JP3422004B2 - Expansion type disk array device - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention has a plurality of disk drives, and even if any one of the drives fails, the disk drive is configured based on data recorded in another drive. The present invention relates to an improvement in a disk array device capable of restoring data. 2. Description of the Related Art FIG. 1 shows a basic configuration of a disk array device. A plurality of, here, N (N ≧ 2 integer) disk drives 1 (where N disk drives are distinguished from each other) , # 2,..., #N), a disk array controller 2 for controlling the disk drives # 1 to #N, and a disk drive #,
The disk array device 4 is composed of the internal bus 3 connecting the disk array controllers 1 to #N and the disk array controller 2. The disk array controller 2 is connected to the host computer 6 via the external bus 5 and accesses each of the disk drives # 1 to #N according to a read / write command from the host computer 6. In order to enhance the reliability of the disk array device, a redundant configuration is adopted so that even if one of the disk drives fails, the data recorded by the failed drive can be restored. The simplest configuration among the redundant configurations is to provide a parity disk. For example, exclusive OR operation is performed on each data of the same byte written in the same physical sector of each disk drive, and a parity is calculated based on the calculation result. This is to generate and record data. In the configuration shown in FIG. 1, when the host computer 6 issues a data write command to the disk array controller 2, the disk array controller 2 transmits the data transferred from the host computer 6 to the disk array controller 2, for example, as shown in FIG. As shown in (1), for example, even-number parity data is generated for the divided data to generate a total of five divided data. The data generated in this manner is divided into five disk drives 1 and recorded. [0005] When the recorded data is read, usually, four divided data excluding parity data are read from four disk drives 1, combined and transferred to the host computer 6. When one disk drive fails, the divided data and parity data are read from the four failed disk drives as shown in FIG. 3, and an even parity operation is performed on the data. Then, the divided data that cannot be reproduced due to the failure of the disk drive is restored, the restored data is written to a new disk drive, the written data is read again, and the data is combined and transferred to the host computer 6. Usually, a data error can be detected from parity data, but data cannot be restored. On the other hand, in the disk array device, a failed disk drive can be specified, that is, an error position on data can be specified, so that data can be restored. As described above, when a disk drive fails, the disk array controller uses the parity data recorded in another disk drive to perform a process of restoring the divided data that cannot be reproduced due to the failure. The parity data is recorded on a disk drive dedicated to parity or divided and recorded on each disk drive.
Since the method of recording data on a disk drive dedicated to parity lowers the access speed, a method of recording divided data is usually adopted. When a disk drive fails, the disk array controller restores the data of the failed disk drive on the new disk drive by using parity data. By the way, the disk array device has a plurality of disk drives, so data security against failure is high, and since each disk drive can transfer data in parallel, the data transfer speed is high. Since a plurality of disk drives need to be installed in advance, there is a disadvantage that the initial investment of the system is large. In particular, when the storage capacity is large and the data storage period when the system is introduced is long, such as a medical image data storage system that requires storage for at least five years, the system scale is gradually increased. It was desired. However, in the conventional disk array device, the initially set data division method cannot be changed in the middle, so that the maximum number of disk drives expected at the time of system construction is required. It is necessary to install the system, and there is a problem that the initial investment of the system becomes extremely large, and there is a problem that it is not possible to cope with a case where a storage capacity exceeding the initial expectation is required due to expansion of business or the like. An object of the present invention is to provide an expansion type disk array device which can start with a minimum storage capacity at the time of system introduction, requires a small initial investment, and can arbitrarily increase the storage capacity as the system is expanded. To provide. According to the present invention, in order to solve the above-mentioned problems, N (N ≧ 2 integer) disk drives for recording and reproducing data, and data such as bits, bytes, words, etc. Divided data generation unit that divides the divided data group into a predetermined number, a redundant data generation unit that generates redundant data for the divided data group, and restores the divided data that cannot be reproduced due to a disk drive failure using the redundant data A disk array controller having a divided data management table for recording an arrangement relationship between each disk drive and the divided data and the redundant data, wherein the disk array controller stores the data when recording the data. Is divided into a predetermined number of divisions M (an integer of M ≧ 2) by a divided data generation unit, and the redundant data generation unit In addition to generating redundant data for the divided data group, the divided data and the redundant data are recorded in a distributed manner on a plurality of disk drives, and the arrangement relationship between the recorded divided data and the redundant data and the disk drives is represented by the divided data. When the data is recorded in the management table and the data is reproduced, the required divided data is reproduced based on the divided data management table, and the divided files are combined to reproduce the original file. When the divided data cannot be reproduced by the data restoration unit, the data restoration unit reproduces the redundant data, and restores the divided data that could not be reproduced due to a failure from the redundant data to reproduce the original data . This
(N-1) is 1 or divided for the number N of drive
When there is a divisor relationship of number M, M divided data
(N-1) group and (N-1) group
Redundant data for the divided data group
To generate the divided data and the redundant data of the (N-1) group.
Data can be arbitrarily distributed and recorded on N disk drives
If (N-1) is 2 or more and there is no relationship with the divisor of M
The largest divisor of M that does not exceed (N-1)
Extract number K and divide M divided data into K groups
Together with the divided data group divided into K groups.
Generated redundant data, and the divided data and
(K + 1) disk drives and redundant data
We propose an extended disk array device that records in a distributed manner. According to the above configuration, when the system is introduced, 2
A RAID-1 level disk array device composed of two disk drives (one that writes the same data to two disk drives by mirroring and can continue processing even if one of the disks fails) RAID-3 level (can be configured as one disk drive by having one parity disk drive for n data disk drives) This enables processing to be continued even if a failure occurs. Striping is performed in units of 1 bit or 1 byte.) Or RAID-5 level (n data and one disk drive for n + 1 disk drives) Rather than fixedly assigning parity, disperse parity equally to enable simultaneous rewriting of two or more locations Thing.) Because it can configure the disk array device can be started at the required minimum storage capacity at the time of system installation, requires less initial investment, moreover it can arbitrarily increase the storage capacity in accordance with the expansion of the system, further
If the process of expanding the disk drive becomes easier,
Finally, there is no danger of file destruction . FIG. 4 shows an example of an embodiment of an extended disk array device according to the present invention. In this example, the number of file divisions M is 4, and the number of disk drives N is 2 (minimum). Constitution)
Here is an example of the case. In the figure, 1 is a disk drive (however, when discriminating N disk drives,
# 1, # 2,... #N. ), 3 an internal bus, 7 a disk array controller for controlling each disk drive, each disk drive and the disk array controller 7 are connected by an internal bus, and these constitute a disk array device 8. The disk array controller 7 is connected to the host computer 6 via the external bus 5 and accesses each disk drive in response to a read / write command from the host computer 6. In the case of this example, (N-1) =
Since it is 1, the data divided into four are grouped into one, and the grouped divided data and the redundant data generated for the group are recorded on two disk drives # 1 and # 2. Is done. FIG. 5 shows the details of the disk array controller. In FIG. 5, reference numeral 71 denotes a divided data generator for dividing the data transferred from the host computer 6; A redundant data generating unit 73 for generating parity data by using a divided data arrangement management table 73 for recording the arrangement relationship between each disk drive and the four divided data and the redundant data. A data restoration unit for restoring the divided data that cannot be reproduced due to the failure of the disk drive; 75, an internal bus interface for accessing the disk drive;
Reference numeral 76 denotes an external bus interface for connecting to a host computer. Also in the disk array controller 7 of the present invention, the recording and reproduction of data as described with reference to FIG. When any one of the disk drives fails, the data is restored in the same manner as in FIG. 3, but conventionally, the divided data is recorded separately on different disk drives. On the other hand, the present invention is different in that the divided data is grouped according to the number of disk drives and recorded in group units. Specifically, in the configuration of FIG. 4, when the host computer 6 issues a file A write command to the disk array controller 7, the divided data generation unit 71 of the disk array controller 7 is transferred from the host computer 6. Although the file A is divided into four as shown in FIG. 4, in a configuration having two disk drives, the four divided data are grouped into one and all are recorded in the disk drive # 1. At the same time, parity data for the divided data of the file A divided into four by the redundant data generation unit 72 of the disk array controller 7,
, Are generated and recorded on the disk drive # 2. Thereafter, the divided data arrangement management table 73 of the disk array controller 7 records on which disk drive the divided data of the file A is arranged. As described above, when the number of disk drives is two, the disk array controller 7
However, since the four divided data are treated as one group, they are not distributed and recorded on a plurality of disk drives. In this case, in the disk drive # 2, in the case of the even parity system, the same data as the file A recorded in the disk drive # 1 is recorded. The same data is recorded. That is, if the inversion and non-inversion of the bit are ignored, the same data as the file A recorded on the disk drive # 1 is recorded in any case, and a mirror disk corresponding to the RAID-1 level is configured. When reproducing a file, the data divided into four parts is reproduced from, for example, a disk drive # 1, combined by a disk array controller 7 and transferred to a host computer 6. If a failure occurs in the disk drive # 1 and the divided data of the file A cannot be read,
The divided data that could not be read by the data restoration unit 74 of the disk array controller 7 is stored in the disk drive #
2 and are combined on the disk array controller 7. Next, when the number of disk drives of the disk array device 8 is expanded to three as shown in FIG.
4, (N-1) = 2, and since this is a divisor of M, the data divided into four are grouped into two, and the divided data and The redundant data generated for the group is recorded on three disk drives # 1 to # 3. As described above, when the number of disk drives is expanded to three, with respect to files recorded before the number of disk drives is expanded to three, divided data grouped into one is divided into two groups. In addition, it is necessary to generate new redundant data for the divided data grouped into two. In this case, according to the present invention, first, the parity data recorded on the disk drive # 2 is rewritten with a new 1 /
The parity data having a capacity of 2 is generated and recorded in the disk drive # 3. Specifically, as shown in FIG. 6, of the divided data of file A, 1 /
4, 2/4, and the remaining 3 in disk drive # 2.
Assuming that / 4 and 4/4 are grouped and recorded, new parity data is generated from parity data corresponding to 1/4 and 3/4 of the divided data, and further divided into 2/4 and 4/4 of the divided data. New parity data is generated from the corresponding parity data, and the two parity data are grouped as new parity data;
Record on disk drive # 3. Thereafter, 3/4 and 4/4 of the divided data are deleted from the disk drive # 1, and the disk drive # 1 is deleted.
If 1/4 and 2/4 of the divided data are deleted from 2, 3
It is possible to expand files corresponding to one disk drive. Thereafter, the divided data arrangement management table 73 is rewritten. As described above, in the disk array device composed of three disk drives, the file A
Are recorded in a distributed manner, so that a RAID-3 level device is configured. At this time, it is easy to realize a configuration corresponding to a RAID-5 level device so as not to lower the access speed. In this case, for example, for some files recorded in the disk array device, new parity data generated from the parity data recorded in the disk drive # 2 is replaced by the disk drive # 2.
2, and the divided data 3/4 and 4/4 remaining in the disk drive # 1 are transferred to the disk drive # 3. Alternatively, parity data is newly generated from the divided data recorded in the disk drive # 1 and recorded in the disk drive # 1, and the parity data recorded in the disk drive # 2 (in the case of even parity, the parity data is divided). Data itself) to divided data 1 /
4, 2/4, 3/4, 4/4, and 3 /
The divided data corresponding to 4 and 4/4 is transferred to the disk drive # 3. Thereafter, the unnecessary divided data is deleted. As a result, the disk drive where the parity data is recorded is arranged differently for each file,
A RAID-5 level device is configured. Next, when the number of disk drives of the disk array device 8 is expanded to four as shown in FIG.
For (4), (N-1) = 3. In this case, since the relationship of the divisor of M is not satisfied, the largest divisor not exceeding 3 from the divisor of M, that is, 2, is selected. While the divided data is still grouped into two, the grouped divided data and the redundant data generated for the group are recorded on any three of the four disk drives Is done. At this time, the disk array controller 7
In order to prevent the access speed from being reduced, divided data and parity data are moved between disk drives, and operations are performed so that the used storage capacities of the four disk drives become substantially equal. In the example of FIG. 7, the parity data recorded on the disk drive # 3 is transferred to the newly added disk drive # 4, and the unnecessary parity data is deleted. Thereafter, the divided data arrangement management table 73 is rewritten. Further, when the number of disk drives of the disk array device 8 is expanded to five as shown in FIG.
= 4 and (N-1) = 4, and since this is a divisor of M, the data divided into four are grouped into four, and the divided data And the redundant data generated for the group are recorded on five disk drives. As described above, when the number of disk drives is expanded to five, for a file recorded before the number of disk drives is expanded to five, divided data grouped into two is divided into four groups. In addition to the division, it is necessary to generate new redundant data for the divided data grouped into four. In this case, as shown in FIG. 8, of the divided data of file A, 1/4 is recorded on disk drive # 1, 3/4 on disk drive # 2, 2/4 on disk drive # 3, Assuming that 4/4 is recorded in the disk drive # 4, first, in FIG.
4 from the parity data recorded in
Is generated and recorded in the added disk drive # 5. Next, the 1/4 and 2/4 of the divided data are transferred from the disk drives # 1 to 2/4 to the disk drive # 3. Delete the split data. Also, the disk drive # 2, which records 3/4 and 4/4 of the divided data, transmits the divided data of 4/4 to the disk drive # 4.
And deletes 4/4 divided data from disk drive # 2. Thereafter, the divided data arrangement management table 7
3 is rewritten. By doing so, it is possible to expand files corresponding to five disk drives. When the number of disk drives becomes six or more, (N−1) ≧ 5 for M = 4, and M
Does not satisfy the divisor relationship. Therefore, the largest divisor that does not exceed 5 from the divisor of M, that is, 4, is selected, and the data divided into four are grouped into four, and the grouped division is performed in the same manner as in FIG. The data and the redundant data generated for the group are recorded on any five of the six or more disk drives. As described above, according to the present invention, the file to be recorded in the disk array device is divided into an appropriate number in advance. Read, combine, subdivide according to the number of expanded disk drives, generate new parity data for the subdivided data,
The expansion process is completed only by transferring the divided data and performing the degeneration operation of the parity data, while the data is again recorded on the disk drive. At this time, the amount of the file to be transferred is almost half of all the files, which is extremely small. Furthermore, in most cases, the degeneration operation of parity data is
The processing ends with the parity operation on the parity data recorded on the same disk drive. For this reason, the processing involved in expanding the disk drive is extremely simple and can be executed at high speed, and the risk of destroying already recorded files is extremely low. Since the number of file divisions is fixed in advance, not only can the file division processing be implemented by firmware or the like to speed up the division processing, but also if the number of disk drives is increased. Also, there is no need to reconstruct the divided data arrangement management table. Therefore, when expanding the number of disk drives, there is no danger that the divided data arrangement management table is destroyed and the file cannot be restored. As described above, according to the present invention, even when the number of disk drives of the disk array device is expanded, a part of the divided data is transferred without rewriting the divided data arrangement management table. However, since expansion processing can be completed only by performing degeneration operation of parity data,
When the disk array device is expanded, there is a very low risk of occurrence of a failure such as a failure to restore a file, and the expansion process can be executed at high speed.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a basic configuration diagram of a disk array device. FIG. 2 is an explanatory diagram showing a normal read / write process. FIG. 3 is an explanatory diagram showing a read / write process when a failure occurs. FIG. 4 is a disk array device (two drives) of the present invention.
FIG. 5 is a configuration diagram showing details of a disk array controller. FIG. 6 is a disk array device (three drives) of the present invention.
FIG. 7 is a disk array device of the present invention (four drives)
FIG. 8 shows a disk array device (five drives) of the present invention.
[Description of reference numerals] 1 ... disk drives (# 1 to #N), 3 ... internal bus,
5 external bus, 6 host computer, 7 disk array controller, 8 disk array device, 71
Divided data generation unit, 72: redundant data generation unit, 73: divided data arrangement management table, 74: data restoration unit, 75
... internal bus interface, 76 ... external bus interface.

──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Nobuyoshi Izawa 3-19-2 Nishi Shinjuku, Shinjuku-ku, Tokyo Nippon Telegraph and Telephone Corporation (72) Inventor Kosuke Sakurai 3-19 Nishi Shinjuku, Shinjuku-ku, Tokyo No. 2 Nippon Telegraph and Telephone Corporation (72) Inventor Takashi Matsumoto 3-19-2 Nishi-Shinjuku, Shinjuku-ku, Tokyo Nippon Telegraph and Telephone Corporation (72) Inventor Keisho Shiramizu Naka-ku, Yokohama-ku, Kanagawa Prefecture No. 2-9-1, NTT-Intelligent Technology Co., Ltd. (72) Inventor Takashi Kono 2-9-1, Furo-cho, Naka-ku, Yokohama-shi, Kanagawa Prefecture NTT-Intelligent (56) References JP-A-7-141121 (JP, A) JP-A-8-147111 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) G 06F 3/06-3/08 G06F 12/00-12/12 G11B 20/10-20/16

Claims (1)

(57) [Claims] [Claim 1] N (N ≧ 2 integer) disk drives for recording and reproducing data, and dividing data into a predetermined number in units of bits, bytes, words, etc. A divided data generation unit, a redundant data generation unit that generates redundant data for the divided data group, a data restoration unit that restores the divided data that cannot be reproduced due to a failure of the disk drive using the redundant data, A disk array controller having a divided data management table for recording the arrangement relationship between the divided data and the redundant data.
The data is divided by a division data generation unit into a division number M
(An integer of M ≧ 2), a redundant data generation unit generates redundant data for the divided data group,
The divided data and the redundant data are recorded in a distributed manner on a plurality of disk drives, the arrangement relationship between the recorded divided data and the redundant data and the disk drives is recorded in a divided data management table, and the data is reproduced. At the time of reproduction, the required divided data is reproduced based on the divided data management table, the divided data is combined to reproduce the original file, and if the divided data cannot be reproduced due to a disk drive failure, the data is restored. It reproduces the redundant data by part, in the expanded form a disk array apparatus for reproducing original data to repair the divided data could not be reproduced due to a failure from the redundant data to the disk drive number N, (N- If 1) is 1 or a divisor of the number of divisions M,
In this case, the M pieces of divided data are divided into (N-1) groups.
And the divided data divided into (N-1) groups
Generating redundant data for a group, the (N-1) group
Drive divided data and redundant data on N disk drives
When (N-1) is 2 or more and there is no divisor of M
Is the largest divisor of M that does not exceed (N-1)
Extract K and divide M data into K groups
Both are for the divided data group divided into K groups.
Generating redundant data, the divided data of the K group and
Redundant data to any (K + 1) stand disk drives
An extended disk array device characterized by recording in a distributed manner.