KR960001080B1 - A data backup method for a real time database management - Google Patents

Abstract

storing a physical offset of a data base file during performing a data backup block function; if there is the data backup request, opening the file, and regulating a boundary of the block of the file system after counting the physical offset; and outputting a corresponding data to the data base file by regulating the block boundary of the output data.

Description

Fast Data Backup Method in Real-Time Database Management System (DBMS)

1 is a distributed structure diagram of an electronic exchanger DBMS including the subject matter of the present invention.

2 is a conventional data backup flowchart.

3 is a data backup flowchart according to the present invention.

The present invention relates to a high speed data backup method in a real time database management system (hereinafter referred to as "DBMS") applied to an electronic switch.

In a conventional voice exchange, data that must be kept permanently, such as subscriber data, is often referred to only by real-time events such as call processing after loading into main memory from disk at system startup. Particularly, in case of subscriber data, occasional change is required by the operator in case of new registration of subscriber, cancellation of subscriber, and request for change of station number by subscriber, the effect of data backup due to change on subscriber data has little effect on the system.

Therefore, the DBMS of the conventional voice exchanger has been used without considering the normal primitive real-time provided by the file system for data backup since there is no need for real-time processing for data backup. Therefore, the conventional data backup method cannot cope with real-time data backup.

Unlike the voice exchange, the mobile exchange frequently moves the subscriber data between the exchanges so that events such as data insertion and deletion occur frequently. In this case, a change of subscriber data needs to be backed up to disk, so a method of processing data backup is required at high speed.

Accordingly, an object of the present invention is to provide a high-speed data backup method that can be applied to a system in which events frequently occur due to data insertion and deletion, such as a mobile exchange, to improve the performance of the system.

In order to achieve the above object, the present invention provides a processor for performing a distributed call processing function (ASP); Processors (NTP, INP) that perform centralized call processing; A processor (OMP) that performs centralization functions related to operation and maintenance; And a data backup method applied to an electronic switch having a database management system (DBMS) mounted in each processor, the data backup method comprising: storing a physical offset of a database file when a function of a data back block (DBBG) is started; Stage 1 ; A second step of opening a file upon calculating a data backup request, calculating a physical offset of a raw disk, and adjusting block boundaries within the file system; And a third step of adjusting the block boundary of the output data and outputting the corresponding data to the database file.

Hereinafter, with reference to the accompanying drawings will be described an embodiment according to the present invention;

1 is a structural diagram of a main memory DBMS (distributed structure) having a distributed structure in which an electronic switching system is configured to provide a data backup function.

As shown in the figure, the electronic switch is composed of three subsystems: an Access Switching Subsystem (ASS), an Interconnection Network Subsystem (INS), and a Central Control Subsystem (CCS). ASS consists of Access Switching Processors (ASPs) that perform distributed call processing functions, and INS consists of INP (Interconnection Network Process) and NTP (Number Translation Processors), which perform part of the centralized call processing function. CCS performs centralization functions related to operation and maintenance, and consists of Operation and Maintenance Processor (OMP) and Man-Machine Processor (MMP).

Among the processors, OMP DBMS is composed of four blocks of DBBG (Data Base Backup Group), DBSG, DBTG (Data Base Transaction Group) and DBKG (Data Base Kernel Group), and MMP's DBMS is DBQG (Data Base Query). Group), DBSG, DBTG, and DBKG, with four blocks. The remaining processors consist of three blocks: DBSG, DBTG, and DBKG. The DBSG performs the user matching function and the communication between the DBMS, and the DBQG performs the processing of the database query word coming through the MMP. DBBG performs change control, data backup and recovery of duplicate data, DBTG performs transaction processing, and DBKG actually retrieves, alters, inserts and deletes the database.

Among them, DBBG is closely related to the backup proposed in the present invention. DBBG receives the wall-up request of the changed contents from DBSG when data change occurs in each processor and uses the system calls provided by the file system to change the disk information. Back up to

2 shows a process of backing up data in a DBMS of a conventional voice changer.

When DBBG receives a data backup request, DBBG uses the normal interface provided by the operating system's file system to back up the data.

First, the descriptor of the database file is obtained using an open system call provided by the file system (1). At this point, the file system needs several disk I / Os to find the database from the super block.

It then computes the attribute offset in the database file with the address of the attribute (2).

Next, use the lseek system call to move the file's pointer from the database file to the corresponding part of the attribute (3), and use the data write system call to write the database file up to the size of the attribute. Output (4). The conventional DBMS does not match the block boundary of the data size and the offset of the data output portion, so that the file containing the changed data is read from the disk into the cache buffer from the file system and the contents of the user buffer are cached. Reflect it in the buffer and write it back to disk.

Therefore, more than one disk I / O is required for one data output.

Finally, close the database file using the close system call (5). At this time, an entry of the file descriptor table allocated for opening a file is returned and the contents of the buffer are flushed to the disk. A single I / O is required to flush the contents of the buffer to disk. The data backup result is then sent to the process requesting the data backup.

It can be seen that the above-described conventional data backup method requires a plurality of disk input / outputs (at least four times) to output one data.

3 is a flowchart of a database backup method according to the present invention, and the database backup procedure is as follows.

When DBBG receives a data backup request, DBBG uses the raw interface provided by the operating system's file system for high-speed backup of the data. The raw interface allows the user to know the physical address of the disk and to directly input and output data to and from the disk.

First, at the start-up of the function of the data backup block (DBBG), the physical offset of each database file is read and stored in a static structure (10).

Second, open the raw disk using the raw interface system call to output the data on demand for data backup (11). I / O is not required when opening a raw disk.

Third, in step 10, the physical offset of the raw disk is obtained by adding the physical offset of the database file stored in the static structure and the relative offset of the attribute (12). At this time, the block boundary of the physical offset of the raw disk is adjusted (13).

Fourth, call the seek system call with the adjusted offset to move the file pointer (14).

Fifth, the block boundary of the output data is adjusted (15).

Sixth, when an attribute is backed up to disk, the image of the disk exists in the main memory, so the changed system and the surrounding data to match the block boundary are called together to write to the disk (16). . Here, in the case of outputting data, since the block boundaries are matched in the previous processing steps 13 and 15, the data can be backed up to disk with one output.

Seventh, use the close system call to close the raw disk and transfer the data backup results (17, 18). When releasing the raw interface, there is no need to flush the contents of the buffer, thus eliminating the need for disk I / O.

As a result, the steps 11 to 18 show that the disk input / output is required only once to output the data to the disk. In general, the disk I / O time is hardly affected by the amount of data input / output at one time (within 5 blocks). However, the disk I / O time is proportional to the number of input / output disks, which shows that the disk backup can be performed more effectively than the conventional method.

Therefore, the present invention carried out by the above processing procedure can be expected to improve the performance several times (more than four times) compared to the data backup used in the conventional DBMS, and the addition / deletion of data like the mobile communication exchange In the frequently occurring systems, the method proposed in the present invention has a significant contribution to improving the performance of the system.

Claims (2)

A processor (ASP) that performs distributed call processing functions; Processors (NTP, INP) that perform centralized call processing; A processor (OMP) that performs centralization functions related to operation and maintenance; And a data backup method applied to an electronic switch having a database management system (DBMS) mounted in each processor, the data backup method comprising: storing a physical offset of a database file at the start of performing a function of a data backup block (DBBG); Step 1 (10); A second step (11 to 13) of opening a file upon calculating a data backup request, calculating a physical offset of a raw disk, and adjusting a block boundary inside the file system; And a third step (14 to 18) of adjusting the block boundary of the output data and outputting the corresponding data to the database file. 16. A high speed data backup method in a database management system (DBMS). The method of claim 1, wherein the file opening process during the second steps (11 to 13) is performed by using a raw interface function. .