JPH04296926A - Recording medium controller - Google Patents

Abstract

PURPOSE:To improve the total performance of a RAID system of an array type disk device by adding a cache to only a checking use recording medium or to the recording medium with preference. CONSTITUTION:When an input/output request is given from a host computer 1, a host interface 2 received a command or the data to store them in a memory unit 4. A microprocessor 3 gives an instruction to a channel 5 to perform the input/output operations for the data and the parity of a disk device 8 or 9. Thus the channel 5 carries out this instruction. The RAID 2-4 have the concentrated accesses to a check disk 9 in a write state. In this respect, a cache is provided only to the channel 5 that is connected to the disk 9. Thus the access frequency is reduced. In such a way, the cache is provided to the disk 9 only so that the cache back-up operation is not needed at all.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an array type disk drive system in which storage mechanisms of input/output devices of a computer system are arranged in an array type.

0002

2. Description of the Related Art Recently, many documents and patents regarding array type disk systems have been published. It became generally known after the University of California, Berkeley (UCB) in the United States published a paper on array-type disks. UCB paper, ■A Cas
e for Redundant Arrays of
Inexpensive Disks (RAID) ■
, David A. Patterson, Garth
Gibson, and Randy H. Kats
, Report No. UCB/CSD87/39
1. In December 1987, array type disks were changed to RAID (Redundant Arrays).
of Inexpensive Disks) and is classified into five levels.

[0003] The first level is the conventional method of using mirror disks. The mirror disk is
Reliability is very high because another identical data disk is prepared. Furthermore, since the one with the shorter seek time is accessed when reading, it has the advantage that the access time is shorter than with a normal disk. However, since it is necessary to write to both the data disk and the check disk when writing, the response time becomes slow. Also, only 50% of the entire disk can be used as a data disk.

[0004] The second level uses Hamming codes, which are used in memories and the like. In the past, it was impossible to tell which part of the disk was malfunctioning. Therefore, a check disk was needed to determine which disk had failed. Data is interleaved for groups of several disks to reduce the number of check disks as much as possible. At the second level, the number of disks is 10
In contrast, 4 check disks are required, and 5 check disks are required for 25 disks.

[0005] The third level is that modern discs have EC.
C: A function has been added that allows you to know which disk is faulty, so one check disk is used. Then, the data disks are XORed and the contents are written to one check disk as parity information.

The fourth level requires access to all disks in a group in order to read/write one data at the third level, but the fourth level eliminates striping of data on disks in a group. Block interleaving allows multiple reads/writes to be processed simultaneously.

The fifth level is a method in which check data, which has been fixed on one disk, is distributed to data disks. At the fourth level, when multiple writes come in, parallel processing cannot be performed because the check disk is fixed. At the fifth level, check data is distributed and multiple writes can be processed in parallel.

FIG. 5 shows a conceptual diagram of RAID4. 28 to 31 indicate four data disks; 32
is the check disk. S0 to S3 below the disks 28 to 32 indicate logical sectors of the respective disks. 33 is sector 0 of the data disk 29, 34 is sector 0 of the check disk 32, 35 is sector 1 of the data disk 30, 3
6 is sector 1 of the check disk 32. As can be seen from the figure, in RAID4, check disks are concentrated on a single disk. In that respect, RAID1~RAI
Almost the same can be said about D3.

FIG. 6 shows a conceptual diagram of RAID5. 37 to 41 are disks in which data and checks are mixed, and S0 to S3 below disks 37 to 41 indicate logical sectors of the respective disks. 4
2 is data in sector 0 of the disk 38, 43 is check data in sector 0 of the disk 41, 44 is data in sector 1 of the disk 39, and 45 is check data in sector 1 of the disk 40. In this way, in RAID5, check data is distributed to each disk.

Next, the operation will be explained. In RAID4 in FIG. 5, when reading from sectors 33 and 35 simultaneously, data is simply read from each sector in parallel. When writing to the sectors 33 and 35, check data is written to the sectors 34 and 36 of the check disk 32 at the same time as data is written to the respective sectors. However, since the check disk 32 is one disk, it is not possible to write to sectors 34 and 36 at the same time. Therefore, when writing, the check disk tends to become a bottleneck.

In RAID5 in FIG. 6, sector 42
When simultaneously reading data from the sectors 44 and 44, data is simply read from each sector in parallel. When writing to sectors 42 and 44, data is written to the respective sectors and at the same time, check data is written to sectors 43 and 45 of check data belonging thereto. In this case, since the check data is distributed to each disk, it does not become a bottleneck for a particular disk.

The five levels of RAID have been shown above.
Other conventional examples include JP-A-62-192984.
There is a method and apparatus for storing digital words in which data and redundant data are separated and stored in separate storage media (MT in the embodiment), which is disclosed in Japanese Patent Laid-Open No. 62-293.
In the data protection mechanism of H.355, there is one that describes the technology indicated by RAID. Array Technology's array type disk drive system and method in Japanese Patent Application Laid-Open No. 2-236714 allows for flexible system configuration of RAID levels of RAID 1 to 5 and reliability levels such as one-device failure and two-device failure according to each user. It also describes the system in general. Array Technology's array type disk device H/W
Taking the configuration diagram as an example, the main part has a configuration as shown in FIG. 4.

In FIG. 4, 1 is a host computer, 2 is a host interface that is a control and buffer between the host computer and the disk controller, 3 is a microprocessor that controls the entire disk controller, and 4 is a host interface that is a control and buffer between the host computer and the disk controller.
is a memory unit that acts as a buffer for the disk controller; 5 is a channel that is a control and buffer between the disk controller and the data disks; 6 is a memory unit that acts as a buffer for the disk controller;
is a channel that is a control and buffer between the disk controller and the redundant disk, 7 is a system bus within the disk controller, 8 is a disk device in which data is stored, 9 is a disk device in which redundant code is stored, 10 is an array disk controller that controls array type disks.

In FIG. 4, when an input/output request is issued from host computer 1, host interface 2 receives commands and data and stores them in memory unit 4. Microprocessor 3 issues an instruction to channel 5 to input/output data and parity to and from disk device 8 and disk device 9, and channel 5 executes the instruction. Here, in RAIDs 2 to 4, data disks and check disks are completely separate and independent, but in RAID 5, check disks and data disks are not separated but are mixed together. Therefore, according to this H/W configuration, in RAID 5, the check data is distributed over multiple disks, so it does not become a bottleneck for the check disk, but in RAID 1 to 4, the check disk is accessed intensively when writing. This was a bottleneck for check disks.

[0015]

[Problem to be solved by the invention] The conventional method is as follows.
The array type disk drive was controlled at two levels. performance, disk space overhead, reliability,
From the viewpoint of control methods, RAID5 is the mainstream among RAID products. RAID5 is RA
Although it is superior in terms of performance compared to RAID4 or lower, there is a problem in that the control method is more complicated than RAID4 or lower.

[0016] This invention was made to solve the above-mentioned problems, and it provides a recording medium control system that can bring out performance equal to or better than RAID 5 despite the simple control method of RAID 1 to 4. The purpose is to obtain equipment.

[0017]

[Means for Solving the Problems] A recording medium control device according to a first aspect of the invention has the following elements. (a) a first channel means for connecting a data recording medium for recording data; (b) recording data regarding the reliability of data recorded on the data recording medium; (c) an interface means for connecting another system that attempts to access the recording medium; (d) a second channel means for connecting the recording medium for checking; (d) a second channel means for connecting the recording medium to the recording medium; a cache memory for temporary storage; (e) processor means for controlling the cache memory by allocating or preferentially allocating the cache memory for the recording medium connected by the second channel means;

Further, in the recording medium control device according to the second invention, when the recording medium connected by the second channel means fails, the processor means controls the operation of the cache memory allocated to the recording medium. The feature is that the control is stopped and other controls are continued.

[0019]

[Operation] The recording medium control device according to the first invention includes the first
A data recording medium (data disk) is connected to the second channel means, a check recording medium (redundant disk) is connected to the second channel means, and the recording medium connected by the second channel means. By adding cache only to (redundant disks) or preferentially adding cache, high-performance control becomes possible.

[0020]Furthermore, the processor means in the second invention is arranged such that when the recording medium (redundant disk) connected by the second channel means fails, the processor means operates the cache memory allocated to the recording medium (redundant disk). However, since it is a disk cache only for the recording medium (redundant disk) connected by the second channel means, there is no problem even if the operation is stopped, and in this point there is no need for a backup device at all. Therefore, a simple configuration or simple control is required.

[0021]

[Example] Example 1. An embodiment of the present invention will be described below with reference to the drawings. Figure 1
1 is a host computer, 2 is a host interface (interface means) that is a control and buffer between the host computer and the disk controller, 3 is a microprocessor (processor means) that controls the entire disk controller, and 4 is a disk. a memory unit acting as a buffer for the controller; 5 a channel (first channel means) which is a control and buffer between the disk controller and the data disk; 6 a control and buffer between the disk controller and the redundant disk; 7 is a system bus in the disk controller, 8 is a disk device in which data is stored, and 9 is a disk device in which redundant code is stored (hereinafter referred to as redundant disk or check 10 is an array disk controller (an example of a recording medium control device according to the present invention) that controls an array type disk, and 11 is a cache memory unit added only to the channel of the redundant disk.

Next, the operation will be explained. When an input/output request is issued from the host computer 1, the host interface 2 receives the command and data and processes the memory.
Store in unit 4. Microprocessor 3 issues an instruction to channel 5 to input/output data and parity to and from disk device 8 and disk device 9, and channel 5 executes the instruction. In RAIDs 2 to 4, the check disk is intensively accessed during writing, so a cache is provided only for the channel connected to the check disk to reduce the access load. In other words, taking RAID4 as an example, when writing, old data and old parity are read from the data disk and check disk into the array disk controller.
XOR calculation is performed on new data, old data, and old parity to generate new parity. When writing multiple data to update parity, such as during transaction processing, accesses to the check disk will inevitably be concentrated, so by providing a cache and accessing the cache, the access load on the check disk can be reduced. It can be done. By doing so, the total throughput can be improved.

[0023] Even if the cache becomes defective for some reason, this cache is only for the redundant disk, so there is no problem with the data disk, so the check that could not be updated. - The data is treated as invalid, and the check data is updated by XORing the data group of the data disk corresponding to the check data.

As described above, in this embodiment, in an array-type disk control device that exchanges data with a host computer, the array-type system includes a plurality of disk drives configured in an array-type, and the array-type disk control device The device includes a microprocessor for managing the array-type disk controller, a host interface that serves as a control and buffer between the host computer and the array-type disk controller, and a memory unit for use as a data buffer. A plurality of channels connected to the system bus are connected to a plurality of the disk drives, and perform input/output of a plurality of data to and from the plurality of disk drives, and the disk drive devices operate on the same plurality of sectors. It is divided into a data disk drive device and a check disk drive device, and a cache mechanism is added only to the redundant disk (check disk) in the disk drive mechanism inside the array type disk control device. The characteristic array type disk drive system has been explained.

Example 2. In order to flexibly respond to reliability levels, the array type disk control device can change the number of cache memory units in response to one failure, two or more failures. Since one check disk is required for reliability in the event of one failure, a disk cache is provided for only one check disk. Since two check disks are required to maintain reliability that can withstand two failures, a disk cache is provided for only the two check disks. The same applies to two or more units.

As described above, in the second embodiment, in order to flexibly respond to the reliability level of the array type disk control device, the reliability level that can withstand the failure of one mirror and the failure of two or more disks is set. An array type disk drive system has been described in which the array type disk controller is provided with a disk cache mechanism corresponding only to the number of redundant disks.

Example 3. The array type disk control device can change the number of cache memories in order to flexibly correspond to the RAID level. In FIG. 2, 9, 16, 17 are three redundant disks corresponding to the data disk 8, and 6, 12,
13 is a channel corresponding to each redundant disk, and 11, 12, 13 is a cache memory mechanism corresponding to each channel. . For the above RAID 1, three redundant disks 9, 16, 17 are required for three data disks 8, so three cache memory mechanisms 11, 1 are required for three redundant disks.
2 and 13 are provided. Also, for RAID2, 10
Since four check disks are required for one disk, four cache memory mechanisms are provided. R.A.
For ID4, one check disk is sufficient, so one cache memory mechanism is provided.

In the third embodiment described above, the array type disk controller has RAID1, RAID2, RAID3, RAID
An array-type disk drive system has been described which is compatible with a data distributed arrangement method called 4, and is characterized by providing a cache mechanism in the check disk.

Example 4. Arrayed disk drive systems are RAID 1-4.
In order to flexibly respond to the RAID level and the reliability level of mirroring, single failure, failure of two or more disks, in addition to the buffer, a cache for only redundant disks is provided in the memory unit in the array type disk control mechanism. . For example, in FIG. 3, 18 is a data disk;
19 and 20 are check disks, 21 is a buffer area, and 22 is a cache area. The cache area corresponds to each redundant disk. 23 is a cache area corresponding to the check disk 9, and 24 is a cache area corresponding to the check disk 16.
25 is a cache area corresponding to the check disk 17, 26 is a cache area corresponding to the check disk 18, and 2 is a cache area corresponding to the check disk 17.
7 is a cache area corresponding to the check disk 19, and 27 is a cache area corresponding to the check disk 20. Disc 8, Disc 9, Disc 16
, disk 17 is one reliability level, one redundancy group for determining the RAID level, disk 18,
Disk 19 and disk 20 are assumed to be one redundant group. If the former redundancy group is a redundancy group to support RAID level 1, and the latter redundancy group is a redundancy group to support a reliability level of up to 2 failures, then the cache area is Since the cache areas are set corresponding to the disks, there is no need to provide multiple cache mechanisms as shown in FIG. As described above, flexible support for RAID levels and reliability levels can be provided very easily.

As described above, in the fourth embodiment, a cache mechanism is integrated into a memory unit, the memory unit is used as a buffer and a cache, and the array type disk is characterized in that it dynamically corresponds to the RAID level and reliability level. The drive mechanism system was explained.

[0031] In this way, by attaching a plurality of cache mechanisms to corresponding redundant disks or by providing a cache area in a memory unit in an array type disk device, it is possible to flexibly deal with the RAID level and reliability level. There is a possible effect.

Example 5. If one or more redundant disks in a disk drive in an arrayed disk drive system partially or completely fail, the cache area in the cache mechanism or memory unit belonging to that disk drive is invalidated. . Even if the cache mechanism or cache area is disabled, there is no problem at all since the cache is for redundant disks. After disk replacement is performed by automatic replacement by hot standby of the check disk, the data group of the data disk corresponding to the check data is XORed to correct and restore the check data.

As described above, in the fifth embodiment, the array type disk drive system performs correction and restoration when a part or all of a certain disk fails, but if a disk drive other than the redundant disk fails, Normal correction and restoration is carried out using the generally known XOR, but if a redundant disk fails, the array type disk drive system is characterized by stopping the disk cache mechanism including the redundant disk. explained.

In the first to fifth embodiments described above, in an array type disk drive system, a cache is added only to the check recording medium (redundant disk), but the cache is also added to the data recording medium. In the case of adding a cache, it is sufficient to adopt a configuration in which a cache is preferentially added to the check recording medium (redundant disk). for example,
In this case, preferential addition includes increasing the cache capacity of the checking recording medium (redundant disk) compared to others, or making the cache of the checking recording medium (redundant disk) faster than others. Also,
In the first to fifth embodiments described above, an array-type disk drive mechanism system has been described as an example, but the present invention is not limited to an array-type system, and any system having a plurality of disk drive mechanisms may be used. Further, the control device is not limited to a disk drive mechanism, and may be any control device that controls a recording medium such as an optical disk or a CD.

[0035]

As described above, according to the first invention, since the configuration is such that cache is added only to the check recording medium (redundant disk) or preferentially to the check recording medium (redundant disk), The device can be made inexpensively, and
It has the effect of increasing performance to the same level or higher than RAID5.

Additionally, according to the second invention, since a cache is provided on the redundant disk, even if the redundant disk or its cache fails, there will be no problem with the data disk, so backup can be performed. Even if you don't use it,
This has the effect of obtaining sufficient reliability.

[Brief explanation of the drawing]

FIG. 1 is a sectional view showing an array type disk control device according to an embodiment of the present invention.

FIG. 2 is a diagram showing a cache mechanism for accommodating RAID levels and reliability levels.

FIG. 3 is a diagram showing a cache mechanism for flexibly responding to RAID levels and reliability levels.

FIG. 4 is a sectional view showing a conventional array type disk control device.

FIG. 5 is a data configuration diagram of RAID4.

FIG. 6 is a data configuration diagram of RAID5.

[Explanation of symbols]

1 Host computer 2 Host interface 3 Microprocessor 4 Memory unit 5 Data disk channel 6 Check disk channel 7 System bus 8 Data disk 9 Check disk 10 Array type disk controller 11 Cache memory unit 12 Check・Disk channel 13 Check disk channel 14 Cache memory unit 15 Cache memory unit 16 Check disk 17 Check disk 18 Data disk 19 Check disk 20 Check disk 21 Buffer area 22 Cache areas 23 to 27 Redundant disk Cache area 28-3
1 Data disk 32 Check disk 33 to 36 Sectors 37 to 41 Disk 4 where data and checks are mixed
2-45 sectors

Claims (2)

[Claims] Claim 1: A recording medium control device having the following elements: (a) first channel means for connecting a data recording medium on which data is recorded; (b) a first channel means for connecting a data recording medium on which data is recorded; (c) a second channel means for connecting a checking recording medium for recording data regarding the reliability of the recorded data; (c) an interface means for connecting another system that attempts to access the recording medium; (d) Cache memory that temporarily stores data transferred to and from the recording medium; (e)
Processor means for controlling the cache memory by allocating or preferentially allocating the cache memory for the recording medium connected by the second channel means. 2. The recording medium control device according to claim 1, wherein when the recording medium connected by the second channel means fails, the processor means stops the operation of the cache memory allocated to the recording medium. 2. The recording medium control device according to claim 1, wherein the recording medium control device continues to perform other controls.