KR20010011204A - Method and apparatus for overcoming raid failure by cache mirroring - Google Patents

Abstract

The present invention relates to a RAID system having a redundant RAID controller each having a cache for controlling a RAID disk array. The two redundant RAID controllers always compare the contents of each cache and match each other, and mirror each other's cache by deleting data in each cache when writing to the RAID disk array is completed. If one of the two redundant RAID controllers detects a failure in the active RAID controller or a failure in the connection between the RAID controller and the RAID disk array, the reserved RAID controller writes the data in its cache to the RAID disk array. This prevents data loss due to an active RAID controller failure. According to the present invention, the cache mirroring provides an excellent input / output performance by using IEEE 1394, thereby reducing the degradation of the host server.

Description

METHOD AND APPARATUS FOR OVERCOMING RAID FAILURE BY CACHE MIRRORING}

The present invention relates to a Redundant Array of Inexpensive Disk (RAID) system, in particular, in the event of a failure between a RAID controller and a RAID array, to ensure data consistency by storing data remaining in the cache of the RAID controller through a high speed serial communication protocol, The present invention relates to a RAID failover method and apparatus through cache mirroring.

The RAID system replaces one large and expensive drive by forming a RAID array by connecting many small and inexpensive drives. RAID systems treat a set of drives as if they were a single storage device, so that in the event of a failure, data is not lost and each drive can operate independently. Therefore, RAID system can reliably store a lot of data without any trouble.

The problems of the prior art RAID system will be described with reference to FIGS. 1 and 2.

1 is a configuration diagram of a RAID system according to the prior art. As shown, the RAID controller 100 is connected to the RAID array 200 to process the input and output requests of the host server 300, all data transmitted between the host server 300 and the input and output through the cache through the cache Improve the function

As described above, when connecting between the RAID array 200 and the host server 300 by using one RAID controller 100, if a failure occurs in the RAID controller 100 or its connection portion, the host server 300 may not use the RAID array 200. To this end, a standby RAID controller 110 is additionally installed, and a redundant structure is formed with the active RAID controller 100.

2 is a block diagram of a redundant RAID controller according to the prior art. As shown, the two RAID controllers 100 and 110 have the same structure, and when one RAID controller, for example, RAID controller 100 is activated, its cache 101 is converted into a local cache. The cache 111 of the other RAID controller 110 is used as a remote cache.

As described above, if the path is redundant between the RAID system and the host server, service downtime of the host server may be minimized in the event of a failure of the RAID controller or the connection part. In this case, however, data written to the cache of the active RAID controller by the command of the host server is lost. If so, the host server recognizes that the data has been written successfully, but the data actually written to the cache is not written to the RAID disk array, resulting in data inconsistency between the host server and the RAID disk array.

However, in order to prevent such data inconsistency, if the write-through mode of writing to the RAID disk array is used immediately without using a cache in the RAID controller, the data read / write speed may decrease.

That is, the conventional RAID system provides a means for preventing the data written to the buffer memory from being lost due to a sudden power failure, but a method for matching data between two caches of a redundant RAID controller in the event of a failure is provided. There was a problem that there was no bar.

Accordingly, an object of the present invention, which is designed to solve the problems of the prior art operating as described above, is to synchronize data using cache mirroring between two caches of a redundant RAID controller, thereby enabling the cache to be activated in the event of a failure. A method and apparatus for failover of a RAID through cache mirroring is provided so that data stored in a standby cache can be written to a RAID disk array even if data stored therein is lost.

1 is a block diagram of a RAID system according to the prior art.

2 is a block diagram of a redundant RAID controller according to the prior art.

3 is a configuration diagram of a RAID system according to the present invention.

4 is a flowchart of a RAID failover method according to the present invention;

5 is a configuration diagram of a redundant RAID controller for performing cache mirroring according to the present invention.

6 is a message flow diagram illustrating a communication procedure between a cache block and a serial interface block according to the present invention.

<Explanation of symbols for main parts of drawing>

100,110,400,410: RAID Controller

200,500: RAID disk array

300,600: host server

101,111,401,411: Cash

102,112,402,412: Cache Block

103,113,403,413: RAID Management Block

104,114,405,415: SCSI Interface Blocks

105,115,406,416: failover block

106,116,407,417: DMA block

107,117,408,418: Structural Blocks

108,118,409,419: Initialization block

404,414 serial interface block

1: IEEE1394 cable

Embodiment of the RAID failover method through the cache mirroring in accordance with the present invention, which is invented to achieve the above object, controls a RAID (Redundant Array of Inexpensive Disks) disk array according to the needs of the host server In a redundant RAID controller each having a cache for

The two redundant RAID controllers always compare the contents stored in the respective caches and match each other, and when writing to the RAID disk array is completed, deleting the data in the respective caches, thereby mirroring each other's cache (s10). )Wow;

Among the two redundant RAID controllers, if a failure is detected in an active RAID controller or a connection between the RAID controller and the RAID disk array is detected, the reserved RAID controller transfers the data in its cache to the RAID disk array. Recording step (s20, s30).

An embodiment of a RAID failover device through cache mirroring according to the present invention,

A host server 600;

RAID disk array 500;

A redundant RAID controller (400, 410) having a cache for controlling the RAID disk array at the request of the host server; And

An interface means (1) for connecting the redundant RAID and for cache mirroring to mutually replicate data between caches in the redundant RAID controller, wherein each RAID controller comprises:

RAID management blocks 403 and 413;

Caches 401 and 411 that match data with other RAID controllers through cache mirroring;

A cache block (402) (412) for controlling the cache by controlling the RAID management block; And

It is connected to the other RAID controller, and includes interface blocks 404, 414 to perform cache mirroring under the control of the cache block.

Two redundant RAID controllers according to the present invention always compare the contents stored in their respective caches and match each other, and when writing to the RAID disk array is completed, by deleting data in their caches, mirroring each other's cache ( Mirroring). To this end, the two redundant RAID controllers are connected to each other via a separate interface means, that is, a serial interface cable for cache mirroring.

When a failure is detected in an active RAID controller or a connection failure is detected, the reserved RAID controller immediately activates and writes data in its cache to the RAID disk array and then continues to control the RAID disk array.

Hereinafter, with reference to the accompanying drawings will be described in detail the operating principle of the preferred embodiment of the present invention.

3 is a block diagram of a RAID system according to the present invention, Figure 4 is a flowchart of a RAID failover method according to the present invention. As shown, in step s10, the redundant RAID controller 400, 410 connected to the RAID disk array 500 to process the input and output requests of the host server 600, each cache through a cache mirroring Match the contents in.

If the first RAID controller 400 is initially set to Active and the second RAID controller 410 is set to Standby, the data stored in the cache of the first RAID controller 400 is stored in the second RAID. The same is stored in the cache of the controller 410. When data stored in the cache of the first RAID controller 400 is recorded and deleted in the RAID disk array 500, data stored in the cache of the second RAID controller 400 is also deleted.

In operation S20, during the above-described cache mirroring operation in a normal state, an error occurs in the first RAID controller 400 or the connection between the first RAID controller 400 and the RAID disk array 500. In the case where an abnormality occurs in the site, in step S30, the first RAID controller 400 first sends a message to the second RAID controller 410 to store the contents of the cache. In addition, the host server 600 reports that the abnormality has occurred, so that the host server 600 can switch the operation to the second RAID controller 410.

When the second RAID controller 410 that is waiting to receive a message to switch to active from the host server 600, the second RAID controller 410 first stores the contents stored in the cache before performing the normal operation. Next, switch to the active state.

The second RAID controller 410 that is switched to the active state writes the contents stored in its cache to the RAID disk array 500. Therefore, the contents stored in the cache are not lost but are written to the RAID disk array.

In this case, the first RAID controller 400 and the second RAID controller 410 use a very fast serial interface in order not to degrade the disk read / write speed by continuing the cache mirroring.

Hereinafter, an operation of an embodiment of the present invention using IEEE 1394 (Institute of Electrical and Electronic Engineers 1394), which is a standard of a serial bus, as a data transmission medium for cache mirroring in a redundant RAID control period will be described.

5 is a block diagram of a redundant RAID controller for performing cache mirroring according to the present invention. As shown, in the first RAID controller 400 in an activated state, the cache block 402 controls the cache 401 under the control of the RAID management block 403, and the SCSI (Small Computers System Interface) interface block 405 controls the RAID disk array 500.

In this case, the contents changed in the cache 401 are transmitted to the cache 411 of the second RAID controller 410 in the standby state by the serial interface block 404 composed of the IEEE 1394 blocks in order to support the IEEE 1394.

That is, data transmitted between the caches 401 and 411 is transmitted through the serial cable 1 composed of an IEEE 1394 cable by the serial interface blocks 404 and 414 supporting IEEE 1384.

The cache block and the IEEE 1394 block implement cache mirroring using messages as shown in FIG. Hereinafter, a message transmitted between the cache block and the IEEE 1394 block will be described.

When the cache block of the active RAID controller sends SrCacheSetup (s1) to the IEEE1394 block of the active RAID controller to be connected with the spare RAID controller, the IEEE1394 block responds with SrCachePeerConnected (s2) and is connected with the IEEE1394 block of the spare RAID controller.

When new data is added to the cache by the host server, the active cache block sends SrCacheBlockCopyOut (s3) to the IEEE1394 block to copy the currently added cache data to the spare RAID controller.

If SrCacheBlockCopyIn (s4) is received from the active cache block, the IEEE1394 block requests the spare RAID controller to store the received cache data, and if the received cache data is stored, it responds with SrCacheBlockCopyDone (s5).

When cache data stored in the cache of the active RAID controller is written to the RAID disk array, the cache block sends SrCacheBlockInvalidate (s6) to the spare RAID controller through the IEEE1394 block. The spare controller then deletes the data stored in the cache, and the IEEE 1394 block responds with SrCacheBlockPurge (s7).

In addition, the cache block transmits SrCacheIsPeerAlive (s8) through the IEEE1394 block to check whether the spare RAID controller is normal. If the spare RAID controller is normal, the IEEE1394 block responds by transmitting SrCachePeerAlive (s9).

In the present invention operating as described in detail above, the effects obtained by the representative ones of the disclosed inventions will be briefly described as follows.

According to the present invention, reliability is improved by maintaining the consistency of data even in the event of a failure in a RAID system using a redundant RAID controller, and the effect of maintaining the consistency between the data of the host server and the data actually stored in the RAID disk array is effective. have.

In addition, IEEE1394, which provides the fastest speed among current serial communication methods and can be implemented as a simple interface, is selected as a data transmission medium for cache mirroring in a redundant RAID control period, providing excellent I / O performance and reducing the speed of the host server. Can be reduced.

Claims (6)

In the failover method of the redundant RAID controller which has a cache to control the Redundant Array of Inexpensive Disks (RAID) Disk Array according to the request of the host server, and transfers data to each other through the interface means. , Mirroring each other's stored data via the interface means between the caches in the two redundant RAID controllers; If a failure is detected in an operation of an active RAID controller among two redundant RAID controllers, the RAID disk is transferred to an inactive state, and data copied to the cache in the RAID controller transferred to the active state is transferred to the RAID disk. A method for failover of RAID through cache mirroring, comprising writing to an array. The method of claim 1, wherein the inactive RAID controller is switched to an active state to write data in its cache to a RAID disk array. When the failure is detected, sending the message to the inactive RAID controller to store the cache data to the inactive RAID controller; The active RAID controller reporting the detection of the failure to the host server; The host server instructing the inactive RAID controller to activate a switchover; Writing the contents stored in its cache to the RAID disk array after the inactive RAID controller is actively switched by the command; And And controlling the RAID disk array by the actively switched RAID controller. The method of claim 1 or 2, wherein the interface means is operated by an IEEE 1394 serial communication method. A host server; Redundant Array of Inexpensive Disks (RAID) disk arrays; A redundant RAID controller having a cache for controlling the RAID disk array at the request of the host server; And And means for interfacing the redundant RAID and cache mirroring to mutually replicate data between caches in the redundant RAID controller. 5. The apparatus of claim 4, wherein the interface means supports a data transmission scheme according to IEEE1394. The method of claim 4 or 5, wherein each of the RAID controller, A RAID management block; Cache matching data with other RAID controllers through cache mirroring; A cache block controlling the cache by controlling the RAID management block; And And an interface block connected to the other RAID controller through the interface means, the interface block performing the cache mirroring under the control of the cache block.