KR20010028691A - RAID Level Y Using Disk Management Method - Google Patents

Abstract

PURPOSE: A method for managing a disk using RAID(Redundancy Array Independent Disk) level Y is provided to divide disks into a master channel and a slave channel, and to perform stripping by channel simultaneously and mirroring between the channels, thereby processing date by chunks of each disk. CONSTITUTION: SCSI(Small Computer System Interface) controllers(21,31) of the first and the second channels(20,30) manage disks by using a striping system, which sequentially records and reads chunks of each disk. A RAID(Redundancy Array Independent Disk) disk controller(10) generates virtual disks. The first and the second channels(20,30) manage disks by using a mirroring system, which designates a master channel to read data recorded in chunks of each disk in the corresponding channel. A RAID disk controller(10) generates virtual disks. In a corresponding RAID level Y, restoration data are distributed by chunks to enable prompt data processing.

Description

RAID Level Y Using Disk Management Method

The present invention relates to a disk management method using redundancy array independent disk (RAID) level Y. In particular, in managing each disk in a system including a RAID disk controller and a plurality of disks, striping of RAID level 0 The present invention relates to a disk management method using RAID level Y, which complements the mirroring method of the RAID level 1 and improves data processing speed and reliability for each disk.

Generally, RAID level is defined as a disk management method for creating a virtual hard disk or restoring a hard disk. Currently, RAID levels 0 to RAID level 5 are defined and used.

To illustrate this, the RAID level 0 enables fast I / O to the data, so it performs a striping method that distributes and manages the data over a large number of disks, while not allocating extra storage space in case of failure. If one disk fails, overall data loss can occur.

The RAID level 1 is a mirroring method in which all data written to one disk is copied and managed on another disk, so that the data can be recovered while the same data must be written to two disks. The processing speed is lowered.

RAID level 2 uses Hamming error correction codes for disks that do not have error detection capability. Disks managed by a small computer system interface (SCSI) controller perform error detection by the corresponding SCSI controller. Do not use levels.

The RAID level 3 stores parity information on one disk among multiple disks, and distributes data in bytes and manages the data on the remaining disks. need.

The RAID level 4 stores parity information on one drive among multiple disks, and distributes and manages the data in block units on the other disk, but the parity information enables data recovery in case of failure of a given drive. give. That is, a disk array of the corresponding RAID level 4 may have the same processing speed as that of the aforementioned RAID level 0 when reading data, but when writing the data, time for updating parity information is required every time. The processing speed was delayed.

The RAID level 5 is similar to the RAID level 4 described above, but since the parity information is recorded and managed on all disks, there is no bottleneck for the parity processing. Provides convenience when recording from time to time.

On the other hand, as in the conventional Voice Message System (VMS) system, a RAID level 1 mirroring method is used between the RAID disk controller and the disk. The RAID level 1 has a disadvantage in that the processing speed is slow. Because of this, we tried to use striping method with RAID level 0, but this had the disadvantage of losing data in the event of a disk failure.

To explain this in detail, the RAID level 0 is defined by the need for industrial development. The biggest problem is that if one disk fails, the data will be inaccessible and the data will be lost. That is, there is no reliability due to the lack of data resilience.

For example, RAID level 0, which includes four physical disks, may be illustrated in the accompanying drawings, as shown in FIG. 1, where a user's logical view of the data stored on each disk is referred to as "chunk," such as A-H. ) Into segments. That is, the first chunk A is physically located on the first disk, and the second chunk B is physically located on the second disk, but physically the same as where chunk A was located on the first disk. Similarly, the third chunk C is physically located on the third disk, physically the same as where it was located on chunks A and B, and the fourth chunk D is physically located on the same disk.

The remaining chunks (ie, chunks E, F, G, H, ...) are also distributed in turn to the next position of each disk, where the size of the chunks distributed to each disk is determined according to user requirements. That is, the user request in chunk B of FIG. 1 is a part of chunk B shown in the hatched area, and the access by the user request, that is, the hatched area of chunk B of the second disc when the recording or reading is made. Is approached.

In other words, the user's request to read the data array of each disk is accessed from where the data is located on a physically separate disk, where the data is read from that location and even if the data is written to the same location. do.

In other words, because user writes and reads are performed in one chunk, different user requests can be performed on each disk, and because they are performed simultaneously, processing speed can be improved, but reliability cannot be guaranteed. Recovery and access is not possible in the event of a disk failure.

In contrast, RAID level 1 arranges the chunks in the same order as the general disks as shown in FIG. 2, but writes the same chunks to the two disks for mirroring, thereby providing reliability. In other words, each of the chunks A, C, and E of the first disk is located at the same place of the second disc, and each of the chunks B, D, and F of the third disc is also located at the same place of the fourth disc.

At this time, the user request for reading data of each disk is accessed by reading the disk set as the master among the same two disks (master, slave), but the user request for writing data is set as the disk and slave set as the master. The disc is accessed and recorded at the same time.

Therefore, when one disk, that is, the disk set as the master, fails, the disk set as the slave can perform writing and reading on behalf of the disk set as the failed master.

However, the RAID level 1 requires twice as much storage capacity as the disk operation of the system, which also requires twice the cost.

As described above, the conventional RAID level 0 provides improved input / output function, but data is lost in the event of a disk failure, so it is not suitable as a storage device for data that is not reliable and cannot be easily reproduced, and thus ensures reliability. In the case of using parity information such as RAID levels 3 and 5, there is a disadvantage in that the processing speed is slower than using RAID level 1 due to the delay in processing speed.

The RAID level 1 is guaranteed for the data due to the mirroring of the disks designated as the master and the slave, but the data is processed at the same time as the RAID level 0 because the data must be written to the master and slave disks simultaneously. There was a disadvantage that the speed is lowered.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problems, and an object thereof is to separate a plurality of disks managed by a RAID disk controller and a SCSI controller into a master channel and a slave channel, respectively, and simultaneously perform stripping for each channel. By performing mirroring between channels to perform data processing for each chunk of each disk, fast data processing is possible, and data can be restored using the disk on the slave channel side in the event of a disk failure on the master channel side. To make it possible.

1 illustrates a conventional RAID level 0 including four disks.

Figure 2 illustrates a conventional RAID level 1 comprising four disks.

3 illustrates RAID level Y in accordance with the present invention with reference to a RAID disk controller.

4 is a diagram illustrating a disk management state using the RAID level Y in accordance with the present invention.

Explanation of symbols on the main parts of the drawings

10: RAID disk controller 20: first channel

21: first SCSI controller 30: second channel

31: second SCSI controller

A feature of the present invention for achieving the above object is a process for separating a plurality of disks managed by a RAID disk controller and a SCSI controller into a master channel and a slave channel; Performing data processing for each chunk of each disk by performing striping and mirroring according to RAID level Y according to a user's request, and performing data processing for each chunk of each disk includes: Accordingly, stripping is performed for each channel and mirroring is performed between the channels, and data is recorded and read based on the chunk of each disk.

Further, in the process of performing data processing for each chunk of each disk, if a disk failure occurs on the master channel side, the slave channel side further comprises reading and restoring data for each chunk of each disk. It is done.

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

In the present invention, in managing each disk in a system including a RAID disk controller and a disk, such as a VMS system, a new RAID level (hereinafter, referred to as' RAID level) that can ensure the processing speed of RAID level 0 and the reliability of RAID level 1 Y '), the RAID level Y performs a stripping method that processes data based on chunks for each channel, and a mirroring method that is divided into a master and a slave to improve reliability between each channel. Will be performed.

In order to explain this in detail, as shown in FIG. 3, three disks each having 8 gigabyte (Gbyte) capacity on the first channel 20 and the second channel 30 based on the RAID disk controller (10). Suppose it exists.

First, in the case of applying the conventional RAID level 0 between the RAID disk controller 10 and the disk, each disk writes and reads a certain amount of chunks equally whenever there is a user request based on the chunks. Chunks of data are processed at the same time, so processing speed is fast and 48 gigabytes (8 gigabytes * 6) are available in terms of disk capacity.

However, if the disk fails while the data is being written to or read from the disk, the data cannot be recovered, thereby reducing the reliability in terms of service to users.

Next, when the conventional RAID level 1 is applied between the RAID disk controller 10 and the disk, the same data is stored on the disks of the first channel 20 and the second channel 30, respectively. Although it can be very useful for a system requiring reliability, as the data must be simultaneously recorded in the first channel 20 and the second channel 30, the processing speed is reduced.

Meanwhile, in the case of performing disk management using the RAID level Y according to the present invention, as shown in the accompanying drawings, as shown in FIG. 4, in each channel 20 and 30, the RAID is controlled by the corresponding SCSI controllers 21 and 31. The disk is managed by stripping method of level 0, and the disk is managed by mirroring method of RAID level 1 between each channel (20, 30). At this time, the RAID disk controller 10 can create a virtual disk.

That is, the SCSI controllers 21 and 31 of the first and second channels 20 and 30 use a striping scheme for simultaneously writing and reading the chunks of the disks each having a predetermined size of chunks. In the case of (20), the SCSI controller 21 writes and reads the first chunk (chunk 0) of the first disk at the first identifier, and writes and reads the first chunk (chunk 1) of the second disk at the second identifier. In the third identifier, a sequential process of recording and reading the first chunk (chunk 2) of the third disk is performed.

Similarly, in the case of the second channel 30, the corresponding SCSI controller 31 writes and reads the first chunk (chunk 0) of the first disk at identifier 1, and the first chunk ( Chunk 1) is recorded and read, and a sequential process is performed to record and read the first chunk (chunk 2) of the third disc at the third identifier.

At this time, when reading the data recorded in the chunk of each disk in the first and second channels 20 and 30, the data reading process recorded in the chunk of each disk in the corresponding channel is performed by designating the master channel as in the mirroring method. RAID level Y does not store the restoration data in units of disks as in RAID level 1, but because the restoration data is distributed in chunks of each disk, it performs processing faster than RAID level 1 in terms of processing speed. You can do it.

When data is written to the chunk of each disk in the corresponding first and second channels 20 and 30, the disk is simultaneously written in the first and second channels 20 and 30 as in the RAID level 1. In terms of capacity, the effect cannot be expected, but since the RAID level 1 writes data on the disk corresponding to the disk identifier of each channel (20, 30), while recording data on any one disk, While the next data has to wait, processing speed slows down, while the corresponding RAID level Y simultaneously writes data for each chunk distributed to each disk in the first and second channels 20, 30, so the next data waits. The processing speed does not occur as it should.

As described above, the disk management operation using the RAID level Y according to the present invention is divided into a plurality of disks managed by the RAID disk controller 10 and the SCSI controllers 21 and 31 into a master channel and a slave channel. Then, if required by the user, stripping is performed for each channel (20, 30) according to the corresponding RAID level Y, and mirroring is performed between each channel (20, 30), and the corresponding RAID level Y is the level of RAID level 1. The difference with mirroring is that RAID level 1 stores the restore data on a disk basis, whereas RAID level Y stores the data on one disk because the restore data is distributed across all disk areas by chunk of each disk. Data processing is faster than processing speed.

When data is written to each chunk of each disk, the corresponding RAID level 1 writes data for each disk identifier of each channel (20, 30), whereas the data must be waited while recording data on one disk. Since the RAID level Y writes data for each chunk of all disks in each channel 20 and 30, there is no waiting time, thereby improving processing speed.

In addition, since the mirroring method is used between the channels 20 and 30, it is possible to solve the problem of data restoration in the event of a disk failure at RAID level 0. If you want to read the data stored in the disk on the master channel side If the disk fails while reading the corresponding data, the data is read from the disk on the slave channel side from that moment, so the reliability by data restoration can be secured. .

In addition, the embodiments according to the present invention are not limited to the above-described embodiments, and various alternatives, modifications, and changes can be made within the scope apparent to those skilled in the art.

As described above, the present invention divides a plurality of disks managed by a RAID disk controller and a SCSI controller into a master channel and a slave channel, respectively, and performs striping for each channel, and at the same time, performs mirroring between each channel, thereby By performing the data processing for each chunk, it becomes possible to process the data quickly and at the same time, it is possible to restore the data using the disk on the slave channel side in case of a disk failure on the master channel side, thereby ensuring reliability.

Claims (3)

Separating a plurality of disks managed by a RAID disk controller and a SCSI controller into a master channel and a slave channel; And performing data processing for each chunk of each disk by performing striping and mirroring according to RAID level Y according to a user's request. The method of claim 1, In the process of performing data processing for each chunk of each disk, mirroring is performed between each channel while stripping for each channel according to the RAID level Y, and recording and reading data based on the chunk of each disk. Disk management method using a raid level Wye, characterized in that. The method of claim 1, In the process of performing data processing for each chunk of each disk, if a disk failure occurs on the master channel side, the slave channel side further comprising reading and restoring data for each chunk of each disk, characterized in that it further comprises: How to manage disk using RAID level wire.