JP2006277042A - Array controller, disk array control method, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To shorten a time required for the data writing of the whole disk array while maintaining the security of the data of the disk array. <P>SOLUTION: This disk array 10 is configured of HDDs 110 to 113 which are equipped with write cache memories 110a to 113a, respectively. The HDD 110 and 111 are used for storing data requested from a host 30, and the HDD 112 and 113 are used for storing the redundant data of the data requested from the host 30. A WC control part 22 of an array controller 20 sets only the HDD 112 and 113 used for storing the redundant data among the HDD 110 and 113 of the disk array 10 in a write-back mode. Thus, data (redundant data) writing to the HDD 112 and 113 is completed in a stage that the redundant data are written in the write cache memories of the HDD112 and 113. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an array controller, a disk array control method, and a program suitable for controlling at least one disk array having redundancy composed of a plurality of disk drives having a write cache memory.

  A recent disk drive, for example, a hard disk drive (HDD), generally has a cache (write cache memory) in a controller (disk controller) that controls access to the disk (for example, Patent Document 1). reference). This cache is used in a mode selected from the write back mode and the write through mode. The write back mode is a mode in which when the host system (host computer) requests the HDD controller to write data, the controller notifies the host system of completion of writing when the requested data is written to the cache. It is. On the other hand, the write-through mode is a mode in which requested data is written to a cache as well as to a disk (disk medium). In the write-through mode, the writing is completed when the requested data is written to the disk. For this reason, using the cache in the write-back mode can speed up writing to the disk. Using the cache in write back mode is sometimes called turning on the write back cache.

  On the other hand, as a technique for improving the reliability of data by having redundant data, a redundant disk array (redundant disk array) configured using a plurality of disk drives (HDDs), that is, a RAID (Redundant Array of) Inexpensive Disks, or Redundant Array of Independent Disks) are known. Several levels (RAID levels) are defined for RAID, and RAID 1 (mirroring), RAID 5 (striping with parity), and the like are known. Both are technologies that enable data to be restored even if a failure occurs in any one HDD by arranging data and redundant data in a redundant disk array configured using a plurality of HDDs.

  A plurality of HDDs constituting a disk array used at the RAID 1 level are divided into a master HDD and a slave HDD. The same data is written in the master HDD and the slave HDD. A plurality of (three or more) HDDs constituting a disk array used at RAID 5 level are all used for storing data and parity data (redundant data). Parity data is distributed to each HDD. Stored.

Recently, the cost of HDDs has decreased, and in order to further improve the reliability of data storage devices, the storage devices are often composed of two sets of disk arrays (logical units). It has become to. Here, one of the two disk arrays is defined as the master disk array (master logical unit), the other is defined as the backup disk array (backup logical unit), and the data written to the master disk array is copied to the backup disk array. Is done.
JP 2003-263276 A (paragraph 0014)

The time required to write data to each HDD (disk drive) in the disk array is mainly
・ Head seek time in HDD ・ Rotation waiting time ・ Data processing time in HDD controller ・ Data transfer time to disk

  Of these times, the seek time and the rotation waiting time vary among HDDs, and as a result, appear as a difference in processing completion time between HDDs. This difference in processing completion time causes a delay until the processing is completed in the entire disk array.

  Recently, the time for writing data to the disk of the HDD has been increased. Assuming that the transfer rate to the disk is 50 MB / s, the data writing time of 64 kB is 1.28 ms. On the other hand, the average seek time of the HDD is 4 to 5 ms, and the average rotation waiting time is about 3 ms (the HDD rotation speed is 10,000 rpm class). Therefore, in order to speed up access to each HDD constituting the disk array, the head seek time and rotation waiting time of the HDD are shortened to shorten the HDD processing completion time and complete the processing between the HDDs. It is necessary to reduce variations in time.

  Therefore, by using the cache (write cache memory) mounted in each HDD in the write back mode, that is, by turning on the write back cache, the seek time, the rotation waiting time, and the processing completion time between the HDDs. It is conceivable to reduce the data write time for the entire disk array by avoiding the influence of the variation of the disk array.

  However, data writing to the HDD is completed when data is written to the cache of the HDD. For this reason, if the power is cut off before the data writing to the HDD disk is completed, there is a possibility that the data is not written to the HDD and the data is lost. Therefore, conventionally, all HDDs constituting a disk array are generally used in the write-through mode so as not to cause such data loss, that is, to ensure data integrity.

  The present invention has been made in view of the above circumstances, and its purpose is to write data in the entire disk array while maintaining data integrity of the disk array by utilizing the redundancy of the data in the disk array. An object of the present invention is to provide an array controller, a disk array control method, and a program that can reduce the time required.

  According to one aspect of the present invention, an array controller that controls at least one disk array having redundancy composed of a plurality of disk drives having a write cache memory for temporarily storing write data to be written to a disk medium Is provided. The array controller includes write control means for performing control for distributing and writing data and redundant data of the data to the plurality of disk drives of the disk array, and a disk to which redundant data is written among the plurality of disk drives. Only the drive is composed of write cache control means for setting the write back mode for completing writing to the drive when data is written to the write cache memory.

  In the array controller having such a configuration, only the write cache memory of the disk drive into which redundant data is written is used in the write back mode. That is, in a disk drive to which redundant data is written, writing to the disk drive is completed at the stage where the redundant data is written to the write cache memory. For this reason, the data write completion time of the entire disk array depends only on the data write completion time of other disk drives, that is, other disk drives to which data is written in the non-write back mode (write through mode). Therefore, it is possible to reduce the time required for data writing of the entire disk array, compared to the conventional technology in which the data writing completion time of the entire disk array depends on the data writing completion time of all the disk drives constituting the disk array. It becomes. Here, in the disk drive to which redundant data is written, after the redundant data is written to the write cache memory of the disk drive, the redundant data written to the write cache memory is actually written to the disk medium of the disk drive. If the power is shut off during the period up to this point, redundant data written in the write cache memory may be lost. However, this redundant data can be restored from the data of other disk drives constituting the disk array. In the control of the disk array to which RAID 1 or RAID 3 or the like in which the disk drive to which the redundant data is written is applied, the setting to the write back mode by the write cache control means is performed in advance at the time of system startup or the like. Is possible. On the other hand, in the control of the disk array to which RAID 5 or the like in which the disk drive into which redundant data is written is changed, every time redundant data is written, the disk drive to which the redundant data is written is checked and Only the disk drive should be set to write back mode.

  Further, the master-side disk array is read as the master-side disk array into which the data requested by the write request from the host and the redundant data of the data is written, and has a write cache memory. A copy control means for performing control for copying to a backup side disk array having redundancy composed of a plurality of disk drives is added, and all of the plurality of disk drives of the backup side disk array are set to the write back mode. A configuration that is set is good.

  In such a configuration, it is possible to reduce the time required for the process of copying the data on the master side disk array to the backup side disk array. Here, after the data read from the master disk array is written to the write cache memory of each disk drive of the backup disk array, the data written to the write cache memory is actually the disk medium of the disk drive. If the power is interrupted before the data is written to the memory, the data written in the write cache memory may be lost. However, this data can be restored using the data in the master side disk array.

  In addition, a cache memory for temporarily storing data read from the master-side disk array by the copy control means is added, and one write size that is the maximum size of data that can be written to the backup-side disk array at one time. The copy control means reads the data of the master side disk array into a cache memory with a larger size, and the copy control means copies the data read into the cache memory to the backup side disk array in a plurality of times. Good.

  In this way, using the period during which the data read into the cache memory is copied to the backup side disk array, the data write to the master side disk array requested from the host or the data read from the master side disk array, Furthermore, reading of data to be copied next from the master side disk array can be executed efficiently. In other words, using the period of data writing to the master side disk array requested from the host or data reading from the master side disk array, and further reading of data to be copied next from the master side disk array, Data read into the cache memory can be efficiently copied to the backup-side disk array.

  In addition, the cache memory is configured to be able to retain data stored in the memory even when the power is cut off, and is read from the master side disk array into the cache memory in order to effectively use the cache memory. The data may be cleared by the copy control unit after waiting for a time when the data is predicted to be actually written from the write cache memory of the plurality of disk drives of the backup side disk array to the disk medium of the disk drive. . In this way, even if the power cut occurs during the period when the copy control unit waits for the time, when the power is restored after that, the plurality of disk drives of the backup side disk array are utilized using the cache memory. It becomes possible to restore the data.

  In addition, the data read from the master side disk array and the address of the data are stored in the cache memory in units of data blocks of a certain size, and the data stored in the cache memory is stored in the backup side disk array. If the data is cleared from the cache memory after writing to the disk drive in the write back mode, and only the address of the data is cleared after waiting for the predicted time, the area of the cache memory can be used more effectively. In addition, even if the power is cut off during the period when the copy control unit waits for the above time, when the power is restored after that, the master side disk array and the backup side disk are used by using the address stored in the cache memory. If data is read from the array and compared, if the two data do not match, the data in the backup side disk array can be restored with the data read from the master side disk array.

  According to the present invention, only the write cache memory of the disk drive into which redundant data is written in the disk array is used in the write back mode, so that the data of the entire disk array is maintained while maintaining the integrity of the data of the disk array. The time required for writing can be shortened.

Embodiments of the present invention will be described below with reference to the drawings.
[First Embodiment]
FIG. 1 is a block diagram showing a configuration of a computer system according to the first embodiment of the present invention. The computer system shown in FIG. 1 includes a disk array 10, an array controller (disk array controller) 20 that controls the disk array 10, and a host (host computer) 30 that executes various applications. The host 30 uses the disk array 10 as an external storage device.

  The disk array 10 is composed of a plurality of disk drives, for example, four HDDs (magnetic disk drives) 110 (# 0), 111 (# 1), 112 (# 3), and 113 (# 4). RAID. Here, the disk array 10 is used as a RAID 1 level disk array, a so-called mirroring disk array. The disk array 10 includes a RAID1 master 11M and a RAID1 slave 11S. The RAID1 master 11M is used to write data requested from the host 30. The RAID1 slave 11S is used to write the same data as the data requested from the host 30 as redundant data (mirror data). That is, the RAID1 slave 11S is used to store a copy of the RAID1 master 11M. In the example of FIG. 1, the RAID1 master 11M is composed of HDDs 110 (# 0) and 111 (# 1), and the RAID1 slave 11S is composed of HDDs 112 (# 2) and 113 (# 3).

  The HDDs # 0 to # 3 have write cache memories (WC) 110a to 113a for temporarily storing data to be written to the disks (disk media) of the HDDs # 0 to # 3, respectively.

The array controller 20 includes a write control unit 21 and a WC control unit (write cache control unit) 22. In response to a write request from the host 30, the write control unit 21 writes the requested data (request data) to the RAID1 master 11M, and redundant data corresponding to the request data (here, the same redundancy as the request data). Data) is written to the RAID1 slave 11S. The WC control unit 22 uses each of the write cache memories 110a to 113a of the HDDs # 0 to # 3 in the write back mode or the write through mode, that is, the write back cache ON (WBC-ON). Or the write back cache OFF (WBC-OFF) state. In the present embodiment, the write control unit 21 and the WC control unit 22 are realized by the microprocessor reading and executing a control program installed in a program memory of a microprocessor (computer) (not shown) of the array controller 20. The This program can be stored in advance in a computer-readable storage medium and distributed. Further, this program may be downloaded (distributed) via a network.

  Next, control by the array controller 20 in the system of FIG. 1 will be described with reference to FIGS. FIG. 2 is a flowchart showing a procedure of control of the write cache memory by the WC control unit 22 of the array controller 20, and FIG. 3 shows HDDs # 0 to # 3 executed in accordance with write control by the write control unit 21 of the array controller 20. It is a timing chart for demonstrating operation | movement contrasting with the past.

  First, the WC control unit 22 sets each HDD on the RAID1 master 11M side, that is, each of the HDDs # 0 and # 1, to the write-through mode (WBC-OFF), for example, when the system is started up (step S1). Further, the WC control unit 22 sets each HDD on the RAID1 slave 11S side, that is, each of the HDDs # 2 and # 3, in the write back mode (WBC-ON) (step S2). Thereafter, the write cache memories 110a and 111a included in the HDDs # 0 and # 1 on the RAID1 master 11M side are used in the write-through mode, and the write cache memories 112a and 113a included in the HDDs # 2 and # 3 on the RAID1 slave 11S side are written. Used in back mode.

  Now, it is assumed that the host 30 requests the array controller 20 to write data D to the disk array 10. This data D is divided into data D1 and D2, for example. The data D1 and D2 are written in a distributed manner on the HDDs # 0 and # 1 of the RAID1 master 11M. The data D1 and D2 are also written to the HDDs # 2 and # 3 of the RAID1 slave 11S, respectively.

  Conventionally, data writing to the RAID1 master 11M and the RAID1 slave 11S is performed in the write-through mode. In this case, the time required for data writing to the HDDs # 0 and # 1 of the RAID1 master 11M and the HDDs # 2 and # 3 of the RAID1 slave 11S, that is, data for the HDD #i (i = 1, 2, 3, 4). As shown in FIG. 3 (a), the time required for writing includes the data processing time in HDD #i (the period of the hatched bar), the head seek time in HDD #i, the rotation waiting time, It is the sum of the data transfer times (the period of the bar line filled with black).

  The data processing time is the time required for the HDD #i to receive and interpret a write request from the array controller 20 (its write control unit 21). The head seek time is the time required to move the head to the first track on the disk where writing of the requested data should be started. The rotation waiting time is the time for which the disk rotates after the head is moved to the head track (that is, after the seek is completed) until the head sector to start writing the requested data reaches the head position. . Data transfer time is the time required to transfer data to the disk.

  Here, the data processing time and the data transfer time hardly depend on the HDD #i. On the other hand, the seek time and the rotation waiting time are different for each HDD #i. For this reason, as in the example of FIG. 3A, even if the data writing process is started simultaneously at time t1 in HDDs # 0 to # 3, the time at which the data writing process is completed differs for each HDD #i.

  In the example of FIG. 3A, the completion time of the data writing process of the RAID1 master 11M is determined by the completion time of the data writing process of the HDD # 0 and is time t11. On the other hand, the completion time of the data writing process of the RAID1 slave 11S is determined by the completion time of the data writing process of the HDD # 2, and is time t12 after time T1 from time t11. That is, in the entire disk array 10, the completion time of the data writing process is determined by the completion time of the data writing process of HDD # 2 of the RAID1 slave 11S, and becomes t12.

  However, in this embodiment, the write cache memories 112a and 113a of the HDDs # 2 and # 3 of the RAID1 slave 11S are different from the write cache memories 110a and 111a of the HDDs # 0 and # 1 of the RAID1 master 11M. Used in mode. In this write-back mode, when the data (D1 and D2) to be written to the HDDs # 2 and # 3 are written to the write cache memories 112a and 113a, the writing to the HDDs # 2 and # 3 is completed. As a result, the HDD # 2 and # 3 notify the array controller 20 of the completion of writing.

  Therefore, in this embodiment, as shown in FIG. 3B, assuming that the data writing process is started at the time t1 at the same time in HDDs # 0 to # 3 as in the case of FIG. In the HDDs # 2 and # 3 of the slave 11S, writing is completed at time t13 that is earlier than the HDDs # 0 and # 1 of the RAID1 master 11M by the seek time + rotation waiting time. The time required for the data writing process in HDDs # 2 and # 3 in FIG. 3B is the data processing time (the period of the hatched bar) and the data transfer time (black bars). The period of the line portion). In this case, the time represented by t12-t13 is an effect obtained by writing data in the write back mode (WBC-ON) in the RAID1 slave 11S. However, in the RAID1 master 11M, as shown in FIG. 3B, the data writing is not completed until the time t11, so the time actually shortened is t12−t11 = T1. That is, in the entire disk array 10, the completion time of the data writing process is determined by the completion time of the data writing process of the HDD # 0 of the RAID1 master 11M and becomes t11.

  As described above, in this embodiment, the write cache memories 112a and 113a of the HDD # 2 and # 3 on the RAID1 slave 11S side are used in the write-back mode. The data (D1 and D2) written in the write cache memories 112a and 113a are written to the disks of the HDDs # 2 and # 3 using the processing free time in the HDDs # 2 and # 3 (write back). ) For this reason, after the data write processing completion time t11 in the entire disk array 10, the power of the write cache memories 112a and 113a of the HDDs # 2 and # 3 is not written to the disks of the HDDs # 2 and # 3. When the interruption occurs, the data in the write cache memories 112a and 113a may be lost.

  However, in the present embodiment, the write cache memories 110a and 111a of the HDD # 0 and # 1 on the RAID1 master 11M side are used in the write-through mode. For this reason, at the time t11, the data (D1 and D2) are written on the disks of the HDDs # 0 and # 1. Therefore, the data (redundant data) of the HDDs # 2 and # 3 on the RAID1 slave 11S side can be restored using the data (D1 and D2) written to the disks of the HDDs # 0 and # 1.

[Second Embodiment]
FIG. 4 is a block diagram showing a configuration of a computer system according to the second embodiment of the present invention. In FIG. 4, the same elements as those in FIG. The computer system of FIG. 4 is different from the computer system of FIG. 1 in that, for example, a disk array 40 to which RAID 5 is applied is used instead of the disk array 10 to which RAID 1 is applied. Here, it is assumed that the disk array 40 includes three HDDs 41 (# 0), 42 (# 1), and 43 (# 2). The HDDs 41 (# 0), 42 (# 1), and 43 (# 2) have write cache memories 41a, 42a, and 43a, respectively.

  In RAID5, redundant data (parity data) P is distributed (arranged) in HDDs 41 (# 0), 42 (# 1), and 43 (# 2). In this respect, RAID 5 is different from RAID 3 in which redundant data (parity data) P is always arranged in a specific HDD.

  Now, in the system of FIG. 4, it is assumed that data (request data) D requested from the host 30 is divided into data D1 and D2 and written to the disk array 40. In this case, the write control unit 21 generates the parity data P of the data D1 and D2 as redundant data P. Which of the HDDs # 0, # 1, and # 2 writes the redundant data P is determined by the write control unit 21 based on the storage destination address of the request data D. For this reason, in the second embodiment, unlike the first embodiment, the WC control unit 22 uses each of the write cache memories 41a to 43a included in the HDDs # 0 to # 2 in the write back mode, or writes It is necessary to set whether to use in the through mode each time data is written by the write control unit 21.

  Therefore, the control of the write cache memory by the WC control unit 22 at the time of data writing will be described with reference to the flowchart of FIG. First, at the time of data writing by the write control unit 21, the WC control unit 22 inquires of the write control unit 21 about the HDD to which redundant data (parity data) P is written (step S11). Then, the redundant data write destination HDD is notified from the write control unit 21 to the WC control unit 22. It is also possible to adopt a configuration in which the redundant data write destination HDD is automatically notified from the write control unit 21 to the WC control unit 22 when the write control unit 21 writes data. The WC control unit 22 sets only the redundant data write destination HDD notified from the write control unit 21 to the write-back mode (WBC-ON), and sets all remaining HDDs to the write-through mode (WBC-OFF) (steps). S12). The state of the system in FIG. 4 is that the write control unit 21 notifies the WC control unit 22 of HDD # 2 as the write destination HDD of the redundant data P. As a result, the HDD # 2 is set to the write back mode. The remaining HDDs # 0 and # 1 are shown in the write-through mode. If the disk array 40 applies RAID3, this mode setting can be performed in advance (for example, when the system is started up) as in the first embodiment. Also, the write back mode and the write through mode can be set for each write command to the HDD.

  In the system shown in FIG. 4, after the completion of the data writing process in the entire disk array 40, the power is cut off before the data in the write cache memory 43a of the HDD # 2, that is, the redundant data P is actually written to the disk of the HDD # 2. Is assumed to occur. At this time, the data D1 and D2 corresponding to the redundant data P are written in the disks of HDD # 0 and HDD # 1, respectively. Therefore, the redundant data P to be written on the HDD # 2 disk can be restored using the data D1 and D2 written on the HDD # 0 and # 1 disks.

[Third Embodiment]
FIG. 6 is a block diagram showing a configuration of a computer system according to the third embodiment of the present invention. In FIG. 6, the same elements as those in FIG. 1 are denoted by the same reference numerals for the sake of convenience. The computer system of FIG. 6 is characterized in that it has two disk arrays corresponding to the disk array 10 in FIG. 1 as a master logical unit (master LU) 50 and a backup logical unit (backup LU) 51, respectively. is there. The backup LU 51 is used for periodically backing up the contents of the master LU 50, for example. The master LU 50 includes a RAID 1 master 500 (# 0) and a RAID 1 slave 501 (# 0), which correspond to the RAID 1 master 11M and the RAID 1 slave 11S in FIG. 1, respectively. Similarly, the backup LU 51 includes a RAID 1 master 510 (# 1) and a RAID 1 slave 511 (# 1).

In the computer system of FIG. 6, an array controller 200 is used instead of the array controller 20 in FIG. The array controller 200 has a copy control unit 23 in addition to the write control unit 21 and the WC control unit 22. The copy control unit 23 performs control for copying the data of the master LU 50 to the backup LU 51 in order to collect the backup of the data of the master LU 50 to the backup LU 51. This copy control function is called in-box replication (IBR). The array controller 20 further includes a data restoration unit 24, a cache memory 25, and a backup battery 26. However, since the data restoration unit 24, the cache memory 25, and the backup battery 26 are used in a modified example to be described later, description thereof is omitted here. 1
Next, control by the array controller 200 in the system of FIG. 6 will be described with reference to FIGS. FIG. 7 is a flowchart showing a procedure of control of the write cache memory by the WC control unit 22 of the array controller 200, and FIG. It is a timing chart for explaining operations of # 0 and RAID1 master # 1 and RAID1 slave # 1 in comparison with the conventional art.

  First, the write control unit 21 sets each HDD on the RAID1 master 500 (# 0) side included in the master LU 50 to the write-through mode (WBC-OFF), for example, when the system is started up (step S21). . The WC control unit 22 also includes each HDD on the RAID1 slave 501 (# 0) side included in the master LU 50 and each HDD included in the backup LU 51, that is, each HDD on the RAID1 master 510 (# 1) side. Each HDD on the RAID1 slave 511 (# 1) side is set to the write back mode (WBC-ON) (step S22). Thereafter, the write cache memory of each HDD on the RAID1 master # 0 side is used in the write-through mode, and the write cache memory of each HDD on the RAID1 slaves # 0 and # 1 and the RAID1 master # 1 side is used in the write-back mode. Is done. Data writing in response to a write request from the host 30 is performed only on the master LU 50. In this case, the writing of the requested data is completed when the data writing to the RAID1 master 500 included in the master LU 50 is completed, as can be understood from the first embodiment.

  Assume that the host 30 requests the array controller 200 to back up the data of the master LU 50. In this case, the copy control unit 23 of the array controller 200 executes a data copy process for copying the data of the master LU 50 to the backup LU 51. The data copy process includes a process of reading out data of each HDD of the RAID1 master # 0 and RAID1 slave # 0 included in the master LU 50 for each fixed size (here, one write size), and a fixed size (1 After the completion of the reading of the write size data, the read data is written to the HDDs of the RAID1 master # 1 and the RAID1 slave # 1 included in the backup LU 51.

  If the HDDs of the RAID1 master # 1 and the RAID1 slave # 1 on the backup LU 51 side are set in the write-through mode as in the conventional case, the read processing and write processing of data of a certain size are shown in FIG. It is executed at the timing shown in a). The example of FIG. 8A shows a state where the data read processing from the RAID1 master # 0 and the RAID1 slave # 0 (each HDD) on the master LU 50 side is started at time t20 and completed at time t21. . In the example of FIG. 8A, the data writing process for writing the data read from the RAID1 master # 0 and the RAID1 slave # 0 to the RAID1 master # 1 and the RAID1 slave # 1 on the backup LU 51 side is started at time t21. A state of completion at time t22 is shown. In FIG. 8, the length of the hatched bar represents the data processing time as in FIG. Similarly to FIG. 3, the length of the bar filled with black represents the sum of the head seek time, the rotation waiting time, and the data transfer time in the write-through mode, and the data transfer time in the write-back mode.

  However, in this embodiment, the HDDs of the RAID1 master # 1 and the RAID1 slave # 1 on the backup LU 51 side are set to the write back mode. For this reason, the reading process and the writing process of the data of the certain size are executed at a timing as shown in FIG. In the example of FIG. 8B, the data write process for writing the data read from the RAID1 master # 0 and the RAID1 slave # 0 to the RAID1 master # 1 and the RAID1 slave # 1 on the backup LU 51 side is as shown in FIG. ) Is started at time t21 as in the case of (), but the time at which the data writing process is completed is time t23 (t22−t23 = T2), which is shorter than the case of FIG. 8A by time T2. In other words, by setting the HDDs of the RAID1 master # 1 and the RAID1 slave # 1 on the backup LU 51 side to the write back mode, the time required for copying data of a certain size can be shortened by a time T2 compared to the conventional case. .

[First Modification]
Next, a first modification of the third embodiment will be described. In the first modification, when data write / read is requested from the host 30 during the process of backing up (copying) the data of the master LU 50 to the backup LU 51, the requested data write / read is executed. It is assumed that The feature of the first modification is that the requested data writing / reading is performed only for the master LU 50 (the RAID 1 master # 0 and the RAID 1 slave # 0), and the backup LU 51 side is also in this period. The point is that data writing for data copying is effectively performed.

  Hereinafter, a first modification of the third embodiment will be described with reference to FIGS. FIG. 9 is a flowchart showing a copy control procedure performed by the copy controller 23 of the array controller 200. FIG. 10 shows a RAID 1 master # 0 and a RAID 1 slave # 0 executed in accordance with the copy control performed by the copy controller 23 of the array controller 20. Data read processing, data write processing in RAID1 master # 1 and RAID1 slave # 1, and data write processing in RAID1 master # 0 and RAID1 slave # 0 executed according to control by the write control unit 21 are conventionally performed. It is a timing chart for comparing with and explaining. In the example of FIG. 10, the data writing process in RAID1 slave # 0 executed in accordance with the control by the write control unit 21 is the same as the data writing process (conventional data writing process) in RAID1 master # 0. It shall be done in write-through mode.

  In general, the array controller 20 has a cache memory 25 as shown in FIG. Here, the size of the cache memory 25 is sufficiently larger than the maximum size (one write size) at which data can be written to the backup LU 51 (the RAID1 master # 1 and the RAID1 slave # 1) by one write operation. It is assumed to be large (for example, twice or more). The copy control unit 23 reads data (data to be copied to the backup LU 51 side) read from the master LU 50 side (the RAID 1 master # 0 and the RAID 1 slave # 0), and the backup LU 51 side (the RAID 1 master # 1 and the RAID 1 slave # 1). ) Is read into the array controller 200 with a size (two times or more, for example, six times the size) larger than the size of the data to be written to (one write size), and temporarily stored in the cache memory 25 in the array controller 200. (Step S31). In the cache memory 25, for example, for each data block of a certain size, not only the data of the block but also an address indicating the write destination of the data of the block is stored.

  The copy control unit 23 takes out the data stored in the cache memory 25 in units of one write size and writes it to the backup LU 51 side (the RAID1 master # 1 and the RAID1 slave # 1) (step S32). The copy controller 23 repeats this step S32 for all data stored in the cache memory 25 (step S33). The copy controller 23 repeats step S31 for all data to be copied from the master LU 50 to the backup LU 51 (step S34).

  Under the control of the copy control unit 23, a period in which data requested by the host 30 is written to the master LU 50 (or read from the master LU 50) can be used effectively. If data to be copied from the master LU 50 to the backup LU 51 is read from the master LU 50 in units of one write size, the read processing and write processing of data of one write size are performed at the timing shown in FIG. Executed. In FIG. 10, the length of the bar line hatched in the direction from the upper right to the lower left represents the data processing time as in FIG. The length of the bar line hatched in the direction from the upper left to the lower right represents the sum of the head seek time and the rotation waiting time. The length of the bar line filled with black represents the data transfer time.

  In the example of FIG. 10A, the data read processing from the RAID1 master # 0 and the RAID1 slave # 0 on the master LU 50 side is started at time t30 and completed at times t31 and t32 (t31> t32), respectively. It is shown. In this case, in the entire master LU 50, the data read processing from the master LU 50 is completed at time t32.

  Then, in the RAID 1 master # 0 and the RAID 1 slave # 0 of the master LU 50, as shown in FIG. 10A, the processing for writing the data requested by the host 30 is started from time t32 and t31, respectively. The process of writing the data requested by the host 30 to the RAID1 master # 0 and the RAID1 slave # 0 is completed at time t34 and t33 (time t34> t32), respectively. In this case, the process of reading the next write size data to be copied from the master LU 50 to the backup LU 51 from the RAID1 master # 0 and the RAID1 slave # 0 is started from time t34 and t33, respectively. This process is completed at time t35 in the case of FIG. 10A for the entire master LU 50.

  On the other hand, in the RAID1 master # 1 and the RAID1 slave # 1 of the backup LU 51, the process of writing the data of one write size read from the RAID1 master # 0 and the RAID1 slave # 0 is timed as shown in FIG. Starting from t32. Since this process is executed in the write-back mode, it is completed in a short time. However, the process for writing the next one-write-size data is not executed unless waiting until time t35.

  On the other hand, in the first modification, the data to be copied from the master LU 50 to the backup LU 51 is a size larger than one write size (here, 6 times the size) in the read process starting from time t30. The data is read into the cache memory 25 of the array controller 200. Therefore, the data read into the cache memory 25 is copied from the master LU 50 to the backup LU 51 after the time T11 required for writing the data requested from the host 30 to the master LU 50 (or reading from the master LU 50). As shown in FIG. 10B, using the period T12 necessary for reading data having a size larger than one write size to be performed, the backup LU 51 sequentially in units of one write size from time t37. Can be written to. Here, the time T required to divide the size of data read from the master LU 50 into the cache memory 25 and write the data to the backup LU 51 in units of one write size is the above-described times T11 (= t39-t37) and T12 (= The size may be a maximum time that does not exceed the sum of t40-t39).

[Second Modification]
Next, a second modification of the third embodiment will be described. In this embodiment, the cache memory 25 in the array controller 200 is backed up by a backup battery 26. However, the data to be copied to the backup LU 51 read from the master LU 50 to the cache memory 25 is preferably cleared quickly for the next data. Therefore, it is conceivable that the data in the cache memory 25 is cleared when all the data read into the cache memory 25 is written in the write back mode of the HDD on the backup LU 51 side in the write back mode. However, after the data in the cache memory 25 is cleared, if a power shutdown occurs before the data written in the write cache memory of the HDD on the backup LU 51 side is actually written in the disk of the HDD, the write cache memory Data may be lost.

  In order to prevent such a problem, in the second modification of the third embodiment, copy control by the copy control unit 23 of the array controller 200 is executed according to the flowchart shown in FIG. In the flowchart of FIG. 11, the same parts as those in the flowchart of FIG.

  Hereinafter, copy control by the copy control unit 23 of the array controller 200 will be described with reference to FIG. When all the data having a size larger than one write size read from the master LU 50 to the cache memory 25 of the array controller 200 is written in the write cache memory of the HDD on the backup LU 51 side (step S33), the copy control unit 23 is constant. Wait for time (step S33a). This fixed time is set to a time required for actually writing the data to the disk of the HDD after the data is written to the write cache memory of the HDD on the backup LU 51 side. When the predetermined time elapses, the copy control unit 23 clears the data read to the cache memory 25 of the array controller 20 (step S33b). Thereafter, the copy control unit 23 performs the determination in step S34.

  As a result, in the second modification, even if the system power is cut off after the data read to the cache memory 25 of the array controller 20 is cleared, the data is taken out from the cache memory 25 at that time. Thus, the data written to the write cache memory of each HDD on the backup LU 51 side is actually written to the disk of the HDD. Therefore, the cache memory 25 of the array controller 20 can be used effectively while ensuring data integrity.

  The data restoration unit 24 of the array controller 20 stores the data held in the cache memory 25 by the backup of the backup battery 26 when the power supply of the system is interrupted while the copy control unit 23 waits for the predetermined time. Then, after power is restored, data is written to each HDD on the backup LU 51 side in the write-through mode. That is, the data restoration unit 24 restores the data of the backup LU 51 using the data held in the cache memory 25.

  As described above, the cache memory 25 stores an address indicating a write destination for each data block in addition to the data of the block. However, after the data held in the cache memory 25 is written to the HDD of the backup LU 51 in the write back mode, the data that requires a larger memory size than the address is cleared from the cache memory 25 and The cleared area of the cache memory 25 may be quickly released. Here, when the power supply of the system is interrupted while the copy control unit 23 is waiting for the predetermined time, the data restoration unit 24 is based on the address held in the cache memory 25 after the power is restored. It is necessary to read data from the master LU 50 and the backup LU 51. Then, the data restoration unit 24 compares the two data to determine whether the two data match, and when they do not match, restores the data on the backup LU 51 side with the data on the master LU 50 side.

  In addition, this invention is not limited to the said embodiment or its modification example as it is, A component can be deform | transformed and embodied in the range which does not deviate from the summary in an implementation stage. In addition, various inventions can be formed by appropriately combining a plurality of constituent elements disclosed in the embodiment or the modification thereof. For example, you may delete a some component from all the components shown by embodiment or its modification. Furthermore, you may combine suitably the component covering different embodiment or a modification.

1 is a block diagram showing a configuration of a computer system according to a first embodiment of the present invention. 7 is a flowchart showing a procedure for controlling the write cache memory by the WC control unit 22 of the array controller 20 in the first embodiment. 6 is a timing chart for explaining the operation of HDDs # 0 to # 3 executed in accordance with the write control by the write control unit 21 of the array controller 20 in the first embodiment in comparison with the prior art. The block diagram which shows the structure of the computer system which concerns on the 2nd Embodiment of this invention. 14 is a flowchart showing a procedure for controlling the write cache memory by the WC control unit 22 when writing data in the second embodiment. The block diagram which shows the structure of the computer system which concerns on the 3rd Embodiment of this invention. 14 is a flowchart showing a procedure for controlling the write cache memory by the WC control unit 22 of the array controller 200 according to the third embodiment. The operations of RAID1 master # 0 and RAID1 slave # 0 and RAID1 master # 1 and RAID1 slave # 1 executed in accordance with the copy control by the copy control unit 23 of the array controller 20 in the third embodiment are compared with the conventional one. The timing chart for explaining. 14 is a flowchart showing a copy control procedure by the copy control unit 23 of the array controller 200 in the first modification of the third embodiment. The data read processing in RAID1 master # 0 and RAID1 slave # 0 executed according to the copy control by the copy control unit 23 of the array controller 20 in the first modification of the third embodiment, and the RAID1 master # 1 and Timing chart for explaining the data writing process in RAID1 slave # 1 and the data writing process in RAID1 master # 0 and RAID1 slave # 0 executed in accordance with the control by the write control unit 21 14 is a flowchart showing a copy control procedure by the copy control unit 23 of the array controller 200 in the second modification of the third embodiment.

Explanation of symbols

  10, 40 ... disk array, 11M, 500, 510 ... RAID 1 master, 11S, 501, 511 ... RAID 1 slave, 20, 200 ... array controller, 21 ... write control unit, 22 ... WC control unit (write cache control unit), 23 ... Copy control unit, 24 ... Data restoration unit, 25 ... Cache memory, 26 ... Backup battery, 50 ... Master LU (master logical unit, master side disk array), 51 ... Backup LU (backup logical unit, backup side disk) Array), 41a, 42a, 43a, 110a, 111a, 112a, 113a ... write cache memory (WC), 41, 42, 43, 110, 111, 112, 113 ... HDD (magnetic disk drive)

Claims (9)

In an array controller for controlling at least one disk array having redundancy composed of a plurality of disk drives having a write cache memory for temporarily storing write data to be written to a disk medium,
Write control means for performing control to distribute and write data and redundant data of the data to the plurality of disk drives of the disk array;
Write cache control means for setting a write back mode to complete writing to the write cache memory only when the redundant data is written in the plurality of disk drives when the data is written to the write cache memory; An array controller comprising: As the master side disk array in which the data requested by the write request from the host and the redundant data of the data is written to the at least one disk array, the data of the master side disk array is read out and a plurality of write cache memories are provided. Further comprising copy control means for performing control for copying to a backup side disk array having redundancy composed of disk drives,
The array controller according to claim 1, wherein the write cache control unit sets all of the plurality of disk drives of the backup-side disk array to the write-back mode. A cache memory for temporarily storing data read from the master disk array by the copy control means;
The copy control means reads the data of the master side disk array into the cache memory with a size larger than one write size that is the maximum size of data that can be written to the backup side disk array at one time, and the cache memory The array controller according to claim 2, wherein the data read in is copied into the backup-side disk array in a plurality of times. The cache memory is configured so that data stored in the memory is not lost when the power is turned off.
The copy control means actually reads the data read from the master side disk array into the cache memory from the write cache memory of the plurality of disk drives of the backup side disk array to the disk medium of the disk drive. The array controller according to claim 3, wherein the array controller clears after waiting for a time predicted to be written to.   When the power cut occurs during the period when the copy control unit waits for the time, and then the power is restored, the cache memory is used to restore the data of the plurality of disk drives in the backup disk array 5. The array controller according to claim 4, further comprising data restoration means. The cache memory is configured so that data stored in the memory is not lost when the power is turned off.
The copy control means stores the data read from the master side disk array and the address of the data in the cache memory in units of a fixed size data block, and the data stored in the cache memory is stored in the backup memory. The data is cleared from the cache memory after writing to the plurality of disk drives of the side disk array in the write back mode, and only the address of the data is cleared after waiting for the predicted time. 3. The array controller according to 3.   When the power cut occurs during the period when the copy control unit waits for the time, and the power is restored after that, the master side disk array and the backup side disk array are stored using the addresses stored in the cache memory. The data is read from the data and compared, and when the two data do not match, the data further comprises data restoration means for restoring the data on the backup side disk array with the data read from the master side disk array. The array controller according to claim 6. In a disk array control method for controlling at least one disk array having redundancy composed of a plurality of disk drives having a write cache memory for temporarily storing write data to be written to a disk medium,
Performing control to distribute and write data and redundant data of the data to the plurality of disk drives of the disk array;
Setting only a disk drive to which the redundant data is written among the plurality of disk drives to a write-back mode for completing writing to the drive when data is written to the write cache memory. And a disk array control method. A program for causing an array controller that controls at least one disk array having redundancy to be configured of a plurality of disk drives having a write cache memory for temporarily storing write data to be written to a disk medium,
In the array controller,
Performing control to distribute and write data and redundant data of the data to the plurality of disk drives of the disk array;
Only the disk drive to which the redundant data is written out of the plurality of disk drives is set to a write back mode for completing the writing to the drive when the data is written to the write cache memory. Program for.

Images