JPH06139027A - Disk driver, disk array device, data storage system, and data backup method for disk array system - Google Patents

Abstract

(57) [Summary] [Purpose] To reduce the backup occupation time of data in the disk array system. [Structure] A data / parity disk drive 5 for storing data and parity information is connected to the disk array controller 2 via drives I / F 2-6-1, 2-6-2, ... In an array. A predetermined number of spare disk drives 4 are connected to the drive I / F 2-6-3 of the disk array controller 2.
When the data / parity disk drive 5 fails, the spare disk drive 4 becomes a data / parity disk drive in place of the failed data / parity disk drive. 3-1, 3-2 or 3
-3 also serves as a data transfer buffer between the data / parity disk drive 5 and the backup device when transferring data for backup.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a computer system provided with a predetermined number of data storage devices such as disk drives, and more particularly to backing up data stored in the storage devices.

[0002]

2. Description of the Related Art A disk array device is disclosed in, for example, Japanese Patent Laid-Open No. 1-250128. The feature of this device is that it achieves high speed by simultaneous multiple operation of a plurality of disk drives and high reliability by having a redundant disk drive and a spare disk drive.

In order to speed up the disk drive,
It is necessary to increase the data transfer speed and improve the number of read / write processing (IO throughput) per second, but in the disk array device, data transfer is performed in parallel for multiple disk drives. By making it possible, the data transfer speed has been increased. A configuration in which redundancy described later is added to such a configuration to increase reliability is generally called RAID3. Also,
There is a configuration in which multiple disk drives are made to wait for multiple seek / rotation in order to improve IO throughput, and a configuration in which redundancy is added to this configuration to increase reliability is called RAID5. There is.

In order to improve the above performance, a plurality of disk drives are required in any of the above configurations. The current standard failure time (MTBF) of a disk drive is about 50,000 hours, but the average failure time (MTBF) of a disk array device using N disk drives is 50,000 /
It can be suppressed to N hours. However, this value is 10
In the case of a disk array device with a table configuration, it takes 5,000 hours, that is, less than seven months, which is generally an unacceptable value for a system.

To solve this problem, the disk drives of the disk array device are divided into groups consisting of several data disk drives and one parity disk drive, and the data is divided into data units called stripes. The parity information is calculated by forming a group called a parity group with all stripes starting from the same logical address in all disk drives in the above group and calculating the exclusive OR of all data stripes in this parity group. Is known, and is stored in the parity disk drive of the parity group as a parity stripe. Such a configuration is called RAID, and not only is data distributed to a plurality of disk drives, but even if any one of the parity groups fails, the parity information stored in the parity disk drive is used. The data of the failed disk drive can be recovered, and the reliability is improved.

By the way, the mean failure time of RAID (MT
BF) is expressed by the following equation 1.

[0007]

[Equation 1]

According to this equation 1, the time required to replace a failed disk drive with a normal disk drive and recover data to this disk drive, that is, the average repair time (MTTR) of the disk drive is the average failure time (MTTR).
It turns out that it is an important factor in determining BF). In order to minimize this mean time to repair (MTTR), a spare disk drive that replaces the failed disk drive is provided in the parity group in advance, and when a failure occurs, the spare disk drive is used to promptly use the data. It is effective to carry out recovery. Such a spare disk drive is also disclosed in Japanese Patent Laid-Open No. 1-250128 as a spare disk drive.

By adopting the RAID configuration and the spare disk drive, the problem of a decrease in mean time to failure (MTBF) in a disk array device composed of a plurality of disk drives can be solved.

[0010]

However, the mean time to failure (MTBF) expressed by the above equation 1 is based on the assumption that the occurrence of failures of a plurality of disk drives in the system is independent of each other. In the case of a large-scale disaster such as an earthquake or fire, or when a double failure occurs due to an error during failure recovery work, etc. are not taken into consideration. Also,
No consideration is given to data loss due to operation mistake.

In order to solve such a problem, there is no practical method other than the regular backup of data. In the above-mentioned prior art, effective backup corresponding to the increase in capacity due to the arraying of disk drives. No consideration is given to the method. In fact, the standard 5.2
When a 5 "disk drive (1.5 GB) is used, a parity group is configured with 10 such disk drives, and one of them is assigned as a spare disk drive and the other one is assigned as a parity disk drive. , The total capacity of this parity group is 10GB,
Standard backup tape drive (eg 10
0 kbyte / sec or about 200 kbyte / se
The time required for backup in c) is less than 14 hours. This is a major obstacle when applied to an online transaction system that makes use of the high IO throughput of the disk array device. This is because, if the data is backed up every day, the service of the system must be stopped and the backup work must be performed for these 14 hours.

An object of the present invention is to solve the above problems, to significantly reduce the data backup occupancy time, and to effectively perform the data backup, the disk drive, the disk array device, and the data storage. A system and a data backup method for a disk array system.

Further, another object of the present invention is to provide a disk drive, a disk array device, a data storage system and a data backup of a disk array system, which are capable of performing effective data backup without blocking the system service. To provide a method.

[0014]

In order to achieve the above object, the present invention stores data in place of the fault information storage device when a fault occurs in the information storage device for storing data. The spare storage device serving as the information storage device is used as a data transfer buffer between the information storage device and the data backup device when the data is backed up.

The present invention further comprises means for recording access information to the backup target file system on the information storage device backed up by the data backup device, and based on the access information. It is possible to back up the data of the backup target file system while accessing the backup target file system from a host computer.

Further, according to the present invention, in a disk driver having a semiconductor memory for temporarily storing data read from a disk, first means for setting a predetermined portion of a recording area of the disk as a write-protected area, and A second means for managing the recording area of the semiconductor memory by dividing it into a plurality of areas, and setting by the first means in one or a plurality of storage areas divided and formed by the second means in the semiconductor memory. The data read out from a part or all of the write-protected area is recorded and saved, and thereafter, the recording area that is the write-protected area on the disc is made writable.

[0017]

In a storage device arranged in an array in a computer system, an information storage device for storing data (file) used by a host computer and parity information calculated from this data, and the information storage device There is a spare storage device for logically replacing this fault information storage device when a fault occurs. In the present invention, such a backup storage device is used as a data transfer buffer between the information storage device and the data backup device when backing up data in the information storage device.

According to this, even if the data transfer rate of the data backup device is sufficiently slower than the data transfer rate of the information storage device, the data transfer rate from the information storage device to the data transfer buffer is high, so at least from the information storage device. The data transfer to the data transfer buffer is performed at high speed, and therefore, for the information storage device, the time required for data backup is extremely short.

An 8 mm tape drive and a data D which are generally used as a large capacity data backup device at present.
The data transfer rate of an AT (Digital Audio Tape Recorder) device is about 200 kB / sec, and the data transfer rate of a disk drive is about 3 MB / sec, which is more than 10 times the difference from the above large capacity data backup device. is there. Therefore, by buffering the data transfer to the backup device by the spare disk drive, the occupation time for data backup in the disk drive can be reduced to 1/10 or less. Explaining with the example given as the prior art, it means that the occupied time of the disk drive for data backup, which conventionally took about 14 hours, can be reduced to 1.4 hours or less. This means that in a system that backs up data once a day, the system downtime for this backup is reduced from 14 hours per day to 1.4 hours.

Further, according to the present invention, since the access information to the backup target file system on the information storage device backed up by the data backup device can be recorded, this backup target file system can be stored in the host computer. This file system can be transferred to the data backup device after being accessed by, for example. Therefore, even if access is made during data backup, the file system after access can be backed up.

Further, according to the present invention, the area to be backed up in the recording area of the disc is the write-protected area, and the data read from the write-protected area is temporarily stored in the semiconductor memory and saved. When all the data is read from this write-protected area and saved in the semiconductor memory, the write-protection in this area is released and normal external data writing becomes possible. The data saved in the semiconductor memory is transferred to the backup device. Therefore, even if the data backup transfer time, which is the time required from the reading of the data on the disk to the final transfer to the backup device, is sufficiently long, the data transfer from the disk to the semiconductor memory is performed at high speed. Even if the write-protection is set, it is canceled in a short time, so that the occupation time of the data backup is short.

[0022]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of a disk array device, a data storage system, and a data backup method of the disk array system according to the present invention, wherein 1 is a host computer, 1-1 is a CPU (central processing unit), -2 is a main memory unit, 1-3 is an I / O buffer, 1
-4 is an I / O (input / output) controller, 1-5 is a disk I
/ F (interface), 1-6 is an OS (operating system), 1-7 is a file management unit, 1-8 is a device driver, 2 is a disk array controller,
2-1 is a host I / F, 2-2 is a CPU, 2-3 is a data transfer control device, 2-4 is a parity generation / check unit,
2-5 is a transfer buffer, 2-6-1, 2-6-2, 2-
6-3 and 2-6-4 are drive I / Fs, 3-1 and 3 respectively.
-2 and 3-3 are backup devices, 4-1 and 4-2, respectively.
Are spare disk drives, 5 are data / parity disk drives, 6, 7-1, 7-2, 7-3, 7
-4 is SCSI (Small Computer System Interfac)
e) It is a bus.

In the figure, a host computer 1, a disk array controller 2 and a backup device 3-
1 are connected to each other by a SCSI bus 6, and the disk array controller 2 has SCSI buses 7-1 and 7-.
Disk drives are connected in an array by 2, ..., 7-3. Of these disk drives, S
The disk drives connected to the CSI buses 7-1, 7-2, ... Are the data / parity disk drives 5, and the SCSI buses 7-1, 7-2,
The disk drive connected to one of the ... Is the above parity disk drive, and the rest are data disk drives. The disk drive connected to the SCSI bus 7-4 is the spare disk drive 4-
1,4-2, ... These SCSI buses 7-1,
The number of disk drives connected to each of 7-2, ..., 7-3 is the same, and each SCSI bus 7-1, 7-2 ,.
The same-ranked disk drives (that is, the data / parity disk drives 5 arranged in a row in the horizontal direction in the drawing) constitute the above-mentioned parity group, and each parity group is connected to the SCSI bus 7-3. The spare disk drives 4 are added one by one.

Instead of the backup device 3-1,
Any of the SCSI buses 7-1 to 7-3, or a SCSI bus 7-4 that is separate from the disk array controller 2
By providing a backup device 3-3 or 3-2
May be provided.

A disk array device is constituted by the disk array controller 2, the disk drives arranged in an array, and the backup device described above.

The host computer 1 is an I / O which buffers and caches input / output data to / from the CPU 1-1, the disk drive 5 and the like for processing such as arithmetic operations.
I / O that is hardware that has a buffer 1-3 and that controls the input / output of data to / from the main storage unit 1-2, the disk drive 5, etc. for temporarily storing data and instructions used by the CPU 1-1 Control unit 1-4 (In recent workstations, a dedicated processor is used for the I / O control unit 1-4 to reduce the load on the CPU 1-1 for inputting / outputting data to / from a disk drive, etc. I / O processing) and a disk I / F1 for controlling the interface with the disk array controller 2
-5 and OS 1-6. Input / output processing of data to / from a secondary storage device such as the data / parity disk drive 5 is managed by software such as a file management 1-7 of the OS 1-6 and a device driver 1-8. These determine scheduling of data input / output to / from the data / parity disk drive 5, buffering, access data length, etc. required by the application program.

In this embodiment, each interface (I / F) is SCSI. SCSI is ANSI
Created by (American National Standard Institute), ISO (International Standard Organization)
However, it is an approved interface for computer peripherals and supports not only magnetic disk drives but also optical disk drives, magnetic tape devices, printers, scanners, etc., and is widely used in small computers.

The disk array controller 2 has its drive I / Fs 2-6-1, 2-6-2, ..., 2-6-.
The data / parity disk drive 5 and the spare disk drive 4 connected in the array 3 and arranged in an array are collectively controlled to make the host computer 1 appear as if one normal disk drive is connected. The data sent from the host computer 1 is the host I / F 2
The data / parity disk drives 5 are arranged in an array via the data transfer control unit 2-3 and the transfer buffer 2-5 according to the logical address. When there is a data request from the host computer 1, the data transfer control unit 2-3, the transfer buffer 2-5
The data distributed and arranged in the data / parity disk drives 5 arranged in an array are integrated and transferred from the host I / F 2-1 to the host computer 1 via the host I / F 2-1.

In the disk array device, even if one data / parity disk drive 5 in the same parity group fails, the data is sent from the host computer 1 so that the data can be recovered there. Redundant data (that is, parity information) is added to the incoming data and stored in the parity disk drive of the data / parity disk drives 5. The parity generation / check unit 2-4 generates and checks the parity information. When a data / parity disk drive fails, the parity information stored in the parity disk drive in the parity group and the data of the non-failed data disk drive are used to calculate the failed data / parity disk drive data. It is generated by this calculation and stored in the spare disk drive 4 for this parity group. When all the data on the failed data disk drive has been recovered and stored on this spare disk drive 4, this spare disk drive will occupy the logical location of the failed data disk drive and will replace this failed data disk drive. Becomes a data disk drive.

In this way, even if a failure of the data disk drive occurs, it is possible to automatically and promptly perform the data recovery process without human intervention. It should be noted that the spare disk drive is not used in a normal case where no failure of the data disk drive has occurred.

Backup device 3-1, 3-2 or 3
-3 is for backing up the data stored in the disk array device. As described above, in the disk array device, no data is lost even if any one data disk drive in one parity group fails, but a plurality of data disk drives simultaneously fail in the same parity group. If multiple failures occur, or if data is mistakenly deleted by an operator error, the data will be lost. The backup device 3-1, 3-2 or 3-3 is provided to prevent such a situation, and the same data stored in the disk array device is backed up in the backup device 3-1, 3-2 or 3-3. When the above-mentioned failure occurs, the data of the backup device is sent to the disk array device (recovery of the backup data) so that it can be used. As such a backup device, a storage device such as a magnetic tape device or a magneto-optical disk drive which is a replaceable medium having a low cost per storage capacity is mainly used. In recent years, a large-capacity device such as 5 GB has been developed as an 8 mm magnetic tape device.

The operation of this embodiment will be described below. (1) When the host computer 1 recognizes the spare disk drive and the backup device: FIG. 2 is a diagram showing a logical configuration as seen from the host computer 1 in this case.

In the figure, reference numeral 2 is a data / parity disk in which the disk array is centralized and controlled by the disk array controller 2 in FIG. 1 and is made to appear as one disk drive from the host computer 1. This is a conceptual representation of the entire array of the drives 5, and this will be referred to as a disk array. The buffer disk 8 conceptually represents the spare disk 4 (generally, a plurality of disks) in FIG. From the host computer 1, these are backup devices 3 (backup devices 3-1 and 3-in FIG. 1).
2 or 3-3) with one SCSI bus
As a device having an ID, it looks as shown in FIG. The SCSI bus ID is an address assigned to each device for distinguishing a plurality of devices connected to the SCSI bus.

In the case of the above configuration, the host computer 1
Recognizes the buffer disk 8 as one SCSI bus device, it is possible to perform all transfers required for backup under its own control. However, it goes without saying that it is not always necessary to manage everything.

Data transfer to back up the data in the disk array 2 to the backup device 3 is performed from the disk array 2 to the buffer disk 8 (1).
And (2) transfer from the buffer disk 8 to the backup device 3. As a preferred example, the transfer (1) is performed while the host computer 1 recognizes the file structure of the disk array 2, and the transfer (2) is a method in which the host computer 1 intervenes only a start instruction and an end confirmation. According to this, the host computer 1 can normally access the disk array 2 during the transfer (2).

For example, in the case of UNIX machine, TA
Transfer to the buffer disk 8 using the R command. At this time, the buffer disk 8 is accessed as a RAW device. When the data is transferred by the TAR command, the format is the same as when backing up to a normal MT device. Therefore, in the data transfer (2), the range to be transferred from the host computer 1 to the buffer disk 8 and the transfer destination are instructed at the time of starting. do it. However, for that purpose, the buffer disk 8 is one of the SCSI bus commands.
It is necessary to support one copy command.

In the transfer (1), the reason why the host computer 1 intervenes from the start to the end is a) processing of an open file, b) determination of an area in which data to be backed up exists, and c) disk. The above data format and normal backup device (such as a magnetic tape device. These are usually sequential access devices and have a different format from the magnetic disk which is a random access device.) This is because the transfer (1) can be performed only with the supported commands. When the transfer (1) is performed without the intervention of the host computer 1, as described later, the above a), b),
It is necessary to support the processing of c), commands for judgment, messages, etc.

However, as shown in FIG. 1, although the disk array controller 2 is actually one SCSI bus device, the disk array controller 2 of the disk array 2 and the buffer disk 8 in FIG. You have to play two roles. Although such a structure does not exist in the SCSI bus protocol, it can be implemented.

(2) When the host computer 1 recognizes the backup device but does not recognize the spare disk device: FIG. 3 is a logical configuration diagram viewed from the host computer in this case, and corresponds to FIG. Are given the same reference numerals.

Here, the buffer disk 8 in FIG. 2 is not recognized by the host computer 1. In this case, the data transfer from the data / parity disk drive 5 to the spare disk drive 4 corresponding to the above data transfer (1) is entirely managed by the disk array controller 2 including start-up and end processing. In this case, it is conceivable to perform backup using a normal backup command (TAR command in the above example). Nevertheless, buffering the spare disk drive 4 once has the effect of increasing the time during which the data / parity disk drive 5 can be used for normal access.

The host computer 1 reads the data in the disk array 2 in order to read the file management information necessary for backup, but the host computer 1 needs to read the data in the backup device 3. COP of SCSI bus without reading data
There is a method of backing up the data in the disk array 2 using the Y command. When the COPY command is received, the spare disk drive 4 is used as a data transfer buffer, and the disk array controller 2 performs control, so that faster backup can be performed.

The backup processing in the disk array controller 2 can include three types of suitable examples shown in FIGS. 4, 5 and 6 depending on the physical connection position of the backup device 3.

In FIG. 4, the backup device 3 is used as a host I / F.
The case of connection to the SCSI bus 6 on the 2-1 side is shown in FIG.
Is a drive I to which no other disk drive is connected
/ F2-6, and FIG. 6 shows a case where the spare disk drive 4 is connected to the drive I / F2-6. 4 to 6, one data / parity disk drive 5 and one spare disk drive 4 are shown, but of course, a plurality of them are arranged in an array as described in FIG. The disk array controller 2 conceptually represents one that is integrally controlled like one disk drive.

In the case of the example shown in FIG. 4, the host computer 1 reads the file management information in the area of the data / parity disk drive 5 to be backed up, and from the data / parity disk drive 5, the backup device 3
To the disk array controller 2 (the host computer 1 may intervene in the transfer). The disk array controller 2 temporarily transfers the data whose transfer destination is the backup device 3 to the spare disk drive 4, and then transfers the data from the spare disk drive 4 to the backup device 3. This can reduce the time taken for the data / parity disk drive 5 to be backed up.

In the case of the example shown in FIG. 5, the backup device 3 drives the drive I / F of the disk array controller 2.
2-6, the data transfer from the spare disk drive 4 to the backup device 3 is performed by the host I / O.
It is performed without passing through the SCSI bus 6 on the F2-1 side. Therefore, the deterioration of the access performance from the host computer 1 to the data / parity disk drive 5 during backup is small.

However, in this case, in order to make it appear that the backup device 3 is connected to the host computer 1 as shown in FIG. 3, the disk array controller 2 sends the backup device 3 from the host computer 1 to the backup device 3. Access must be passed transparently. This means that the buffer disk 8 in FIG.
Similarly to the above, although it is a method not specified in the SCSI bus protocol, it can be implemented.

The example shown in FIG. 6 is a case in which the backup device 3 is connected to the spare disk drive 4 on the same SCSI bus. In this case, the data transfer from the spare disk drive 4 to the backup device 3 is performed without passing through the inside of the disk array controller 2 either. Therefore, the host computer 1 during the backup is further transferred to the data / parity disk drive 5. It is possible to prevent the deterioration of access performance. Such an effect is maximized by the spare disk drive 4 supporting the COPY command.

(3) When the host computer 1 does not recognize the backup device or the spare disk: FIG. 7 is a conceptual configuration diagram viewed from the host computer 1 in this case. Similar to the case of recognizing the backup device 3, the actual physical connection includes three types shown in FIGS. 4, 5 and 6 as preferred embodiments. In this case, since the host computer 1 does not recognize the backup device 3, the disk array controller 2 must support the backup command. Although such a command is not specified by the SCSI bus, such a command can also be implemented as vendor unique. In order to perform the backup, first, the host computer 1 reads the file management information of the backup target area, and issues the address to be backed up and the command instructing the backup to the disk array 2 in the order of performing the backup.

(4) When a 2-port SCSI bus disk drive is used: In each of the embodiments described above, the disk drive device is assumed to be a normal 1-port type. However, in all the above embodiments, a configuration using a 2-port SCSI bus disk drive is also conceivable. One example thereof is shown in FIG. The feature of this embodiment is that the data transfer from the data / parity disk drive 5 to the spare disk drive 4 and the command / message transfer from the normal host computer 1
This is the point that there is no conflict with the access to the parity disk drive 5 on the SCSI bus. Spare disk drive 4
The data transfer from the backup device to the backup device does not conflict with the access from the host computer 1 to the data / parity disk drive 5 when the backup device 3 is connected to the position shown by the dotted line, and the backup is performed. There is little performance deterioration inside.

(5) Method of backing up without blocking the system: In the embodiment described above, the time for occupying the disk drive for the backup could be shortened. The file system or partition being backed up must have at least one group of files and its management area during processing in order not to lose the consistency of the internal data and the management information during backup. Must stop write access to the system (block the system). However, there are cases where 24-hour service is required for online transaction applications, and such a system cannot tolerate the above blockage.

In order to perform backup without blocking such a system, the following procedure may be performed. That is, 1) During backup, a file opened in the backup target disk area is recorded. The recording destination may be on the disc or on the memory. The list of open files generally resides on the main memory of the host computer. 2) Back up the target area to the backup device, including open files (of course, it may be omitted). 3) If the file was a backup target area when the file was closed, backup is performed in an area continuous to the backup performed in 2) in the same way as a differential backup.

Instead of 1), all the write accesses after the file open process are recorded, and the area without write access is backed up even if it is open, and only the area with write access is closed. After the processing, differential backup may be performed.

FIG. 9 is a block diagram showing an embodiment of the disk drive according to the present invention, in which 8a and 8b are SCSs.
I bus, 9a and 9b, SCSI controller, 10 memory controller, 11 data memory, 12 format controller, 13 read / write amplifier (hereinafter referred to as R / W amplifier), 14 magnetic head, 15 is a recording disk,
16 is a write-protected area determination unit, 17 is a spindle motor,
Reference numeral 18 is a VCM, and 19 is a battery.

In FIG. 9, the SCSI controller 9a is a SC
The protocol control of the SI bus 8a is performed, and the SCSI control unit 9b performs the protocol control of the SCSI bus 8b.
The memory control unit 10 controls data transfer among the SCSI control units 9a and 9b, the data memory 11, and the format control unit 12, and as shown in FIG.
Is divided into four storage areas and managed. Data memory 1
It is desirable that 1 has a capacity of at least several tracks of the disk 15, and here, the capacity is 1 MB. The data memory 11 is made non-volatile by the battery 19. The format control unit 12 manages the sector division method of the recording track on the disk 15. The R / W amplifier 13 has a disk 15
The magnetic head 14 is driven by the data when the data is recorded on the magnetic head 1 and the magnetic head 1 is reproduced when the data is reproduced from the disk 15.
The reproduced data signal from 4 is amplified and converted into a digital signal. The write-prohibited area determination unit 16 determines whether or not the data write-scheduled area on the disk 15 is the write-prohibited area. If the write-prohibited area is the write-prohibited area, the write operation of the format control unit 12 is stopped.

Next, the operation of this embodiment when backing up the data in the predetermined area of the recording area on the disk 15 will be described with reference to FIG.

Here, the data backup operation includes a lock instruction for saving the backup area to the data memory 11 and a backup read instruction for transferring the data saved in the data memory 11 to an external backup device. It is executed when the disk drive receives a command of a type from a host device (not shown).

When the disk drive receives a partial lock command from the host device (step 101), the disk drive 1
The lock area designated by this partial lock instruction 5 is temporarily set as the write-inhibited area (step 102), and the area for temporarily saving the data in this lock area is secured in the data memory 11 ( 103). The data memory 11 is normally in the state shown as “normal time” in FIG. 10. However, if there is a partial lock command and the capacity of the lock area is 3/4 or less of the capacity of the data memory 11, the data in the lock area is The minimum area that can store
The state is set to any one of “at the time of saving 1”, “at the time of saving 2”, and “at the time of saving 3” shown in 0. If the capacity of the lock area exceeds 3/4 of the capacity of the data memory 11, the data memory 11 is set to the state of “3 at the time of saving” in FIG. 10, and data having a capacity equal to 3/4 of the capacity of the data memory 11 is set. The data is read from the lock area, stored in the save areas of the second to fourth areas of the data memory 11 (step 104), and writing to the lock area of the disk 15 that has been read is permitted (step 105). If all the data in the lock area cannot be stored in the save area of the data memory 11 (step 10
6) The SCSI controller 10 is notified of the completion of the partial lock (step 107). When all the data in the lock area can be stored in the save area of the data memory 11 (step 106), when the save of the lock target area is completed,
Notify that writing is possible (step 108).

After that, when the backup read command is received (step 109), the disk drive detects that the SCSI bus to which this backup read command has been transferred is S.
The backup data is transferred from the save area of the data memory 11 to the host device via either the CSI bus 8a or 8b (step 110). When a notification of the completion of receiving the backup data is received from the host device (step 11
1) If the disk drive finishes transferring all the backup data in the backup area designated by the partial lock command to the host device (step 11)
2) The save area set in the data memory 11 is returned to the normal area (step 113), the partially locked state of the disk 15 is released (step 114), and the backup operation is completed.

When the untransferred portion remains in the backup area (step 112), the step 10 is continued.
The series of operations 4 to 112 are performed, and the data in the remaining lock area is saved in the data memory 11 (step 1
04) and transfer to the host device (step 110) are repeated.

After passing through steps 107 and 108,
When there is no backup read command (step 10
9) When a read command for the disk 15 is received (step 120), normal read processing is performed, and the disk 1 is read.
When a write command to write data 5 is received (step 122), writing is prohibited when it is in the write prohibited area (step 124) (step 126), but normal writing processing is performed when it is not in the write prohibited area (step 12).
5). When neither read nor write command is received, the other commands are processed (step 123).

As described above, in this embodiment, after the backup data is saved in the data memory 11,
It is possible to suppress a decrease in throughput of the disk drive for writing data to the disk 15 and performing backup.

FIG. 12 is a block diagram showing still another embodiment of the disk array device according to the present invention in which the disk drives described in FIGS. 9 to 11 are arranged in an array, and 20 is a host computer and 21. Is a disk array device, 22 is a host interface control unit (hereinafter referred to as host I / F), 23 is a data distribution control unit, 24 is a buffer memory as a cache memory, 25 is a parity generation unit, 26 is a data recovery unit, Reference numeral 27 is a magnetic tape device control unit (hereinafter referred to as MT control unit), 28 is a magnetic tape device (hereinafter referred to as MT device), and 29a to 29e are S.
The CSI control units 30a to 31e and 32 are the disk drives (hereinafter referred to as HDDs) 33 and 34 described in FIG.
Is a SCSI controller.

In FIG. 12, the host I / F 22 receives the command and write data from the host computer 20 to the disk array device 21, and receives the disk array device 21.
From the host computer 20 to the instruction execution result and the read data. The data distribution control unit 23 controls the data transfer between the host I / F 22, the cache memory 24, the parity generation unit 25, the data recovery unit 26, the MT control unit 27, and the SCSI control units 29a to 29e, 33, 34. To do. Further, the data distribution control unit 23
At the time of data writing, the write data from the host computer 20 is divided into data blocks (previous data stripes) in units of 4 kbytes and distributed to the HDDs 31a to 31e. Conversely, at the time of data reading, the HDDs 30a to 31e.
The data blocks from 31e are put together and reconstructed into read data for the host computer 20.

Each data block is written as follows. That is, now, the data block D0 has four data.
If it is divided into 0, D01, D02, and D03, as shown in FIG.
a, the data block D01 is the HDD 30b, the data block D02 is the HD 30c, and the data block D03 is
Will be sequentially arranged in the HDD 30d, and the HDD 3
0e includes these data blocks D00, D01, D0.
2, the parity block P0 generated from D03 is arranged. Next data is data block D10, D11, D
If the parity block is divided into 12 and D13 and these parity blocks are P1, these are also stored in the HDDs 30a to 30e, but the parity block P1 is stored in the HDD 30d. In the same manner, sequential data is stored. As described above, the data blocks divided from the same data and the parity blocks generated by them are collectively referred to as a parity group.

The parity block is the parity generation unit 13
However, as shown in the following equation 2, 4 of the same parity group
It is generated by exclusive ORing two data blocks.

[0066]

[Equation 2]

The cache memory 12 temporarily stores write data and read data. In addition, the data recovery unit 14 recovers the lost data block by using the normal data block and the parity block in the parity group when the data block in the parity group is lost due to the HDD failure. Is. For example, when the HDD 30a fails, the data block D0 is used by using the data blocks D01, D02, D03 and the parity block P0 as shown in the following Expression 3.
Can recover 0.

[0068]

[Equation 3]

SCSI control units 29a to 29e, 33, 3
4 controls the protocol of the SCSI bus connecting the HDD. The MT control unit 27 controls the MT device 28 as a backup device and records the data from the HDD on the magnetic tape in the MT device 16 for backup. The HDDs 30a to 31e and 32 can be connected to two SCSI buses. The HDD 33 is also a spare.

The data recording / reproducing operation of this embodiment will be described below with reference to FIG. 14. First, data from the host computer 20 is transferred to the D00 area of the HDD 30a and the HDD.
The case of recording in the D01 area of 30b will be described.

When a write command and write data are transmitted from the host computer 20 at time t01, the host I / O
The F22 transfers these write command and write data to the data distribution control unit 23. At time t02, the data distribution control unit 23 transfers the write data to the cache memory 24, and also uses the SCSI to generate the parity block.
It requests the control units 29a, 29b and 29e to read data. Therefore, the SCSI control unit 29a requests the HDD 30a to read the data block D00, the SCSI control unit 29b requests the HDD 30b to read the data block D01, and the SCSI control unit 29e requests the HDD 30e to read the parity block P0.

At time t03, the HDDs 30a, 30b, 3
0e reads the data block and the parity block of the same parity group, respectively, and the SCSI control unit 29
These data blocks are transferred to a, 29b, and 29e. Therefore, the SCSI control units 29a, 29b, 29e
The read data blocks to the data distribution control unit 2
3, and the data distribution control unit 23 stores them in the cache memory 12.

Here, the read data blocks and parity blocks from the HDDs 30a, 30b and 30e are rd0, rd1 and rp0, respectively, and the data blocks of the write data from the host computer 20 are wd0 and wd.
Let d1.

At time t04, the data blocks rd0 and rd1, the parity block rp0, and the data blocks wd0 and wd1 are transferred from the cache memory 24 to the parity generation unit 25. In the parity generation unit 25, a new parity block wp0 is calculated from these as in the following Expression 4.
Is generated and transferred to the cache memory 24.

[0075]

[Equation 4]

At time t05, the data distribution control unit 23
A write command and write data blocks wd0, wd1 and a parity block wp0 are sent to the SCSI control units 29a, 29b, 29e, respectively. As a result, the SCSI control unit 29a requests the HDD 30a to write the data block wd0 in the recording area D00, and the SCSI control unit 29b.
Is recorded on the HDD 30b in the recording area D0
1 write request to the SCSI control unit 29e
30e is requested to write the parity block wp0 to the recording area P0.

At time t06, the HDDs 30a, 30b, 3
Completion of the write operation of 0e indicates that the SCSI control units 29a, 29
The data distribution control unit 23 is notified via b and 29e,
As a result, the data distribution control unit 23 causes the host I / F 2
The completion of the write command is notified to the host computer 20 via 2. With the above, the data recording operation is completed.

Next, for example, when the SCSI control unit 29b or the SCSI bus connected thereto fails, the H
A data reproducing operation from the D00 area of the DD 30a and the D01 area of the HDD 30b will be described.

Now, if such a failure occurs at time t10, the SCSI control unit 29b notifies the data distribution control unit 23 of this. Then, at time t11, if a read command is transmitted from the host computer 20, the host I / F 22 transfers the read command to the data distribution control unit 23. At time t12, the data distribution control unit 23
Requests the SCSI controller 29a to read the HDD 30a. Furthermore, the data distribution control unit 23 determines that the SC in failure.
Instead of the SI control unit 29b, H is sent to the SCSI control unit 33.
Request reading of DD 30b.

At time t13, the HDDs 30a and 30b are
The read data block is transferred to the data distribution control unit 23 via the SCSI control units 29a and 33, respectively. Therefore, the data distribution control unit 23 supplies the data blocks from the HDDs 30a and 30b and sends them to the host computer 20. This completes the data reproducing operation.

Next, the data recovery operation of this embodiment will be described with reference to FIG.
This will be described using 5. If the HDD 30a fails at time t20, the HDD 30a detects this and the SCSI 30
This is notified to the data distribution control unit 23 via the control unit 29a. At time t21, the data distribution control unit 23 requests the data recovery unit 26 to recover the data block in the HDD 30a to the spare HDD 32. Data recovery unit 26
At the time t22, the SCSI control unit 33
30b to 30e, data blocks D01, D02, D03 and parity block P of the first parity group
Request a read of 0.

From the time t23, the data recovery unit 26 becomes H level.
Upon receiving the read data from the DDs 30a to 30e, the arithmetic operation of the equation 2 is sequentially executed to recover the data block D00. Then, at time t24, when the transfer and the data recovery operation of each block D01, D02, D03, P0 of the leading parity group are completed, the data recovery unit 26 returns to SC
The recovered data block D00 is written to the spare HDD 32 via the SI control unit 33. At the time t24, when the writing in the spare HDD 32 is completed, this is notified to the SCSI.
The data recovery unit 26 is notified via the control unit 33. The data recovery unit 26 notifies the data distribution control unit 23 of the completion of recovery of the data block D00, and permits access to the data block D00. Subsequently, recovery processing is sequentially performed on other data blocks.

At time t26, the data recovery unit 26 completes the recovery operation of the last data block Dk0, and the SCSI
The restored data block Dk0 is written to the spare HDD 32 via the control unit 33. Then, at time t27, the spare HDD 32 notifies the data recovery unit 26 of the completion of writing via the SCSI control unit 33. Data recovery unit 26
Notifies the data distribution control unit 23 that all data blocks have been recovered, and the data recovery processing is completed.

In the above data recovery processing, the SCS
Since the I control units 29a to 29e are not used, data blocks can be written / read in parallel to the HDDs 31a to 31e other than the parity group of the failed HDD 30a even during the data recovery process, and the HDDs 30b to 3e.
0e is readable.

Next, the data backup operation of this embodiment will be described with reference to FIG. At time t30, it is assumed that the host computer 20 issues a partial backup command for the areas of the data blocks D12 to D53 in the disk array device 21. Therefore, the data distribution control unit 23 uses the S
The HDDs 30a to 30e are transmitted via the CSI control units 29a to 29e.
The partial lock of the data blocks D12 to D53 is requested to 30e, and the acceptance of the write request to the area from the host computer 20 is prohibited. Since each HDD has the 1 Mbyte data memory 11 (FIG. 9) as described above,
All the data in the lock area can be stored and saved in the save area of the data memory 11, and when the save of each block in the lock target area is completed, it is notified that writing in the area is possible. In this example,
Up to 3840 kbytes of data can be saved in one parity group.

Next, at time t32, the SCSI control unit 20
All of a to 20e notify completion of partial lock of the HDD to the data distribution control unit 23, whereby the data distribution control unit 23 notifies the SCSI control unit 33 of the HDDs 30a to 30e.
Request the transfer of backup data from the MT device 28 to the MT device 28. Furthermore, when all the backup areas designated by the host computer 20 can be stored in the data memory 11 in the HDDs 30a to 30e, the acceptance of the write request from the host computer 20 to this area is permitted.

At time t33, the SCSI controller 33 requests the HDD 30c to read the first data block D12 of the partial backup. Upon receiving the data block D12, the SCSI control unit 33 sends the data block D12 to the MT control unit 27 and requests the MT device 28 to write at time t34. As a result, the MT control unit 27 starts writing the data block D12 to the MT device 28.

The SCSI controller 33 uses the data block D
When the transfer of 12 is completed, at the time t35, the HDD 30d
Is requested to read the data block D13, and the next data block D13 is read in advance to the SCSI control unit 33 before the writing of the data block D12 to the MT device 28 is completed. At the time t36, when the writing completion notice of the data block D12 is sent from the MT control unit 27, the SCS
The I control unit 33 notifies the HDD 30c of the completion of reception of the data block D12.

Subsequently, the SCSI controller 33 issues a request to write the read data block D13 to the MT device 28 and a request to read the data block D20 from the HDD 30a. At the time t37, when the writing completion notice of the data block D13 is received from the MT control unit 27,
The SCSI control unit 33 notifies the HDD 30d of the completion of reception of the data block D13, and the data distribution control unit 23 of the completion of backup of the data blocks of the parity group in the first column.

Similarly, when a write completion notification of the data block D23 is received from the MT control unit 27 at time t38,
The SCSI control unit 33 notifies the data distribution control unit 23 of the completion of backup of the data blocks of the parity group in the second column. Upon receiving the write completion notification of the last data block D53 of the backup area from the MT control unit 27 at time t39, the SCSI control unit 33 instructs the data distribution control unit 23 to finish the backup of the data block of the parity group in the second column. Then, the host computer 20 is notified of the completion of the partial backup at time t40.

The case where the backup area can be entirely saved in the data memory 11 (FIG. 9) in the HDD has been described above. However, when the backup area is larger, the data distribution control unit 23 causes the disk 15 (see FIG. 9). From the area saved in the data memory 11 in the backup area (1), acceptance of the write request from the host computer 20 is permitted.

As described above, since the SCSI control units 29a to 29e are not used during the transfer of the backup data, the HDDs are concurrently operated during the backup process.
Access to 30a-31e is possible.

In this embodiment, since a bus used for data recovery and data backup is provided in addition to the normal HDD access bus, performance degradation due to occupation of the bus for data recovery or data backup is prevented. be able to. Furthermore, since the bus used for data recovery and data backup is also used for the spare HDD, there is no need to provide a separate SCSI control unit for controlling the spare HDD 32, and the cost can be reduced accordingly.

Further, in this embodiment, since one parity group is connected by one bus, by reading HDDs sequentially, the data blocks distributed in a plurality of HDDs in the parity group are distributed. Can be rearranged in the original order.

[0095]

As described above, according to the present invention,
The occupation time of the data disk during the backup can be reduced by the ratio of the transfer speed of the backup device and the magnetic disk drive (in the typical example, the magnetic tape device has about 200 kB / sec, whereas the magnetic disk drive has 3MB / sec and 1/15).

Further, according to the present invention, by recording the file opened in the backup target area and performing the backup again after closing, the backup can be performed without blocking the system.

Further, according to the present invention, the backup data is saved in a part of the semiconductor memory in the disk drive and transferred to the backup device during the normal disk access. It is possible to prevent the drive from being occupied for a long time. Further, since the area of the backup area of the disk where the data has not been saved is write-protected, the data written there will not be accidentally destroyed.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of a disk array device, a data storage system, and a data backup method of the disk array system according to the present invention.

FIG. 2 is a configuration diagram showing a specific example of logical connection viewed from the host computer side in the embodiment shown in FIG.

FIG. 3 is a configuration diagram showing another specific example of the logical connection viewed from the host computer side in the embodiment shown in FIG.

FIG. 4 is a diagram showing a specific example of processing on the disk array controller side in the logical connection shown in FIG.

5 is a diagram showing another specific example of the processing on the disk array controller side in the logical connection shown in FIG.

FIG. 6 is a diagram showing still another specific example of the processing on the disk array controller side in the logical connection shown in FIG.

7 is a configuration diagram showing still another specific example of the logical connection viewed from the host computer side in the embodiment shown in FIG. 1. FIG.

FIG. 8 is a block diagram showing another embodiment of the disk array device, the data storage system, and the data backup method of the disk array system according to the present invention.

FIG. 9 is a block diagram showing an embodiment of a disk drive according to the present invention.

10 is a diagram showing a state during a data backup operation of the data memory in FIG.

FIG. 11 is a flowchart showing an operation at the time of data backup of the embodiment shown in FIG.

12 is a block diagram showing still another embodiment of the disk array device according to the present invention using the disk drive shown in FIG. 9. FIG.

13 is a diagram showing a data write state in each disk drive in FIG.

14 is a timing chart showing a data write operation and a data read operation of the embodiment shown in FIG.

15 is a timing chart showing a data recovery operation in the embodiment shown in FIG.

16 is a timing chart showing an operation at the time of data backup in the embodiment shown in FIG.

[Explanation of symbols]

 1 host computer 2 disk array controller 3 backup device 4 spare disk drive 5 data / parity disk drive 6 host side SCSI bus 7 drive side SCSI bus 10 memory control unit 11 data memory 12 format control unit 15 disk 16 write protected area determination unit 20 host computer 21 disk array device 23 data distribution control unit 24 buffer memory 25 parity generation unit 26 data recovery unit 28 magnetic tape device 29a to 29e SCSI control unit 30a to 30e, 31a to 31e disk drive 33, 34 SCSI control unit

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Satoshi Honda, 292 Yoshida-cho, Totsuka-ku, Yokohama, Kanagawa Prefecture, Hitachi, Ltd., Microelectronics Equipment Development Laboratory (72) Inventor Naoto Matsunami, Totsuka-ku, Yokohama, Kanagawa 292 Yoshida-cho, Hitachi, Ltd., Microelectronics Equipment Development Laboratory (72) Inventor, Shoichi Miyazawa, 292, Yoshida-cho, Totsuka-ku, Yokohama-shi, Kanagawa Hitachi, Ltd. Microelectronics Equipment Development Laboratory (72) ) Inventor Souichi Isono 292 Yoshida-cho, Totsuka-ku, Yokohama-shi, Kanagawa Hitachi, Ltd. Microelectronics Equipment Development Laboratory

Claims (20)

[Claims] 1. A disk array controller, a disk drive connected to the disk array controller by a drive interface, for storing a plurality of data arranged in an array, and data stored in the disk drive. A data backup device for backing up, wherein one or more disk drives of the plurality of disk drives are spare disk drives, and the rest are data / parity disk drives for storing data and the like, and the spare disk drives are In a disk array device that uses a data / parity disk drive as a substitute for the failed data / parity disk drive, when the data is backed up, the spare disk drive is replaced with the data / parity disk drive and the data disk. The disk array apparatus, characterized in that the data transfer buffer between the backup device. 2. The disk array device according to claim 1, wherein the data backup device is connected to a drive interface of the disk array controller and controllable from the disk array controller. 3. The disk array device according to claim 2, wherein the data / parity disk drive has two or more interfaces and is connected to the disk array controller and the data backup device. 4. The information representing the recognition result is obtained from a means or a host computer for recognizing a current usage range of a file system on the plurality of disk drives arranged on the array. Means for performing backup of data in the range of use of the file system to the data backup device without intervention of the host computer, and storing the data on the plurality of disk drives arranged on the array. A disk array device for recovering backup data of a data backup device. 5. The host computer according to claim 4, further comprising means for recognizing an address changed from the time of the previous backup or means for acquiring information indicating the recognition result from the host computer, wherein the host computer intervenes. In the disk array device, the differential backup to the data backup device is performed, and the backup data of the data backup device to the plurality of disk drives arranged on the array is recovered. 6. The means according to claim 1, wherein access to the file system to be backed up on the plurality of disk drives arranged on the array in which backup to the data backup device is executed is recorded. A disk array device characterized by having the above, and being capable of backing up data of the backup target file system while accessing the backup target file system from a host computer. 7. A computer system is connected by one or a plurality of interfaces, and a plurality of storage devices are connected to each of the one or a plurality of drive interfaces, and file data can be stored by the plurality of storage devices. In the data storage system, at least one of the storage devices is a spare storage device, and when a failure occurs in the data backup means and the storage device that stores data, instead of the failed storage device, Means for occupying the logical position of the spare storage device, and means for using the spare storage device as a data transfer buffer between the storage device and the backup means at the time of data backup of data in the storage device A data storage system having: 8. The means for recognizing a current usage range of a file system on the plurality of storage devices or a host computer according to claim 7,
A means for acquiring information indicating the recognition result, the progress of the backup of the data in the range of use of the file system to the data backup means, and the storage on the plurality of storage devices without the intervention of the host computer. 2. A data storage system, wherein the backup data of said data backup means is restored. 9. The computer system according to claim 8, further comprising means for recognizing an address changed from the time of the previous backup or means for acquiring information indicating the recognition result from the host computer, wherein the host computer intervenes. In the data storage system, the differential backup to the data backup unit is performed, and the backup data of the data backup unit to the plurality of storage devices is recovered. 10. The computer system according to claim 7, further comprising means for recording access to the file system to be backed up on the plurality of storage devices in which the backup to the data backup means is executed, A data storage system capable of backing up data of the backup target file system while accessing the backup target file system. 11. A disk array controller, a disk drive connected to the disk array controller by a drive interface, for storing a plurality of data arranged in an array, and data stored in the disk drive. A data backup device for backing up, wherein one or more disk drives of the plurality of disk drives are spare disk drives, and the rest are data / parity disk drives for storing data and the like, and the spare disk drives are In a disk array system in which a data / parity disk drive is replaced with a failed data / parity disk drive, the backup disk drive is replaced with the data / parity disk drive when data is backed up. Data backup method for a disk array system characterized by a data transfer buffer between the data backup device. 12. The data backup method for a disk array system according to claim 11, wherein the data backup device is controlled by the disk array controller. 13. The data backup method for a disk array system according to claim 11, wherein the disk drive has two or more interfaces and is connected to the disk array controller and the data backup device. 14. The information indicating the recognition result is obtained from a means for recognizing a current use range of a file system on the plurality of disk drives arranged on the array or a high-level computer. The progress of backup of the data in the range of use of the file system to the data backup device without the intervention of the host computer, and the backup of the data backup device to the plurality of disk drives arranged in the array. A method for backing up data in a disk array system, characterized by recovering backup data. 15. The data according to claim 14, wherein the information indicating the recognition result is obtained from the means for recognizing the address changed from the time of the previous backup or the host computer, and the data is obtained without the host computer intervening. A method of backing up data in a disk array system, which comprises performing a differential backup to a backup device and recovering backup data of the data backup device onto the plurality of disk drives arranged on the array. 16. The access according to claim 14, wherein the access to the file system to be backed up on the plurality of disk drives arranged on the array in which backup to the data backup device is executed is recorded, A data backup method for a disk array system, wherein data of the backup target file system can be backed up while accessing the backup target file system from a host computer. 17. A disk driver having a semiconductor memory for temporarily storing data read from a disk, comprising first means for setting a predetermined portion of a recording area of the disk as a write-protected area, and recording of the semiconductor memory. A second area for managing the area by dividing the area into a plurality of areas, and write-protection set by the first means in one to a plurality of storage areas divided and formed by the second means in the semiconductor memory. A disk driver, wherein data read from a part or all of the area is recorded and saved, and thereafter, a recording area which is a write-protected area on the disk is writable. 18. The disk driver according to claim 17, further comprising a plurality of means for connecting to a host device. 19. A disk array device comprising a plurality of disk drivers according to claim 17 or 18 and an array controller, wherein data backup of the disk driver is performed under the control of the array controller. A disk array device, characterized in that the write-protected area is set in a data backup area and data is saved in the semiconductor memory. 20. The backup path according to claim 19, wherein a plurality of paths are provided for connecting the disk driver and the array control device, and the connection path and backup data for transmitting a command for setting the write-protected area and a save operation are provided. A disk array device, wherein the connection path for transferring data from the set write-protected area of the disk device to the semiconductor memory is separated.