JPS6314385B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention provides a database relocation processing method;
In particular, when reorganizing records in a hierarchically structured database, data information and pointers are temporarily separated and stored in main memory in order to minimize I/O accesses that occur during pointer linkage processing. The present invention relates to a data base relocation processing method in which the above-described pointer combination processing is performed on the device, and then the newly obtained pointer is added to predetermined data information.

In a structured database, pointers are created to clarify the hierarchical relationship between records and indicate their physical addresses, and one record is made up of a pair of data information and a pointer. . When reading/writing a record, the page containing the record is transferred to the main storage device, the record is read/written, and then returned to the database.

For this reason, consideration is given to ensuring that multiple records with a hierarchical relationship exist on a common page as much as possible. That is, when record types A and B have a hierarchical relationship as shown in FIG. It is stored in _S1 . However, assuming a network model, as processing progresses, records that exist in other storage areas S ₂ are generally - Type C has a hierarchical relationship with record type B, and is connected by set type β.

FIG. 2 represents the relationships shown in FIG. 1 in correspondence to individual records.

When the relationship shown in Figure 1 exists, records a ₁ , a ₂ , _{a 3} belonging to record type A, records b ₁ , b ₂ , b ₃ , b ₄ , b ₅ , b ₆ belonging to record type B
For example, on page P ₁ , in the form a ₁ → b ₁ → b ₂ → a ₁ ,
Even if records are stored on page P ₂ in the form a ₂ → b ₃ → b ₄ → a ₂ and on page P ₃ in the form a ₃ → b ₅ → b ₆ → a ₃ , records c ₁ Between b ₁ , b ₃ , b ₅ and record
Generally, c ₂ , b ₂ , b ₄ , and b ₆ do not exist on the same page.

Therefore, when performing processing that uses a join between record types C and B, record type C
Reorganize the arrangement between records c ₁ , b ₁ , b ₃ , b ₅ and between records c ₂ , b ₂ , b ₄ , b ₆ in order to improve the relationship between and B and increase access efficiency. However, for example, each record is made to exist in the same page as much as possible, and connections are established using pointers that exist in each record. Records are also updated, added, and deleted, and even in such cases, it is necessary to establish (newly create) connections between records using pointers that exist in each record.

In assembling the combination of pointers, it is necessary to access each record each time, as the pointers have conventionally existed integrally in each record. That is, for example, in the case of the relationship shown in FIG. 2, the illustrated pages P ₁ , P ₂ ,
Since P ₃ and P ₄ are different, in conventional processing, while tracing and processing set type β,
In the case of tracing record b ₁ → b ₃ → b ₅ , and record b ₂
→b ₄ →b ₆ , each page P ₁ , P ₂ ,
_P3 needs to be accessed repeatedly, which increases the frequency of I/O accesses to the database, making processing extremely complicated. In particular, when the number of records is large, the number of I/O accesses becomes extremely large.

The present invention aims to solve the above-mentioned points, and the data base relocation processing method of the present invention is such that a page consisting of a plurality of records in which data information and pointers are paired is a unit of transfer. It is configured to have a database given as , transfer the page from the database to the main storage, process it, and return the page to the database, and based on the returned result. In a data processing system that processes new pointer updates, a new set of
When reconfiguring to a new set type, one or more pages containing records that require reconfiguration with the new set type are transferred onto main storage, and data information and pointers of the records that require reconfiguration are separated. The data information is stored at the new physical address on the database, the pointer is left on the main memory, and the new physical address and the old physical address where the data information of the record requiring reconfiguration is stored are stored. A new pointer corresponding to the new physical address is created by creating a correspondence with the address, and the new pointer is appended to the data information existing at the new physical address, so that multiple records conform to the new set type. It is characterized by being reconfigured to configure pages. This will be explained in detail below.

3 to 7 are explanatory diagrams sequentially explaining the processing procedure of an embodiment of the present invention, and FIG. 8 shows an embodiment of the processing method of the present invention.

Now record on page P ₁ as shown in Figure 2
a ₁ , b ₁ , b ₂ are records a ₂ , b ₃ , b ₄ on page P ₂
But records a ₃ , b ₅ , b ₆ on page P ₃ are on page
Assume that records c ₁ and c ₂ exist on P ₄ , and regardless of whether or not record types A and B exist on the same page, record types C and B should be kept on the same page as much as possible. The explanation will be given assuming that the relocation according to the new set type β is obtained.

[] First, set record type A.
By tracing the type α, extracting the record type B, and extracting the record type C, the records are sequentially stored in the main storage device (which in this specification includes the work file) as shown in FIG. You can obtain the outputted initial information DB Data.
In FIG. 3, 1 represents record type information, 2 represents record data information, 3 represents relative position information within the record of the next pointer, 4 represents physical position information of the record, and 5 represents next pointer information. In the case of the setting example corresponding to Figure 2, the records to be relocated are a ₁ , b ₁ , b ₂ , a ₂ , b ₃ , b ₄ , a ₃ ,
b ₅ , b ₆ , c ₁ , and c ₂ , and since these relationships are reconstructed, an information unit having information 3, 4, and 5 is created only for each record in FIG. 3. That is, for example, records
Regarding b ₁ , information 5 makes it clear that the record b ₁ has a relationship with record b ₃ . Regarding record c ₁ , the record c ₁
is revealed to have a relationship with record _b1 .

[] Then record a ₁ to be relocated,
For b ₁ , b ₂ , a ₂ , b ₃ , b ₄ , a ₃ , b ₅ , b ₆ , c ₁ , c ₂ , record c ₁ , b ₁ , b ₃ , b ₅ as much as possible
are located on the same page, records c ₂ , b ₂ , b ₄ ,
Store the data information at a new physical address on the database so that b ₆ is located on the same page. The new physical address is recorded in records c ₁ , b ₁ ,
b ₃ , b ₅ and c ₂ , b ₂ , b ₄ , b ₆ are located on the same page as much as possible, and it does not matter which page records a ₁ , a ₂ , a ₃ are located on. You can think that. Record B ₁ or above stored above
Regarding _c2 , a correspondence table (RC Data-1) of old and new physical addresses is created as shown in FIG. 4A. In the figure, 6 indicates old physical address information, and 7 indicates new physical address information. Note that the address with a dash is the new physical address, and the address without the dash is the old physical address. At the same time, as shown in FIG. 4B, pointer information (RP Data-1) is created according to the set type β. The pointer information RP Data-1 can be considered to be a new address given as information 4' to the information unit having information 3 shown in FIG.

[] Next, as shown in Figure 5, the old and new physical address correspondence table RC Data-1 and the pointer information RP
Data-1 is rearranged in the correct order using the old physical address of the record as a key, and the new and old physical address correspondence table RC Data-2 and pointer information RP
Create Data-2. In addition, in the case of drawings, the old and new physical address correspondence table RC Data-1 and RC
It is the same as Data-2.

[] Next, the new and old physical address correspondence table RC Data
While comparing the record old physical addresses of pointer information RP Data-2 and pointer information RP Data-2, as shown in FIG. -Make 3.

[] Finally, as shown in FIG. 7, pointer information RP Data-4 is created by rearranging pointer information RP Data-3 in the normal order using the new physical address information 4' of the record as a key. The pointer information RP Data-4 includes records b ₁ , b ₂ ,
It can be seen that the data information of b ₃ , b ₄ , b ₅ , b ₆ , c ₁ , and c ₂ is pointer information arranged in the order of the new physical addresses stored first.

[] The pointer information RP Data-4 created on the main memory as described above is individually added to the new physical address where the corresponding data information is stored. The new physical address is pointer information
This can be determined from the record physical position information 4' in RP Data-4 and the intra-record relative position information 3 of the next pointer.

[] Through the above processing [] or [], the rearrangement for set type β is obtained. Note that the pointer for set type α is
It can be considered that when the above process [] is executed and the data information is returned to the database, it is added to the data information and stored.

FIG. 8 shows an embodiment of the present invention that executes the processes [I] to [] described in connection with FIGS. 3 to 7 above. 8 in the figure is the database,
9 is a data processing device which typically shows the information processing process on the main memory; 10
1 to 16 are processes (or processing blocks) corresponding to the above processes [I] to [], 17 to 23
are information buffers, and the information DB Data, RC Data-1, and RP Data shown in Figures 3 to 7 are respectively information buffers.
-1, RC Data-2, RP Data-2, RP Data
-3, RP Data-4 indicates what is temporarily set.

As is clear from FIG. 8, in the pointer linking process due to record relocation, it is necessary to access the database 8 in process [I],
Only in processing [], processing []. Then, there is no need to access the same page twice.

As explained above, according to the present invention, storage of record data information and pointer linking processing are separated, and pointers are created by replacing old and new pointers on the main memory. . Therefore, the frequency of I/O access during relocation is significantly reduced, and the processing range can be localized (within the same page) during subsequent processing, improving processing efficiency. Furthermore, there is no need for a relocation processing program that takes into consideration a database with a complex network structure.

[Brief explanation of the drawing]

1 and 2 are explanatory diagrams for explaining the prerequisite problem of the present invention, FIGS. 3 to 7 are explanatory diagrams for sequentially explaining the processing procedure of an embodiment according to the present invention, and FIG. 8 is an explanatory diagram for explaining the processing procedure of the present invention. An example of a method is shown. In the figure, a ₁ , a ₂ , a ₃ , b ₁ , b ₂ , b ₃ , b ₄ , b ₅ , b ₆ ,
c ₁ and c ₂ are records respectively, 8 is a database, 9
10 represents a data processing device, 10 to 16 represent processing (or processing blocks), and 17 to 23 represent information buffers, respectively.

Claims (1)

[Claims] 1 A database is provided in which pages each consisting of a plurality of records consisting of pairs of data information and pointers are given as transfer units, and the pages are transferred from the database to the main storage and then processed. In a database relocation processing method in a data processing system configured to return the page to the above database based on the returned result, a new set
When reconfiguring to a new set type, one or more pages containing records that require reconfiguration with the new set type are transferred onto main storage, and data information and pointers of the records that require reconfiguration are separated. The data information is stored at the new physical address on the database, the pointer is left on the main memory, and the new physical address and the old physical address where the data information of the record requiring reconfiguration is stored are stored. A new pointer corresponding to the new physical address is created by creating a correspondence with the address, and the new pointer is appended to the data information existing at the new physical address, so that multiple records conform to the new set type. A database relocation processing method characterized in that the database is reconfigured to configure pages.