JPH0823838B2 - Index file group division processing method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Summary] Regarding processing when a group of pages forming an index file is divided into a plurality of groups in a computer, the interrupted division processing step is performed in the normal access processing process. For the purpose of a division processing method that can detect and recover the division processing from the detected point, a plurality of pages that store data, a primary group that aggregates the plurality of pages, and a plurality of the primary groups are provided. Group division processing for an indexed index file including a grouped secondary group, group information that is provided for each of the primary and secondary groups and positions the group in the group hierarchy, and a super record that stores page configuration information within the group The method is to select a division source primary group and set it as a super record of the primary group. A flag is set to update the page information and group information of the super record of the secondary group which is the upper order of the selected primary group, and the division destination primary group is selected to become the super record of the primary group. The empty page information is updated and the in-process flag is set, the target page in the division source primary group is copied to the division destination primary group, the index and the super record of the division source primary group are updated, and the division is performed. When all the copy target pages in the original primary group have been copied to the division destination primary group, the division processing information in the super record of each group involved in the division is reset.

[Industrial applications]

The present invention relates to processing when a group of pages forming an index file is divided into a plurality of groups in a computer.

Data is held in a page, the page is classified into a group of a predetermined number of pages according to the value of the index of the data, and the page is classified as a primary group, and a predetermined number of primary groups form a secondary group, for example. The index file is composed of the required number of secondary groups, each primary,
A method is known that enables efficient indexing by evenly distributing data pages and empty pages to secondary groups as much as possible.

[Problems to be solved by conventional technology and invention]

 FIG. 4 is a diagram for explaining the structure of the index file.

The index file 3 is composed of an index section 7 and a data section 8. The minimum unit of the data section 8 is a page, and each page generally stores a plurality of data records.

A predetermined number of pages (3 in the figure) constitutes a primary group 1, a predetermined number (2 in the figure) of a primary group 1 constitutes a secondary group 2, and the required number of pages is determined by the amount of data. The data group 8 is formed by the secondary group 2.

The data section 8, each secondary group 2, and each primary group 1 are respectively provided with super records 6, 5 and 4 for holding information about the free space status of the lower-order groups or pages constituting them and the key order. Set up.

FIG. 5 is a diagram for explaining the super record. The figure is 1
Shows an example of each secondary group, and the super record 5 of the secondary group 2 shows the group numbers of the non-empty groups in the key order of the stored data for the primary group 1 forming the secondary group, and then the groups of the empty groups. Numbers are shown together with appropriate blank indications (in the figure, blank blank numbers are shaded to distinguish them).

Further, the super record 4 of the primary group 1 shows the page numbers of the pages constituting the super record 4 in the same key order and empty display as described above.

It should be noted that the numbers in the page of the figure indicate the key value of the record when the data record is stored, and the part without the number is empty.

Therefore, in the case of key-order access, pages can be referenced in key order by referring to the super record.

Also, in the case of retrieval by key, the page containing the required key value is determined and accessed by the index part. With the above configuration, data is stored in the page in key order, and those pages are stored in key order. By grouping so that the amount of data is distributed to each group as evenly as possible, until a page that may hold the data corresponding to the search key is reached and the data on that page is searched By making the number of steps for each key value uniform, it is possible to maintain good overall search processing efficiency.

In addition, each primary group 1 is provided with an appropriate number of empty pages to cope with the addition of data. However, if the addition of data causes a large imbalance in the amount of data between the groups, it is possible to add an appropriate time. A primary group with a large amount of data
Equalize by splitting into two groups using an empty primary group.

The division process for equalization of the index file is processed by an appropriate program in parallel with the use of the index file in the original work. At that time, if the division process is interrupted due to some kind of failure, the index part and the data part will become inconsistent, so normally so-called log file update recording is used to return to the state before processing. A known recovery process for restoration is performed, and the process of the business using the index file cannot be performed until after the recovery process is completed.

It is an object of the present invention to provide a division processing method capable of detecting a suspended processing status of division processing in the course of normal access processing and recovering the division processing at the detected point.

[Means for solving problems]

 FIG. 1 is a flow chart of processing showing the configuration of the present invention.

The figure shows the flow of page group division processing in the index file, and 10 to 19 are processing steps.

[Work]

Each super record of each group of pages is provided with a division processing information column indicating the division processing status of the group or page whose status is displayed in the super record, and the display of division processing and the division source Also, display the group number or page number of the division destination.

In the division processing, first, in processing step 10, division processing is set in the division processing information column of the super record of the division source primary group.

In processing step 11, the group status display is updated for the super record of the secondary group to which the primary group belongs, and the division processing information column is set. Processing step 12
Then, for the selected super-record of the division-destination primary group, the page state display of the empty page to be the division destination is updated and the division processing information column is set.

Then, in processing step 13, the data is copied from the division source page to the division destination page, and in the processing step 14, the index part is updated to the state after division.

Next, in process step 15, the page state display of the page of the division source of the super record of the division source primary group is updated to an empty page.

In the case of performing the page movement between the same primary groups after being identified in the next processing step 16, the processing is performed from the processing step 10 by the above procedure, and when the movement processing is eliminated, the processing proceeds to the processing step 17, and the division is performed. Destination, each primary group of division source,
The division processing information column of each super record is reset in the order of the secondary group, and the processing ends.

With the above processing method, even if the division processing is interrupted, the necessity of division processing recovery can be identified by the division processing information column of the super record on the access route. The information in the file can complete the recovery.

〔Example〕

FIG. 2 is a diagram for explaining a configuration example of a super record when implementing the present invention, together with an example of an index file.

In the figure, 20 is the super record of the secondary group, 21,
22 is an example of the structure of the super-record of the primary group, and the left part of the super-records 20 to 22 is a status display showing the key order and the empty state of the lower-level group or page as in the conventional case.

Further, the right end portion is the division processing information column 23, and the display N of the division processing in-progress display 24 is a non-processing display and the display is a division processing in progress display.

The number display 25 on the right side is a part for displaying the number of the page or group to be processed together with the division source and the division destination when the division processing display 24 is Y. An arrow from the original to the division destination is shown, and * indicates that no number is set.

The figure shows a primary group 00 consisting of pages 000 to 002 under the control of super record 21 and a primary group 01 consisting of pages 010 to 012 under the control of super record 22 to form a secondary group having super record 20. Shows an example of an index file that has a data section that has been
There is data in the page and all pages in the primary group 01 are empty pages.

There is an index part 26 corresponding to this data part, and a section consisting of a page number column showing the page number and a maximum key column showing the maximum key value of the data held on that page is in ascending order of the key values of the maximum key column. Arranged in.

The primary group 00 in the example of FIG. 2 is replaced by the primary group
A situation in which the division processing for division into 00 and 01 is executed according to the flow of the processing in FIG. 1 will be described below with reference to FIG. In the figure of each step in FIG. 3, the updated portion is indicated by a thick line for the sake of explanation.

In the division processing, for example, it is assumed that the page with a large key value is moved, and in the processing step 10 of FIG.
As shown in FIG. 10A, Y is displayed in the division process information column and “2” is displayed as the division source page number to be moved (relative number is displayed in the group, and the same applies in other cases). To do.

Next, in processing step 11, the two to which the primary group belongs
Regarding the super record of the next group, the division destination primary group is selected, the primary group 01 of the group status display is updated to be in use as shown in FIG. 3B, and the division processing information column is set. .

In processing step 12, as shown in FIG. 3 (c), with respect to the selected super-record of the division-destination primary group, the empty page 010 is selected as the division destination, the page status display is updated, and the division processing information field is displayed. Make settings.

Then, in processing step 13, the data is copied from the division source page to the division destination page, and the state shown in FIG.

Next, in processing step 14, as shown in FIG. 3 (e), the index portion is updated to match the state after division.

Next, in processing step 15, as shown in FIG. 3 (f), the page state display of the super-record of the division-source primary group is updated to make the division-source page 002 an empty page.

It is identified in the next processing step 16 and in this case one more
Since moving the page provides a better balance of the data amount between both primary groups, the process returns to processing step 10, and the processing of moving the page 001 having the next largest key value is performed by the same procedure as described above, and the processing is performed again. In step 15, page 001 is made an empty page as shown in FIG.

If it is identified in the processing step 16 that there is no moving page, the process proceeds to the processing step 17 and the division processing of each super record of the division primary group, the division primary group in the processing step 18 and the secondary group in the processing step 19 The information column is reset to the state shown in FIG. 3 (h), and the process ends.

According to the above processing method, when the division processing is interrupted before the update of the index part in processing step 14 is completed, each super record is simply restored, or when the division processing is interrupted after the index part is updated. The recovery processing can be performed by continuously completing the division processing after the interruption point in the same manner as described above.

Whether or not such recovery processing is necessary is identified by checking the division processing information column of the super record on the access route in the process of key-order access to the index file, and at that time, the maximum key of the division source primary group. It is possible to perform an accurate recovery process by discriminating whether the index part is updated or not, depending on whether or not the page and the contents of the index part match.

Further, in the case of the search by the key, the processing may be performed according to the contents of the index portion, and even if the recovery processing is performed before and after that, no contradiction occurs.

〔The invention's effect〕

As is apparent from the above description, according to the present invention, in a process of dividing a group of pages forming a computer index file into a plurality of groups, the interrupted division process step is performed in a normal access process process. Since the detection can be performed and the division processing can be recovered from the detected point, there is a remarkable industrial effect of improving the utilization efficiency of the computer.

[Brief description of drawings]

 FIG. 1 is a flow chart of processing showing the configuration of the present invention, FIG. 2 is an explanatory diagram of super records of the present invention, FIG. 3 is an explanatory diagram of processing steps of the present invention, and FIG. FIG. 5 is an explanatory diagram of a super record. In the figure, 1 is a primary group, 2 is a secondary group, 3 is an index file, 4, 5, 6, 20 to 22 are super records, 7 and 26 are index parts, 8 is a data part, and 10 to 19 are processes. Steps, 24 indicates divisional display, and 25 indicates number display.

Claims (1)

[Claims] 1. A plurality of pages for storing data, a primary group including the plurality of pages, a secondary group including a plurality of the primary groups, and each of the primary and secondary groups. A group division processing method for an indexed index file consisting of group information that positions the group in the group hierarchy and a super record that stores page configuration information in the group, in which the division source primary group is selected and the primary group is selected. The processing flag is set in the super record of the group, the page information and the group information of the super record of the secondary group which is the upper order of the selected primary group are updated, and the primary group to be divided is selected to select the primary group. Update free page information and set a processing flag in the super record of the group, and The target page of the division destination 1
When copying to the next group, updating the index and the super record of the division source primary group, and copying all the copy target pages in the division source primary group to the division destination primary group, A method for dividing a group of index files, characterized in that the division processing information in the super record of the group is reset.