KR100472204B1 - Logical volume manager system having mapping table structure and snapshot method using the same - Google Patents

Abstract

According to the present invention, in providing a non-stop service in a logical volume management system, it is necessary to determine the necessity of copy-on-write (COW) during a write operation performed after a snapshot is generated and to determine whether to generate a COW during a snapshot delete operation. In the logical volume management system having a mapping table structure in the logical volume management system that can significantly improve the performance of the system and the reliability of data access services by significantly reducing the number of disk I / Os, and an efficient snapshot method using the same. It is about.

The mapping apparatus of the present invention has a structure in which a mapping entry is formed by connecting mapping information for the original volume and mapping information for each snapshot in a pair for each logical data block or extent.

Description

Logical Volume Management System with Mapping Table Structure and Efficient Snapshot Method Using the Logical Volume Management System {LOGICAL VOLUME MANAGER SYSTEM HAVING MAPPING TABLE STRUCTURE AND SNAPSHOT METHOD USING THE SAME}

The present invention relates to a logical volume management system and an online snapshot method having a mapping table structure for providing a non-stop service, and more particularly, to copy-on-write during a write operation performed after a snapshot is created. Logical volume management system that can improve system performance and reliability of data access service by drastically reducing the number of disk I / O required to determine the need of Write) and COW occurrence during snapshot delete operation. A logical volume management system having a mapping table structure in and an efficient snapshot method using the same.

Recently, the computing environment has been actively developed such as electronic commerce and web services due to the development of the Internet. These Internet services should always be provided online to users, and 24 * 7 * 365 nonstop service has become a very important service element.

In order to provide such a nonstop service, snapshot-based online backup support is essential.

Snapshots are used to support this online backup, and store and maintain the state of existing data at a specific point in time that you want to back up.

In addition, if the data changes after the snapshot is created, a new data area is allocated and the data at the time of snapshot creation is copied to the newly allocated data area to maintain the existing data contents. This method is called copy-on-write (hereinafter referred to as COW).

In general, the method of supporting snapshots through COW in the logical volume manager maintains a mapping table to map logical addresses to physical addresses, and also creates a snapshot mapping table for logical mapping to a logical volume. Keep as many snapshots as the manager can provide.

In the following description, the logical volume manager supports 3 snapshots for each logical volume as an example.

1 is a view showing the structure of a volume header area for providing a snapshot of the conventional method.

The volume header 130 of each disk 120 constituting the logical volume 110 includes a volume label 131, a data block, or an extent that holds information about the volume and the disk. Free Space Bitmap (132) for Allocation Management, Source Mapping Table (133) for Translation of Logical Address of the Volume into Physical Address, Mapping Table (1) for Snapshot 1 ), Mapping table 135 for snapshot 2, and mapping table 136 for snapshot 3. The remaining area becomes a data area 140 for storing data.

2 illustrates a logical view of mapping information of a conventional method.

In FIG. 2, when a snapshot creation command occurs, the volume manager copies the original mapping table to a snapshot mapping table area corresponding to the corresponding snapshot number in an appropriate snapshot creation order. When a change operation on the data for the original volume is requested, the volume manager compares the original mapping table information with the corresponding snapshot mapping table information to determine whether the change operation is the first change operation on the data block after the snapshot is created. Done.

If all three snapshots are created, the original mapping table and all three snapshot mapping tables are compared. At this time, if there is a snapshot having the same mapping information as the original mapping information, COW is performed.

In the example of FIG. 2, a change operation for logical block (or extent) 2 is requested. In this case, since the original mapping information and the snapshot mapping information are the same, the COW proceeds.

On the other hand, the size of the original mapping table is determined according to the capacity of the disk, the size is usually composed of a number of basic blocks (basic block, 512 bytes).

In general, since disk I / O is composed of basic block units (512 bytes), in the structure of the conventional mapping table as shown in FIG. 2, the original mapping information and each snapshot mapping information for a specific data block are stored. In order to compare with each other, at least four disk I / Os are required for each mapping information.

This significantly increases the performance of the system by increasing the mapping time by up to four times when snapshots are taken, and creates significant differences in mapping time when there are no snapshots and when snapshots exist. Difficult to provide service.

Also, when deleting a snapshot, the mapping information of the snapshot is compared with the original mapping information and other snapshot mapping information. If the snapshot mapping information is different from other information, the area is released through the free space management bitmap. The process is necessary. In this case, if all three snapshots are created, four disk I / Os are generated to compare one mapping information, thereby degrading system performance.

Accordingly, an object of the present invention is to solve the above-described problems, and an object of the present invention is to provide an online snapshot function for supporting a non-stop service in a logical volume manager, and to check mapping information when data is changed or deleted. Logical volumes that reduce the number of disk I / Os performed for a single time to ensure reliable data service and improve system performance by keeping mapping time the same regardless of the presence of snapshots. The present invention provides a mapping table structure in a management system and an efficient snapshot method using the same.

A logical volume management system having a mapping table structure for achieving the object of the present invention is a pair of mapping information for the original volume and mapping information for one or more snapshots for each logical data block or extent. This is done by including a mapping table structure consisting of a single mapping entry.

In addition, an efficient snapshot method in a logical volume management system for achieving the object of the present invention, after reading the data of the mapping table in the unit of the basic mapping block to the memory and the original mapping information for each mapping entry in the mapping block. Creating a snapshot copying the snapshot mapping information area; If a change operation occurs on a specific data block after taking a snapshot, the mapping block for the data block is read into memory, and the original mapping information and each snapshot mapping information are compared in the corresponding mapping entry, and the same snap as the original mapping information is performed. A snapshot change step of performing copy-on-write on the snapshot if shot mapping information exists; And a snapshot that reads data from a mapping table into a basic mapping block unit and allocates a data block indicated by the target snapshot when the target snapshot mapping information and other mapping information are different for each mapping entry. The shot deletion step is made.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

3 is a block diagram of a logical volume management system to which the present invention is applied.

As shown in FIG. 3, a logical volume management system includes a general disk 310, a volume manager 320, a logical volume 330, a file system, or a database 340.

The disk 310 is accessed through a block number of the disk.

The volume manager 320 stores information necessary for managing the logical volume 330 on disks 310 designated by the user according to a user's request, and logically connects the disks to provide one large logical volume. do.

The logical volume 330 provided by the volume manager 320 is accessed via a logical block number.

The file system or database 340 recognizes a logical volume provided by a volume manager as a storage device, creates a file system on the logical volume, initializes a database, and accesses a logical volume. Use

On the other hand, Figure 4 is a view showing the internal structure of the disk 310 managed by the volume manager 320 according to the present invention.

The volume manager 320 generates a volume label 410 for storing information about a logical volume on each disk 310. It also maintains a Free Space Bitmap (420) to control block (or extent) allocation on the disk, with which physical block (or extent) 440 a particular logical block (or extent). The mapping table 430 is maintained as information about whether the correspondence is to correspond to.

Each disk 310 is a data space (a collection of physical blocks (or extents) for storing data, except for the volume header 450, which is a collection of volume labels, free space bitmaps, mapping table information, and the like. Data Area, 460).

5 is a view showing the structure of the mapping table 430 according to the present invention.

As shown in FIG. 5, the structure of the mapping table according to the present invention includes mapping information 510 for the original volume and mapping information 520 for the snapshot 1 for each logical data block (or extent) 560. ), A mapping entry 530 for snapshot 2 and mapping information 540 for snapshot 3 are connected in pairs to form a mapping entry 550.

In addition, each mapping information 510, 520, 530, and 540 is a pair of a disk ID 570 and a physical block (or extent) ID 580 within the disk for distinguishing various physical disks participating in the volume. It consists of.

The structure of the mapping table according to the present invention can be modified in various ways. For example, the disk ID area 570 allocates 8 bits to allow a maximum of 28 (= 256) physical disks per volume. The physical block (or extent) ID area 580 allocates 56 bits. If the size of the physical block is 4 KB, the physical disk size up to 4 KB * 256 = 268 Byte is supported.

Therefore, one mapping entry 550 has a size of 64 bits * 4 = 8 bytes * 4 = 32 bytes and stores 512/32 = 16 mapping entries in a basic block of 512 bytes.

That is, in a basic block (usually 512 bytes) processed by one disk I / O, a structure includes a plurality of mapping entries having original mapping information and snapshot mapping information for the same data block.

According to the mapping table structure of the present invention, if it is desired to determine whether to perform COW by comparing the original mapping information for each specific data block and each snapshot mapping information with each other, the original mapping information for each data block and each snapshot mapping Since all of the information is contained in one basic mapping block (512 bytes), it is possible to determine whether the original mapping information and the respective snapshot mapping information of the corresponding data block are the same with one disk I / O.

6 is a flowchart illustrating a process of copying contents of an original mapping table to a snapshot mapping table for a corresponding snapshot when a snapshot is generated based on the mapping table structure described above.

As shown in FIG. 6, the volume manager 320 first finds out the number of a snapshot to be created through a snapshot creation command (S601).

In addition, an exclusive mode lock is obtained for each 512-byte block of the mapping table, and the corresponding block is read into the memory (S602).

There are a total of 16 mapping entries in one 512 byte base block. Therefore, the original mapping information, which is the first 8 bytes of each mapping entry, is copied to an area separated by 8 * snapshot numbers (S603).

That is, the first snapshot is copied 8 bytes away, that is, immediately after the original mapping information, and the second snapshot is copied 8 * 2 = 16 bytes away.

When copying of 16 mapping entries is completed, the lock is released (S604), and it is checked whether the corresponding block is the last mapping block on the corresponding disk (S605).

If the result of the test is not the last block, the block number is increased and the above-described processes are repeated (S606).

If it is the last block, it is checked whether the corresponding disk is the last disk (S607).

If it is not the last disc, increase the disc number and repeat the above-mentioned steps (S608). If it is the last disc, the copy process is completed.

After the snapshot is created, if a change operation on the data occurs, it is determined whether the change operation occurs first for the corresponding data block (or extent) after the snapshot is created. If not, no COW is performed.

At this time, the process of determining whether to perform COW is as follows.

If a change operation on the data block (or extent) has not occurred once since the snapshot creation, the original mapping information and the mapping information for the snapshot must be identical. Therefore, if original mapping information and mapping information for a specific snapshot are the same, COW should be performed.

All mapping information (including the original and all snapshots) for a particular data block (or extent) is stored in the same mapping block, so if the block is not in memory, the block is read from disk. Then, the mapping information position for the corresponding data block (or extent) in the block in the memory is calculated to compare the original and each snapshot mapping information.

If there is any snapshot with the same mapping information as the original result, the COW is executed.

COW is performed by requesting the free space manager to allocate a new data block (or extent) and copying the original data block (or extent) indicated by the original mapping information to the data block determined according to the request result.

The snapshot mapping information having the same mapping information as the original mapping information is changed to point to the newly allocated data block (or extent).

Therefore, it is possible to determine whether to perform COW according to data change with only one disk I / O to the mapping table.

On the other hand, when deleting a snapshot, the data block (or extent) allocated through the COW for the snapshot must be deallocated.

7 is a flowchart illustrating space allocation release when deleting a snapshot according to the present invention.

As shown in FIG. 7, the logical volume manager 320 first checks the snapshot number to be deleted by the delete command (S701).

In addition, an exclusive mode lock is obtained on a basic mapping block basis of 512 bytes of the mapping table, and the corresponding mapping block is read into the memory (S702).

If the snapshot mapping information of the identified number is different from all other mapping information (original mapping information and other snapshot mapping information) in each mapping entry of the read mapping block, the corresponding data indicated by the snapshot mapping information. The block (or extent) is deallocated (S703).

The fact that specific snapshot mapping information is different from all other mapping information means that the data block (or extent) indicated by the snapshot mapping information is an area allocated only for the corresponding snapshot.

When the deletion of the 16 mapping entries in the mapping block read through the allocation allocation process is completed, the lock is released (S704), and it is checked whether the corresponding block is the last mapping block in the corresponding disk (S705).

If the test result is not the last block, the block number is increased and the above-described process is repeated (S706).

If it is the last block, it is checked whether the corresponding disk is the last disk (S707).

If it is not the last disk, increase the disk number and repeat the above-mentioned process (S708). If it is the last disk, the deletion process is completed.

The online snapshot method in the logical volume management system according to the present invention as described above uses the mapping table structure of the present invention. The data of the mapping table 430 is read into the memory in basic mapping block units and then mapped. For each mapping entry 550 in the block, a snapshot creation step of copying the original mapping information into the corresponding snapshot mapping information area, and if a change operation occurs for a specific data block after the snapshot is created, mapping to the data block. After reading the block into memory, compare the original mapping information with each snapshot mapping information in the corresponding mapping entry 550, and if there is the same snapshot mapping information as the original mapping information, copy-on-write is performed on the snapshot. The snapshot change step is performed, and data of the mapping table 430 is read into memory in units of basic mapping blocks, and then each mapping entry 550 is read. To remove the target snapshot information comprises a map and that when the other map information different snapshot deletion step of unassigned blocks of data indicated by the object snapshot.

In addition, the snapshot generation step, the step of confirming the snapshot number to be created by the snapshot generation command, and obtaining the exclusive mode lock in the unit of the basic mapping block of the mapping table 430, and the mapping block to the memory Reading, copying the original mapping information for each mapping entry 550 included in the read mapping block into a snapshot mapping information area of the identified snapshot number, and all mappings in the mapping block. When the copy is completed for the entry 550, the step of releasing the lock, confirming whether the corresponding mapping block is the last block, and completing the copy of all the mapping blocks in the mapping table 430 through each of the above steps, such mapping Snapshot of the mapping table for all logical disk units by applying the snapshot copy steps for the table sequentially to each logical disk unit. The copying is completed.

In the snapshot changing step, after reading the mapping block corresponding to the data block on which the change operation has occurred into the memory, checking the mapping entry 550 for the changed data block in the read mapping block and the mapping entry 550. Compares the original mapping information and each snapshot mapping information with each other, and if there is any snapshot having the same mapping information as the original result, the Copy-On-Write operation is performed, and the snap having the same mapping information as the original is performed. If there is no shot, the Copy-On-Write operation is not performed.

In addition, the step of deleting the snapshot, the step of confirming the snapshot number to be deleted through the snapshot delete command, to obtain the exclusive mode lock in the unit of the basic mapping block of the mapping table 430, and the mapping block to the memory Reading and comparing the snapshot mapping information of the identified snapshot number with all other mapping information with respect to each mapping entry 550 included in the read mapping block, and if all are different, for the snapshot. Unassigning an allocated data block or extent, unlocking when all mapping entries in the mapping block are completed, and checking whether the corresponding mapping block is the last block and mapping each table through the above steps Complete the deletion of all mapping blocks in 430, and delete the snapshot of this mapping table for each logical disk device. Applying to the sequential steps to complete the deletion of the snapshot of the mapping table of all logical disk devices.

As described above, the efficient snapshot method in the logical volume management system according to the present invention connects mapping information for the same logical address to the mapping entry without separately storing the snapshot mapping table as many as the number of snapshots. By saving, you can reduce the number of disk I / Os during a write or snapshot delete operation after a snapshot is created, resulting in a significant improvement in system performance when accessing the original volume, and always with or without a snapshot. The same performance can provide access to the data to improve its reliability.

What has been described above is only one embodiment for implementing an efficient snapshot method in a logical volume management system according to the present invention, and the present invention is not limited to the above-described embodiment, and is claimed in the following claims. Without departing from the gist of the present invention, any person having ordinary knowledge in the field of the present invention will have the technical spirit of the present invention to the extent that various modifications can be made.

1 shows the structure of a volume header area for providing a snapshot of a conventional method.

2 illustrates a logical view of mapping information of a conventional method.

3 is a block diagram of a logical volume management system to which the present invention is applied.

4 is a diagram illustrating an internal structure of a disk managed by a volume manager according to the present invention.

5 shows a mapping table structure according to the present invention.

6 is a flowchart of copying a mapping table when creating a snapshot according to the present invention.

7 is a flow chart for de-allocating space when deleting a snapshot according to the present invention.

<Description of the symbols for the main parts of the drawings>

310: Disk 320: Volume Manager

330: logical volume 410: volume label

420: free space bitmap 430: mapping table

450: volume header 510: source mapping information

520,530,540: snapshot mapping information

550: mapping entry

Claims (7)

In a logical volume management system, A mapping table structure for setting original mapping information and snapshot mapping information to support an online snapshot, wherein the mapping table structure includes, for each logical data block or extent, mapping information for the original volume, one or more than two. A logical volume management system having a mapping table structure having a structure in which a pair of mapping information for a snapshot is connected to each other to form a mapping entry. The method of claim 1, The original mapping information and the snapshot mapping information have a mapping table structure comprising a pair of a disk ID for distinguishing several physical disks participating in the volume and a physical block or extent ID in the disk. Logical volume management system. The method of claim 2, 8 bits are allocated to the disk ID, 56 bits are allocated to the physical block or extent ID, and a plurality of mapping entries are included in the basic mapping block having a size of 512 bytes at the time of disk I / O. Logical volume management system having a table structure. The method of claim 1, wherein the logical volume manager supports online snapshots through the mapping table structure. A snapshot generation step of reading data of a mapping table into memory in a basic mapping block unit and copying original mapping information to a corresponding snapshot mapping information area for each mapping entry in the mapping block; If a change operation occurs on a specific data block after taking a snapshot, the mapping block for the data block is read into memory, and the original mapping information and each snapshot mapping information are compared in the corresponding mapping entry, and the same snap as the original mapping information is performed. A snapshot change step of performing copy-on-write on the snapshot if shot mapping information exists; And Snapshot which reads data of mapping table into memory in basic mapping block unit and deallocates data block indicated by target snapshot when mapping target information and other mapping information are different for each mapping entry. Deletion step; effective snapshot method in a logical volume management system, characterized in that consisting of. The method of claim 4, wherein the snapshot creation step, Confirming the snapshot number to be created by the snapshot creation command; Acquiring an exclusive mode lock in units of the basic mapping block of the mapping table and reading the mapping block into memory; Copying the original mapping information for each mapping entry included in the read mapping block into a snapshot mapping information area of the identified snapshot number; Releasing the lock when copying is complete for all mapping entries in the mapping block; Verify that the corresponding mapping block is the last block and complete the copying of all mapping blocks in the mapping table through each of the above steps, and apply the snapshot copying process for these mapping tables to each logical disk device sequentially to Efficient snapshot method in a logical volume management system comprising the step of completing a snapshot copy to the mapping table of the device. The method of claim 4, wherein the step of changing the snapshot, After reading the mapping block corresponding to the data block where the change operation occurred, into the memory, check the mapping entry for the changed data block in the read mapping block, and compare the original mapping information of the mapping entry with each snapshot mapping information. If there is any snapshot having the same mapping information as the original as a result of the comparison, the Copy-On-Write operation is performed, and if there is no snapshot having the same mapping information as the original, the Copy-On-Write operation is performed. Efficient snapshot method in a logical volume management system characterized in that not. The method of claim 4, wherein deleting the snapshot, Determining the snapshot number to delete through the delete snapshot command, Acquiring an exclusive mode lock in units of the basic mapping block of the mapping table and reading the mapping block into memory; For each mapping entry included in the read mapping block, the snapshot mapping information of the identified snapshot number is compared with all other mapping information, and if all are different, the data block or extent allocated for the snapshot is deallocated. To do that,  Releasing the lock upon deletion of all mapping entries in the mapping block; Confirm that the corresponding mapping block is the last block, and complete the deletion of all mapping blocks in the mapping table through each of the above steps, and apply the snapshot deletion process for the mapping table to each logical disk device in sequence to all logical disks. Efficient snapshot method in a logical volume management system comprising the step of completing the deletion of the snapshot on the mapping table of the device.