threshold and a table unnecessary area repair end threshold.

However, here, as a specific example, it is assumed that the table unnecessary area repair processing start threshold is larger than the table unnecessary area repair processing end threshold.
(2-2) Unnecessary Region Repair Processing of Write-Once Database Management System Next, unnecessary region repair processing in the second embodiment will be described in detail with reference to the flowcharts of FIGS. 6 and 7.

  In the flowchart of FIG. 7, the query analysis unit 211, the execution plan creation unit 212, the transaction control management unit 221, the index management unit 231, the page in the second embodiment indicated by steps A 1 to A 6 and step A 7 and step B Since the operations of the management unit 232, the tuple management unit 233, the free space management unit 234, and the statistical information management unit 235 are the same as the operations described in the first embodiment, description thereof is omitted.

  In the first embodiment, the free space management unit 234 manages only the size of the unnecessary area for each page.

  In contrast, in the second embodiment, after the page size information is updated in step A6, the free space management unit 234 updates the total unnecessary area size in the table to the table unnecessary area information storage unit 61.

  That is, the free space management unit 234 updates a value equal to the sum of the unnecessary area sizes stored in the page unnecessary area information storage unit 35 of the referenced table (step D1 in FIG. 7).

  After writing the transaction commit log in step A7 in FIG. 7, the free space management unit 234 obtains the table size and the table unnecessary area repair processing start threshold from the statistical information management unit 235, and stores the table unnecessary area information. The total table unnecessary capacity size is acquired from the unit 61. Then, the free space management unit 234 checks whether the unnecessary area ratio of the table (hereinafter referred to as the table unnecessary area ratio) does not exceed the table unnecessary area repair processing start threshold (step D2 in FIG. 7).

  If the table unnecessary area ratio exceeds the table unnecessary area repair processing start threshold, the page having the largest unnecessary area size is checked from the page unnecessary area information storage unit 35, and the page is displayed. On the other hand, an unnecessary area release process (step B) is performed.

  Then, after performing the unnecessary area release process (step B), the process returns to step D3, and whether or not the previous table unnecessary area ratio is smaller than the table unnecessary area repair end threshold stored in the database statistical information storage unit 34. If it is smaller, the process ends.

  Here, specifically, for example, it is assumed that the table unnecessary area repair processing start threshold is 30% and the table unnecessary area repair processing end threshold is 10%.

  For example, if a certain table unnecessary area ratio is 32% in step D2, the process proceeds to step D3 in FIG. 7 and the table unnecessary area repair processing end threshold is 10%, and therefore the process proceeds to step D4. . In step D4, the page having the largest unnecessary area size is investigated, and after searching for a page having a high unnecessary ratio, the process proceeds to step B, and unnecessary area releasing processing is performed.

  In this case, for example, the unnecessary area release process (step B) is performed on the page having the largest unnecessary area size, so that the table unnecessary area ratio that was 32% is restored in step D3. When the processing end threshold is less than 10%, the processing ends.

  On the other hand, if the table unnecessary area ratio is still larger than 10%, which is the table unnecessary area repair end threshold, in step D3, the page having the next larger unnecessary area size is investigated (FIG. 7). Step D4), an unnecessary area release process (Step B) is performed on the page.

  When the unnecessary area release process (step B) ends, the process returns to step D3, the ratio of table unnecessary areas is determined, and the unnecessary area release process (10% in the above example) is satisfied until the table unnecessary area repair end threshold (10% in the above example) is satisfied. Repeat step B).

  Note that the second embodiment has a feature in the processing of steps D2 and D3 shown in FIG. 7, and the reserved area additional process of data using the reserved empty area is described in the first embodiment. Since it is the same as the case of the form (1-4), the description is omitted.

Alternatively, Step D4 and Step B are provided between Step D2 and Step D3, and when the table unnecessary area repair processing start threshold value (for example, 30%) is not satisfied, the page having the largest unnecessary area size is examined. (Step D4), an unnecessary area release process (Step B) is performed on the page, and a process that satisfies the table unnecessary area repair process end threshold (for example, 10%) may be added.
(2-3) Effect of write-once database system In the second embodiment, by providing a table unnecessary area repair processing start threshold and a table unnecessary area repair end threshold, it is unnecessary depending on the ratio of unnecessary capacity in the entire table. The capacity restoration timing can be measured automatically.

  Similarly to the first embodiment, the free capacity management unit 234 is configured to store the unnecessary capacity generated by the update or deletion in the page unnecessary area information storage unit 35 for each page. Unnecessary capacity restoration can be performed from a highly efficient page.

Further, depending on the operation status, as shown in the first embodiment, by reducing the table unnecessary area repair processing start threshold, it is possible to make it difficult to generate unnecessary capacity, and the table unnecessary area repair processing end threshold. By increasing the size, the load for repairing unnecessary capacity can be reduced as much as possible.
(3) Write-once database management system according to the third embodiment (3-1) Configuration of the write-once database management system The structure in the third embodiment is the same as in the case of the first embodiment, 1 is the same as FIG. 1 except that the page management unit 232 monitors the disk load and processing status through the operating system and has a disk load threshold value that is a condition for starting unnecessary area repair.
(3-2) Unnecessary Region Repair Processing of Write-Once Database System Next, unnecessary region repair processing in the third embodiment will be described in detail with reference to the flowcharts of FIGS. 8 and 9.

  In the flowchart of FIG. 8, the query analysis unit 211, the execution plan creation unit 212, the transaction control management unit 221, the index management unit 231, the page management unit 232, the tuple in the third embodiment indicated by steps A1 to A8 and A9. The operations of the management unit 233, the free space management unit 234, and the statistical information management unit 235 are the same as the operations described in the first embodiment, and thus description thereof is omitted.

  In the first embodiment, when the unnecessary area size ratio exceeds the threshold value, the unnecessary area is repaired by the unnecessary area releasing process in step B after the comparison.

  In contrast, in the third embodiment, after the unnecessary area ratio is checked in step A9, if the unnecessary area size threshold is exceeded, reservation of unnecessary area repair for the page is performed together with the target page position. The part 232 remembers (step E1 in FIG. 8).

  Next, referring to the flowchart of FIG. 9, the page management unit 232 confirms whether reservations for unnecessary area repair are registered at regular time intervals or at designated time zones. (Step F1 in FIG. 9).

  If a reservation for unnecessary area repair is registered, the page management unit 232 checks the disk load and processing status through the operating system (step F2 in FIG. 9).

  If the disk load exceeds the disk load threshold for a certain time, the process waits until the disk load continues to be below the disk load threshold (step F3 in FIG. 9).

  Thus, by waiting until the disk load decreases (step F3 in FIG. 9), if the disk load continues to fall below the disk load threshold, or if the disk load is low in step F2, unnecessary area release processing is performed. Perform (Step B in FIG. 9).

  When the unnecessary area release processing in step B is completed, the process returns to step F1 and continues a series of steps until there is no reservation for unnecessary area repair. If there is no reservation for unnecessary area repair processing, the unnecessary area repair in the third embodiment is performed. The process ends.

The unnecessary area releasing process and the reserved area additional process using the reserved empty area are the same as those in the first embodiment, and thus the description thereof is omitted.
(3-3) Effect of Write-once Database In the present embodiment, the execution timing of unnecessary area repair can be automatically performed when the disk load is low.

  Thereby, as shown in the first embodiment, the area to be repaired is reduced for repairing the unnecessary area in units of pages, and the unnecessary area is repaired when the disk load is low. The impact on other operations of the database and other applications running on the same computer can be minimized.

  In addition, if the time zone where the influence range of the database and other applications is low is known in advance, the effect of unnecessary area repair can be further limited by specifying the execution timing of this unnecessary area repair in advance. You can also

It is a block diagram which shows the structure of the additional type database management system by the 1st Embodiment in this invention. It is a flowchart showing the unnecessary area | region repair process by the 1st Embodiment in this invention. It is a flowchart showing the unnecessary area | region release process by the 1st Embodiment in this invention. It is the figure which showed the securing of the empty area by the 1st Embodiment in this invention by the example. It is a flowchart showing the securing area additional recording process by the 1st Embodiment in this invention. It is a block diagram which shows the structure of the additional type database management system by the 2nd Embodiment in this invention. It is a flowchart showing the unnecessary area | region repair process by the 2nd Embodiment in this invention. It is a flowchart showing the unnecessary area | region repair process by the 3rd Embodiment in this invention. It is a flowchart which performs the unnecessary area | region release process by the 3rd Embodiment in this invention by the presence or absence of a release process reservation.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Database client 2 Database management system 3, 6 Storage device 4 Page before release 5 Page after release 21 Query management part 22 Transaction management part 23 Storage management part 31 Index data storage part 32 Table data storage part 33 Transaction information storage part 34 Database statistics Information storage section 35 Page unnecessary area information storage section 41, 51 Page management information 42 Tuple data 61 Table unnecessary area information storage section 211 Query analysis section 212 Execution plan creation section 221 Transaction control management section 231 Index management section 232 Page management section 233 Tuple Management unit 234 Free capacity management unit 235 Statistical information management unit 421, 423, 424 Valid tuple data 422, 425 Deletable tuple data 521, 522 Valid tuple data

Claims (15)

In a write-once database management system that manages a database that stores data,
A management means for managing unnecessary areas generated in the database by updating or deleting the data while updating each page as a recording unit of the database;
The management unit calculates the size of the unnecessary area from the unnecessary area managed for each page, and stores the size of the unnecessary area and the position of the unnecessary area in the database for each page. Area storage means;
Whether the data belonging to the unnecessary area is released based on the size of the unnecessary area stored in the unnecessary area storage unit and the ratio of the unnecessary area in the page when the unnecessary area occurs in the page Unnecessary area ratio determining means for determining whether or not,
The unnecessary area ratio determining means includes:
Based on the size of the unnecessary area and the ratio of the unnecessary area in the page, when a preset threshold is exceeded, the data belonging to the unnecessary area of the page is released from the position of the unnecessary area. A write-once database management system, further comprising unnecessary area release processing means. The unnecessary area release processing means includes:
Compare the transaction identifier, which is the execution log of the data belonging to the unnecessary area of the page, with the oldest transaction related to the oldest as the system execution log, and determine whether the data belonging to the unnecessary area can be deleted. Tuple data validity determination means for determining whether or not,
When the transaction identifier is older than the oldest transaction identifier by the tuple data validity determination means, the data belonging to the unnecessary area is determined as the deleteable tuple data and the deleteable tuple data Tuple data deletion reservation means for performing deletion reservation for
When there is tuple data that can be deleted by the tuple data deletion reservation unit, index data determination unit that determines whether there is index data assigned to the deleteable tuple data;
When there is the index data assigned to the deleteable tuple data by the index determining means, the index data deleting means for deleting the index data,
The index determining means releases the erasable tuple data to release a free area when the index data is deleted by the index data deleting means or when the index data is not in the erasable tuple data. The write-once database management system according to claim 1, wherein the write-once database management system is secured. When the management means updates the data stored in the database by updating the page,
A tuple length calculating means for calculating a tuple data record to be newly added to the database by updating, as a appending tuple data length;
If the tuple data length is calculated by the tuple length calculation means and the tuple data record of the additional write tuple data length can be inserted in the reserved free area, it is inserted into the free area. And update the page of the inserted tuple data,
When the tuple data having the calculated tuple data length cannot be inserted into the empty area and the tuple data having the tuple data length can be inserted into the last page of the database, the tuple data is inserted into the last page. ,
When the tuple data having the calculated tuple data length cannot be inserted into the empty area and the tuple data having the tuple data length cannot be inserted into the last page of the database, a new page is added to the database. The write-once database management system according to claim 2, wherein the tuple data of the tuple data length is additionally recorded. Table total management means for managing the unnecessary area generated in the database while totaling the unnecessary areas for each table of the data;
The table total management means further comprises table unnecessary area storage means for storing while updating the size of the unnecessary area for each table,
The table total management means includes:
Based on the size of the unnecessary area for each table and the ratio of the unnecessary area in the table, at least a start threshold value for starting the unnecessary area release processing means or an end threshold value for ending the unnecessary area release processing means With one side,
The write-once database management system according to claim 1, wherein the unnecessary area release processing means is controlled by the start threshold value or the end threshold value. The unnecessary area ratio determining means includes:
When the unnecessary area occurs in the page and the preset threshold value is exceeded, an unnecessary area release processing reservation unit that performs a reservation for performing the unnecessary area release processing unit is further provided. ,
The unnecessary area release process reservation means includes:
The write-once database management system according to claim 1, wherein the unnecessary area release processing means is controlled based on a processing load of the database. An unnecessary area repair processing method in a write-once database management system for managing a database storing data,
A management step of managing unnecessary areas generated in the database by updating or deleting the data while updating each page which is a recording unit of the database;
It is unnecessary to calculate the size of the unnecessary area from the unnecessary area managed for each page in the management step, and to store the size of the unnecessary area and the position of the unnecessary area in the database for each page. An area storage step;
Whether the data belonging to the unnecessary area is released based on the size of the unnecessary area stored in the unnecessary area storing step and the ratio of the unnecessary area in the page when the unnecessary area occurs in the page An unnecessary area ratio determining step for determining whether or not,
The unnecessary area ratio determining step includes:
Based on the size of the unnecessary area and the ratio of the unnecessary area in the page, when a preset threshold is exceeded, the data belonging to the unnecessary area of the page is released from the position of the unnecessary area. An unnecessary area repair processing method, further comprising: an unnecessary area release processing step. The unnecessary area release processing step includes:
Compare the transaction identifier, which is the execution log of the data belonging to the unnecessary area of the page, with the oldest transaction related to the oldest as the system execution log, and determine whether the data belonging to the unnecessary area can be deleted. A tuple data validity determination step for determining whether or not;
If the transaction identifier is older than the oldest transaction identifier in the tuple data validity determination step, the data belonging to the unnecessary area is determined as the deleteable tuple data and the deleteable tuple data Tuple data deletion reservation step for performing deletion reservation for
In the tuple data deletion reservation step, if there is the deleteable tuple data, an index data determination step of determining whether there is index data assigned to the deleteable tuple data;
An index data deletion step of deleting the index data when there is the index data assigned to the tuple data that can be deleted by the index determination step;
The index determining step releases the deleteable tuple data and frees the free space when the index data is lost by the index data deleting step or when the index data is not in the deleteable tuple data. The unnecessary area repair processing method according to claim 6, wherein the unnecessary area is repaired. The management step is to update the data stored in the database by updating the page;
A tuple length calculation step of calculating a tuple data record to be newly added to the database by update as a postscript tuple data length;
When the recordable tuple data length is calculated by the tuple length calculating step, if the tuple data record of the recordable tuple data length can be inserted in the reserved space area, it is inserted in the space area. And update the page of the inserted tuple data,
When the tuple data having the calculated tuple data length cannot be inserted into the empty area and the tuple data having the tuple data length can be inserted into the last page of the database, the tuple data is inserted into the last page. ,
When the tuple data having the calculated tuple data length cannot be inserted into the empty area and the tuple data having the tuple data length cannot be inserted into the last page of the database, a new page is added to the database. The unnecessary region repair processing method according to claim 7, wherein the tuple data having the tuple data length is additionally recorded. A table total management step for managing the unnecessary area generated in the database while totaling each data table;
The table total management step further comprises a table unnecessary area storage step for storing the table while updating the size of the unnecessary area for each table,
The table total management step includes:
Based on the size of the unnecessary area for each table and the ratio of the unnecessary area in the table, at least a start threshold value for starting the unnecessary area release processing step or an end threshold value for ending the unnecessary area release processing step With one side,
The unnecessary area repair processing method according to claim 6, wherein the unnecessary area release processing step is controlled by the start threshold or the end threshold. The unnecessary area ratio determining step includes:
When the unnecessary area occurs in the page and the preset threshold value is exceeded, an unnecessary area release process reservation step for making a reservation for performing the unnecessary area release process step is further provided. ,
The unnecessary area release processing reservation step includes:
The unnecessary area repair processing method according to claim 6, wherein the unnecessary area release processing step is controlled based on a processing load of the database. An unnecessary area repair program in a write-once database management system for managing a database for storing data,
A management procedure for managing unnecessary areas generated in the database by updating or deleting the data while updating each page as a recording unit of the database;
The management procedure calculates the size of the unnecessary area from the unnecessary area managed for each page, and stores the size of the unnecessary area and the position of the unnecessary area in the database for each page. Region storage procedure;
Whether the data belonging to the unnecessary area is released based on the size of the unnecessary area stored in the unnecessary area storing procedure and the ratio of the unnecessary area in the page when the unnecessary area occurs in the page An unnecessary area ratio determination procedure for determining whether or not,
The unnecessary area ratio determining procedure includes:
Based on the size of the unnecessary area and the ratio of the unnecessary area in the page, when a preset threshold is exceeded, the data belonging to the unnecessary area of the page is released from the position of the unnecessary area. It further includes an unnecessary area release processing procedure,
An unnecessary area repair program which is executed by a computer. The unnecessary area release processing procedure includes:
Compare the transaction identifier, which is the execution log of the data belonging to the unnecessary area of the page, with the oldest transaction related to the oldest as the system execution log, and determine whether the data belonging to the unnecessary area can be deleted. A tuple data validity determination procedure for determining whether or not,
When the transaction identifier is older than the oldest transaction identifier according to the tuple data validity determination procedure, the data belonging to the unnecessary area is determined as the deleteable tuple data and the deleteable tuple data Tuple data deletion reservation procedure for performing deletion reservation for
When there is the tuple data that can be deleted by the tuple data deletion reservation procedure, an index data determination procedure for determining whether there is index data assigned to the deleteable tuple data;
An index data deletion procedure for deleting the index data when there is the index data assigned to the removable tuple data by the index determination procedure;
In the index determination procedure, when the index data is deleted by the index data deletion procedure or when the index data is not in the deleteable tuple data, the deleteable tuple data is released to free up an empty area. The unnecessary area repair program according to claim 11, wherein the computer executes execution of securing. The management procedure is to update the data stored in the database by updating the page,
A tuple length calculation procedure for calculating a tuple data record to be newly added to the database as a result of the update, as a tuple data length;
When the recordable tuple data length is calculated by the tuple length calculation procedure, if the tuple data record of the recordable tuple data length can be inserted into the reserved space area, it is inserted into the space area. And update the page of the inserted tuple data,
When the tuple data having the calculated tuple data length cannot be inserted into the empty area and the tuple data having the tuple data length can be inserted into the last page of the database, the tuple data is inserted into the last page. ,
When the tuple data having the calculated tuple data length cannot be inserted into the empty area and the tuple data having the tuple data length cannot be inserted into the last page of the database, a new page is added to the database. 13. The unnecessary area repair program according to claim 12, further comprising: causing the computer to execute the step of securing and adding the tuple data having the tuple data length. A table total management procedure in which the management procedure manages the unnecessary area generated in the database for each table of the data;
The table total management procedure further comprises a table unnecessary area storage procedure for storing the table while updating the size of the unnecessary area for each table.
The table total management procedure is as follows:
Based on the size of the unnecessary area for each table and the ratio of the unnecessary area in the table, at least a start threshold value for starting the unnecessary area release processing procedure or an end threshold value for ending the unnecessary area release processing procedure With one side,
12. The unnecessary area repair program according to claim 11, which causes a computer to control the unnecessary area release processing procedure according to the start threshold value or the end threshold value. The unnecessary area ratio determining procedure includes:
When the unnecessary area is generated in the page and the preset threshold is exceeded, an unnecessary area releasing process reservation procedure for performing a reservation for executing the unnecessary area releasing process procedure is further provided. ,
The unnecessary area release processing reservation procedure is:
12. The unnecessary area repair program according to claim 11, which causes a computer to control the unnecessary area release processing procedure based on a processing load of the database.

Images