JP3177997B2 - Storage system - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an efficient data recovery method when a failure occurs in a storage device in a system including a plurality of storage devices, and in particular, stores data by dividing the data into the plurality of storage devices. The present invention relates to a failure recovery method that can be applied effectively in the case of a disk array system or the like.

[0002]

2. Description of the Related Art In a system including a plurality of storage devices, when data is divided and stored in a plurality of storage devices, the reliability of data is reduced by dividing and storing the data.

In the conventional system, redundant data (ECC data) for failure recovery covering all storage devices is added in order to prevent a decrease in reliability, and a failure occurs in any one of the storage devices. In such a case, the data in the storage device is recovered by the ECC data.

[0004]

In the above-mentioned prior art, a failure which can be recovered at the same time in order to enhance a failure recovery function against a decrease in data reliability due to an increase in the number of storage devices due to an increase in the storage capacity of the system. The redundancy of ECC data indicating the number of storage devices in the state has been increased. For this reason, the method of generating ECC data becomes complicated with the expansion of the system, and the EC
The load of C data generation has increased, the number of storage devices to be subjected to ECC data failure recovery has increased, and the amount of data accessed in the failure recovery processing has increased, thus increasing the load of failure recovery. In other words, in the conventional method, the expandability of the system is poor. In particular, when a plurality of storage devices are connected to each path, if multiple failures occur, the path performance becomes a bottleneck and the system performance is degraded. Was.

SUMMARY OF THE INVENTION It is an object of the present invention to solve the above-mentioned problems, and to improve the performance and expandability of failure recovery control in a system in which data is divided into a plurality of storage devices using a plurality of storage devices and stored. Is to improve.

Another object of the present invention is to achieve load distribution of an access load to a storage device and a path associated with a failure recovery process for a failure of a storage device in the above system.

[0007]

SUMMARY OF THE INVENTION In a system for storing data across a plurality of storage devices, the storage system includes a plurality of storage devices, a control device for controlling the storage devices, and a path connecting the control device and each storage device. Means for generating ECC data for failure recovery from a data set stored over a plurality of storage devices, storing the ECC data in another storage device, and storing a data set corresponding to the ECC data in an E storage device.
Means for collectively managing as a CC group, means for collectively accessing ECC groups, means for setting a plurality of ECC groups corresponding to mutually exclusive storage devices or subsets of paths, When the data in the storage device becomes temporarily or permanently inaccessible due to a failure or the like, means for inputting the remaining data of the ECC group to which the data belongs, and means for recovering the data from the input data Thus, the performance of the failure recovery process can be improved.

Further, in the above system, the means for changing the setting of the ECC group for each reference data amount makes it possible to distribute the load of the access load of the storage device and the path accompanying the failure recovery processing.

[0009]

In a failure recovery process associated with a failure that has occurred in an arbitrary storage device, an ECC group including data to be recovered is used to recover the storage device in the failed state from the data in the remaining storage devices excluding the failed storage device. Recover data. ECC
By setting a group for a mutually exclusive storage device or a subset of paths, the amount of data handled by each ECC group can be limited and the amount of ECC data can be reduced, and storage can be performed by increasing the storage capacity of the system. Even if the number of devices increases, a new ECC group is set as a subset of the increased storage device group, so that the amount of data to be accessed in accordance with the failure recovery processing can be kept constant, thereby increasing the load of the failure recovery processing. Without doing so, higher performance of the failure recovery processing can be achieved.

In addition, by changing the setting of the ECC group in advance for each reference data amount and setting the ECC group in advance, the ECC group can be fixedly set to a specific storage device or path, and the ECC group can be set in accordance with the failure recovery processing. Since the total amount of data to be accessed is the same and the number of storage devices or paths to be accessed can be increased, the amount of access data per storage device or one path associated with the failure recovery processing decreases, and The load distribution of the access load can be realized.

[0011]

Embodiments of the present invention will be described below in detail with reference to the drawings.

FIG. 3 shows a system configuration of this embodiment.
Eight drives 101 to 10 operable in parallel using a magnetic disk device (hereinafter referred to as a drive) as a storage device
8 and one spare drive 109 and the drive group 101
And a path 302 that connects the drives 101 to 109 to the disk device 301 and controls all the drives 101 to 109 at the same time. In this system configuration, since a drive and a path correspond one-to-one, selecting a drive and selecting a path have the same meaning. The data is the above 8 drives 1
Recording is performed by dividing the data into six drives 01 to 108, and ECC data is recorded in the remaining two drives. In FIG. 3, E
The CC group 310 includes drives B102, C103, and D
104, the data set 311 is recorded, and the drive A 101
ECC data 312 is recorded in the ECC group 320
Records the data set 313 in the drives F106, G107, and H108, and stores the ECC data 31 in the drive E105.
4 is recorded. Although FIG. 3 shows only one spare drive in order to avoid unnecessary complexity, it is easy to increase the number of spare drives in order to increase the reliability required for the system. is there.

FIG. 1 is a schematic diagram of a storage area of each drive, and vertical columns 101 to 109 show drives.
The horizontal rows 131 to 135 show the tracks of each drive.
It is defined that the drive A101 and the drive H108 are continuous on the drive sequence number, the ECC group is set to four drives continuous by the drive sequence number, and the setting for each drive is changed using the track as the reference data amount.
The case where the setting is performed by sliding one drive at a time is shown. For example, the ECC group 111 of the track 131 is set to the drives A101, B102, C103, and D104, and the next track 132 is slid by one drive, and
Drive CC group 113 to drive H108, A101, B10
2, set to C103.

FIG. 2 schematically shows the storage area of each drive, and has the same format as that of FIG. This shows a case where the ECC group is fixedly set to a specific four drives.
Drives A101, B102, C1
03, D104, indicating that the ECC group 212 has been set to the drives E105, F106, G107, and H108.

FIG. 4 schematically shows the storage area of each drive, and has the same format as that of FIG. The setting of the ECC group is changed in track units, and in selecting four drives necessary for setting the ECC group at the time of the change, the combination of the drives is realized as uniformly as possible, such as selecting each drive at random. Shows the case where it is selected. For example, the ECC group 411 of the track 131 is set to the drives A101, B102, C103, and D104, and the ECC group 413 of the track 132 is set to the drives A101, D104, E105, and F106.

FIG. 5 shows the movement of data and ECC data between a drive and a disk controller in a failure recovery process using an ECC group. Each data and E
The CC data moves according to the arrow.

FIG. 6 shows a flowchart of a failure recovery method using an ECC group.

FIG. 7 shows a system configuration when two drives are connected to each drive path. Drive 101-
The drives 701 to 709 corresponding to the path 109
2 and each drive can operate independently, but cannot use the same path at the same time.

Next, referring to FIGS.
4 describes a failure recovery process using an ECC group when a failure occurs.

The drive failure detection 720 is performed by the disk control device 301 by analyzing response information from the drive and monitoring the status. In the following, the drive D104
Assume that a drive failure has been detected. The disk control device 301 that has detected the drive failure is the drive D10.
In order to recover the data of No. 4, the failure recovery processing 700 is started. In the failure recovery processing 700, it is checked in the spare drive check 701 whether the spare drive J109 is available. As a result, in the condition determination 702, if the spare drive is not in the usable state, the failure recovery processing is terminated 707, and if the spare drive is in the usable state, the failure recovery processing is continued 708. Next, in the ECC group input 703, the EC
Data 312, 60 of drives A 101, B 102, C 103 excluding drive D 104 from C group 310
1, 602 are input to the disk control device 301. The data 603 of the drive D 104 is recovered by the data recovery 704 from the data set input to the disk control device 301. Next, at the recovery data output 705, the spare drive J10
9 to output data 603. In the subsequent condition determination 706, it is determined whether or not all the data in the drive D104 has been recovered.
09, ECC group input 703, data recovery 704,
The recovery data output 705 is repeated. If there is no data to be recovered, 710, the failure recovery processing ends. By the above processing, the data of the drive D104 can be recovered to the spare drive J109, and the drive D10 can be recovered.
By accessing the drive J109 instead of 4, it is possible to perform the same processing as before the failure occurred in the drive D104.

When a failure occurs, data recovery can be achieved by performing the above-described failure recovery processing regardless of the difference in the setting of the ECC group.

Next, an embodiment of a failure recovery method when an ECC group is fixedly set to a specific drive will be described in detail with reference to FIG.

An ECC group is fixedly set in a track group of the same address of the four drives, and ECC data of one track corresponding to data of three tracks is stored.
Since there are eight drives 101 to 108, two ECC groups 211 and
12 can be set.

Assuming that a failure has occurred in the drive D104, the failure recovery processing will be described. When some or all tracks of the drive D104 become inaccessible, the data of the drive D104 is recovered using the ECC group 211. First, in the recovery of the track 131, the drive A10
Tracks 131, B102 and C103 are input, and the track 131 of the drive D104 is recovered and stored in the track 131 of the drive J109 which is a spare drive.
Similarly, the recovery of the track 132 of the drive D104 is performed by inputting the track 132 of the drives A101, B102, and C103. Thereafter, the data of the drive D104 can be recovered by inputting the rest of the ECC group from the same drive A101, B102, and C103 for each failed track and performing the same processing.

According to the present embodiment, the drives to be accessed in accordance with the failure recovery processing are the drives A101, B102,
C103 is a total of three drives, and the other drive E1
05, F106, G107 and H108 have no effect. As described above, the method of fixedly setting the ECC group is as follows.
There is an advantage that the range of drives to be accessed in accordance with the failure recovery processing is limited, and drives in other areas can operate normally. On the other hand, since access is concentrated, if a plurality of faults occur simultaneously, there is a problem that the fault recovery time becomes longer.

Next, an embodiment of the failure recovery method in the case of FIG. 1 will be described in detail.

In FIG. 1, four consecutive drive sequence numbers are used.
A group of tracks at the same address of the drive is set as one ECC group, and one track corresponding to data of three tracks is set.
The ECC data of the track is stored. Drive 101-
Since there are eight 108s, two ECC groups can be set from tracks having the same address. Further, the setting of the ECC group is slid by one drive for each track.

Assuming that a failure has occurred in the drive D104, the failure recovery processing will be described. When some or all tracks of the drive D104 become inaccessible, the data of the drive D104 is recovered using the ECC group. For example, in recovery of the track 131 of the drive D104, EC
The C group 111 is used. Drive A101, B10
2, input the track 131 from C103 and drive D
The track 131 of 104 is recovered and stored in the track 131 of the drive J109 which is a spare drive. Similarly, the recovery of the track 132 of the drive D 104 uses the ECC group 114. Drive E105, F106, G
A track 132 is input from 107 and the same processing is performed. Hereinafter, the recovery of track 133 is ECC group 11
6, the drive C103, E105, and F106 are accessed, and the recovery of the track 134 is performed in the ECC group 1
18, the drive B102, C103, E105
, And the tracks 135 are recovered by using the ECC group 109 and the drives A101, B102, and C10.
3 and perform the same processing, all data in the drive D104 can be recovered.

According to this embodiment, in the setting of the ECC group in FIG. 1, a total of six drives are accessed in accordance with the failure processing, and as shown in FIG.
It is possible to access twice as many drives as compared to three, which is the number of access drives in the method of setting the C group. Moreover, since the total amount of data to be accessed is the same, load distribution of the loads on the drives A101, B102, C103 and the path 302 to each drive can be realized.

Hereinafter, an embodiment of the failure recovery method in the case of FIG. 4 will be described in detail.

In FIG. 4, eight drives 101-108
, All combinations for selecting four tracks from the same address are equally selected and set as an ECC group. For example, the ECC group 411 includes the tracks 13 of the drives A101, B102, C103, and D104.
1 and the ECC group 413 is the drive A1
Track 132 of 01, D104, E105, F106
Is a set of

Assuming that a failure has occurred in the drive D104, the failure recovery processing will be described. When some or all tracks of the drive D104 become inaccessible, the data of the drive D104 is recovered using the ECC group. The recovery of the track 131 uses the ECC group 411. Tracks 13 from drives A101, B102, C103
1, the track 131 of the drive D104 is recovered, and the track 1 of the drive J109, which is a spare drive, is recovered.
31. Hereinafter, regarding the track 132, E
Using the CC group 413, the drives A101, E1
05 and F106, and for the track 133, using the ECC group 416, the drive B10
2, G107 and H108, and access the track 134
Is accessed by using the ECC group 418 to the drives B102, E105, and F106, and the track 135 is accessed by using the ECC group 419 to access the drives A101, G107, and H108, and the same processing is performed. The data in the drive D104 can be recovered.

According to this embodiment, the drives to be accessed in response to the failure processing are all the drives except the drive D104, that is, a total of seven drives. The ECC of FIG.
More than twice the number of drives can be accessed as compared with the three methods in which the group is fixedly set, and the number of access drives can be increased as compared with the method of FIG. Since the total amount of data to be accessed is the same, the drives A101 and B1
02, C103 and the load of the path 302 to each drive can be distributed. On the other hand, since the storage area of each ECC group is fragmented, there is a problem that area management becomes complicated.

In the above three embodiments, the case where the reference data amount for changing the setting of the ECC group is performed by the track has been described. However, it is easy to adopt a sector or a cylinder as the reference data amount. In addition, when bits or bytes are used as the reference data amount, the efficiency is low when the magnetic disk device is used as a storage device as in the above-described embodiment, and the system is configured by a storage device accessed in units of bytes or bits. It is desirable to do.

In the above embodiment, one drive is assigned to each path.
Next, an embodiment of a system in which a plurality of drives are connected to each path will be described with reference to FIG. In FIG. 7, two drives are connected to each path, and drives 101 to 108 and drives 801 to 808 are connected.
And, the setting of the ECC group is shifted by one pass. Therefore, when a failure occurs in the drive D104 and the drive P804 at the same time, access for recovery from the failure of the drive D104 is performed by the paths 831, 833, and 833, and access for recovery of the failure of the drive P804 is to the paths 835, 826. 837.

According to this embodiment, in the case of a system in which a plurality of drives are connected to the same path, the ECC group is set to be shifted by a fixed amount with respect to the path, so that a path in the case where a failure occurs in a plurality of drives at the same time is avoided. Load can be distributed.

[0037]

As described above in detail, according to the present invention, a storage device such as a plurality of magnetic disk devices is provided.
A set of mutually exclusive storage devices in a system that stores data across multiple storage devices and performs a failure recovery process from the data of the remaining storage devices except for the failed storage device when a failure occurs in the storage device Against ECC
By setting the group, even if the number of storage devices increases due to an increase in the storage capacity of the system, it is possible to prevent an increase in the load of the failure recovery processing.

Further, according to the present invention, in the failure recovery processing in the above-described system, the access required for the failure recovery processing is achieved by changing the setting of the ECC group for the storage device or the path for each reference data amount and setting it. Since the operation can be performed for more storage devices and paths, the load can be distributed.

[Brief description of the drawings]

FIG. 1 shows a configuration in which an ECC group is set by shifting one drive at a time in track units.

FIG. 2 shows a configuration in which an ECC group is fixedly set to specific four drives.

FIG. 3 shows a system configuration of an embodiment.

FIG. 4 shows an ECC group of 4 out of 8 drives.
1 shows a configuration in a case where drives are equally selected and set.

FIG. 5 shows data movement in the case of performing recovery from a drive failure in a system configuration similar to that of FIG. 3;

FIG. 6 is a flowchart of a failure recovery process.

FIG. 7 is a system configuration diagram when a plurality of storage devices are connected to the same path.

[Explanation of symbols]

101 to 108: magnetic disk drive, 109: spare drive, 111 to 120: ECC group, 131 to
135 ... truck, 211, 212 ... ECC group,
301: Disk control device, 302: Path, 311, 3
13: Data set, 312, 314: ECC data, 3
10,320: ECC group, 411-420: ECC
Group, 601, 602, 603 ... data, 700 ...
Failure recovery processing, 701: spare drive check step, 702: spare drive check determination step, 70
3 ECC group input step 704 Data recovery step 705 Recovery data output step 706
Recovery end determination step, 707: processing route when there is insufficient spare drive, 708: processing route when there is a spare drive, 709: processing route when there is unrecovered data, 710 ... processing route when there is no unrecovered data , 7
20: drive failure detection step, 801 to 809: drive, 811, 813 ... data set, 812, 814
... ECC data, 810,820 ... ECC group, 8
31-839 ... pass.

──────────────────────────────────────────────────続 き Continuation of the front page (72) Inventor Hiroyuki Kitajima 1099 Ozenji Temple, Aso-ku, Kawasaki-shi, Kanagawa Prefecture Hitachi, Ltd. System Development Laboratory (56) References JP-A-2-157953 (JP, A) JP-A-62-293355 (JP, A) JP-A-2-194457 (JP, A) JP-A-2-291011 (JP, A) JP-A-2-176822 (JP, A) JP-A-3-505935 (JP) , A) (58) Field surveyed (Int.Cl. ⁷ , DB name) G06F 12/16 G06F 11/10 G06F 3/06

Claims (1)

(57) [Claims] 1. A group is formed by a plurality of storage areas existing in different storage devices, and at least one storage area in the one group includes one of the other storage areas in the one group. In a storage device system for storing parity data for reproducing lost data when data stored in the storage device is lost, at least one of a plurality of storage regions in a group to which the first storage region of the first storage device belongs. A storage system, wherein one is provided in a second storage device that does not include a storage region in a group to which a second storage region of the first storage device belongs.