JP2010015347A - Disk array device and control unit and method thereof - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a disk array device for performing write processing control for speeding up read processing of a plurality of mirrored disks. <P>SOLUTION: A disk storage section 4 includes: a disk device 41A to which a logical address has been allocated successively from the outside; a disk device 41B to which a logical address has been allocated successively from the inside; and a disk device 42 for writing data successively in a predetermined order. A controller section 3 writes data to either the disk device 41A or the disk device 41B having shorter access time and the disk device the disk device 42 and 42 when the disk devices 41A, 41B are not in use in write processing, and write data to either the disk device 41A or the disk device 41B not in use when one of the disk devices 41A, 41B is in use. <P>COPYRIGHT: (C)2010,JPO&INPIT

Description

  The present invention relates to a disk array device, a control device thereof, and a control method, and more particularly, to a disk array device, a control device thereof, and a control method for speeding up writing processing of mirrored data.

  In recent years, with the increase in capacity of physical disks, a large amount of data is often stored on one disk. This is preferable from the viewpoint of the bit unit price, but conversely, it causes a lot of IO (Input / Output) processing from the host computer, especially IO processing of write operation. For example, when the cache in the disk array device is depleted, the amount of writing to the physical disk increases. However, when the speed of the writing process is required, the larger the number of physical disks, The load is distributed, which is advantageous for speeding up the writing process.

  On the other hand, in general, a physical disk is often mirrored from the viewpoint of reliability. In this case, in the reading process, data may be read from one disk of the mirroring pair. However, in the writing process, it is necessary to write data to the two disks that form a mirroring pair, so that a high-speed writing process is required. The writing processing speed to the physical disk is greatly influenced by the head positioning time, which is the mechanical operation of the head. For this reason, if the address at which data is to be written, that is, the physical access position in the disk is far from the current head position, it takes time to move the head, resulting in a longer processing time.

  It should be noted that the control mechanism for the mirrored magnetic disk device maintains a communication record between the physical address and logical address of the track and the current position of the converter positioning arm for each magnetic disk device. In the read access to the record, it is known that the arm of the magnetic disk device closest to the stored track accesses the record data.

  In normal operation, the spare disk is used as a copy of the redundant disk, and the load on the redundant disk at the time of read-modify-write is distributed. It is known to improve the reliability of array devices.

Further, it is known that in an array type storage device, the write performance is improved by enabling simultaneous and independent access to a plurality of parity blocks of the parity storage device.
Japanese Patent Laid-Open No. 7-56688 Japanese Unexamined Patent Publication No. 7-44331 JP-A-6-180632

  According to the above-described prior art, in the read access, the arm closest to the track accesses the record data. Therefore, the data reading process can be speeded up. However, this technique cannot speed up the process of writing mirrored data.

  Further, according to the above-described prior art, the spare disk is used as a copy of the redundant disk. Therefore, the load on the redundant disk can be distributed. However, although this technique can indirectly speed up processing as a result of load distribution, it cannot speed up data reading and writing.

  In addition, according to the above-described conventional technology, a plurality of parity blocks can be simultaneously accessed independently. However, according to this technique, the write process to the parity block can be speeded up, but the read process and write process of the mirrored data cannot be speeded up.

  It is an object of the present invention to provide a disk array device that speeds up the write processing of mirrored data.

  Another object of the present invention is to provide a control device for a disk array device that speeds up writing processing of mirrored data.

  It is another object of the present invention to provide a disk array apparatus control method that speeds up the process of writing mirrored data.

  This disk array device includes a first disk, a second disk, a third disk, and a control means. The first disk is assigned a logical address in order from the outside of the disk. The second disk forms a mirroring pair with the first disk, and is assigned a logical address in order from the inside of the disk. The third disk writes data to be written to the first and second disks in a predetermined order. The control means writes the data to the third disk and the disk having a short access time of either the first disk or the second disk when both the first disk and the second disk are not in use. When one of the first and second disks is in use, the data is written to the first and second disks that are not in use and the third disk.

  This control device and control method for a disk array device controls a disk array device including the first to third disks described above. According to the control device and the control method, when both the first and second disks are not in use, one of the first and second disks having a short access time and the third disk And when one of the first and second disks is in use, the data is written to the non-use disk of the first and second disks and the third disk. Include.

  According to this disk array apparatus, its control apparatus, and its control method, when both the first and second disks are not in use, the mirrored data has the access time of either the first or second disk. Written on a short disk and a third disk. Since the third disk writes data in a predetermined order, its access time is very short. Therefore, the access time in this case is determined depending on the shorter access time in the first and second disks. Also, when one of the first and second disks is in use, the mirrored data is written to the first and second disks that are not in use and the third disk. Accordingly, the access time in this case is determined depending on the access times of the first and second disks.

  As a result, the access time in the write processing of the mirrored data can be made the shortest time at that time. As a result, the writing process of the mirrored data can be accelerated.

  FIG. 1 is a diagram showing the configuration of a disk array device and its control device according to an embodiment of the present invention.

  In FIG. 1, the host device 1 includes, for example, a host computer, a server (server device), and the like. The host device 1 includes an HBA (Host Bus Adapter) 11 as data transfer means. The HBA 11 requests the disk array device 2 to write data, and transmits the data to be written to the disk array device 2. In addition, the HBA 11 requests the disk array device 2 to read data, and receives the data read from the disk array device 2. The HBA 11 is, for example, a card adapter that conforms to a PCI (Peripheral Components Interconnect) bus of a computer.

  The disk array device 2 includes a controller unit 3 and a disk storage unit 4. The controller unit 3 controls writing of data to the disk storage unit 4 and reading of data from the disk storage unit 4 in response to a request from the HBA 11.

  The controller unit 3 includes a channel control unit 31, a CPU 32, a cache memory 33, a disk control unit 34, a disk management table 35, and a hot spare disk data management table (HSD management table) 36.

  The disk storage unit 4 includes a plurality of disk devices 41. The disk device 41 is used for storing data. The disk storage unit 4 mirrors data. In the disk device 41, one disk constituting the mirroring pair is represented as “disk device 41A”, and the other disk is represented as “disk device 41B”. The disk device 41A and the disk device 41B constituting the mirroring pair are determined in advance.

  The disk storage unit 4 includes one or a plurality of disk devices 42. The disk device 42 is a spare disk (hot spare disk). That is, the disk device 42 does not constitute a mirroring pair, and is used instead of the failed disk device 41 when the disk device 41 fails.

  However, in this example, the disk device 42 is used not only as a spare disk of the disk device 41 but also as a data buffer for temporarily recording data to be mirrored. In other words, the disk device 42 that is not used as a spare disk is used as a data buffer.

  The disk device 41 and the disk device 42 are, for example, magnetic disk devices, that is, hard disk devices (HDD: Hard Disk Drive). The disk device 41 and the disk device 42 may be disk-type media such as optical disks.

  In the disk storage unit 4, one disk device 41A, one disk device 41B, and one disk device 42 may be provided as a set, and a plurality of sets may be provided. Alternatively, one disk device 41A and one disk device 41B may be set as one set, and one disk device 42 may be provided for each set or for a plurality of sets.

  In the controller unit 3, the channel control unit 31 executes data transfer with the HBA 11 of the host device 1. That is, the channel control unit 31 receives a data write request from the HBA 11 and data to be written, and sends it to the CPU 32. In addition, the channel control unit 31 sends a data read request from the HBA 11 to the CPU 32 and sends the data read from the cache memory 33 or the disk storage unit 4 to the HBA 11.

  The CPU 32 controls the controller unit 3 to execute a data write request and a data read request from the HBA 11. For this purpose, the CPU 32 executes an operation for data transfer in the controller unit 3, and controls data transfer among the channel control unit 31, the cache memory 33, and the disk control unit 34 based on the calculation. The cache memory 33 temporarily stores data when data transfer is performed between the channel control unit 31 and the disk control unit 34.

  The disk control unit 34 controls reading and writing of data with respect to the disk device 41A, the disk device 41B, and the disk device 42 according to the control from the CPU 32. For this purpose, the disk control unit 34 uses a disk management table 35 and an HSD management table 36. These will be described later with reference to FIG.

  FIG. 2 is an explanatory diagram of disks in the disk array apparatus of FIG.

  In FIG. 2A, the disk device 41A is assigned logical addresses in order (in ascending order) from the outside of the disk as indicated by an arrow A. In other words, logical addresses are assigned in accordance with the physical positions of tracks and sectors in order from the outer circumferential direction of the disk. Therefore, in the disk device 41A, the ascending order of the logical addresses matches the ascending order of the physical addresses. Such a disk will be referred to as a “mirror positive” or first disk.

  In FIG. 2B, as indicated by the arrow B, the disk device 41B is assigned logical addresses in order (in ascending order) from the inside of the disk. That is, logical addresses are assigned in accordance with the physical positions of tracks and sectors in order from the inner circumferential direction of the disk. Therefore, the logical address in the disk device 41B is opposite to the logical address in the disk device 41A. In the disk device 41B, the ascending order of the logical addresses is opposite to the ascending order of the physical addresses. Such a disk is referred to as a “mirror secondary” or a second disk.

  In FIG. 2C, the disk device 42 is assigned logical addresses in order (in ascending order) from the outside of the disk, as indicated by an arrow C. That is, it is the same as the disk device 42. However, in a disk device 42 used as a data buffer (hereinafter simply referred to as disk device 42), data to be written is written in a predetermined order. The order of writing is determined in the disk device 42 independently of the address of the data in the disk device 41. For example, data to be written is written in order from the beginning of the physical address on the disk. Specifically, in the disk device 42, the data in the writing process is written by the disk control unit 34 at a physical address subsequent to the physical address where the writing process immediately before the writing process was performed.

  FIG. 2A shows only one physical disk 101A among a plurality of disks (physical disks) provided in the disk device 41A. FIG. 2B shows only one physical disk 101B among the plurality of physical disks provided in the disk device 41B. FIG. 2C shows only one physical disk 101C among a plurality of physical disks included in the disk device 42.

  The physical disk 101A includes a head 103A, an arm 105A, and tracks 111A and 113A. The track 111A is a head position track, and the track 113A is the next adjacent track of the track 111A. The sector 115A is a sector at the head position at the time of the last access on the track 111A. The same applies to the physical disk 101B and the physical disk 101C.

  In the disk device 41A, since a logical address is assigned from the outer peripheral direction, the track 113A, which is an adjacent track inside the track 111A, becomes the next track of the track 111A on the logical address. The same applies to the disk device 42. On the other hand, since a logical address is assigned from the inner peripheral direction in the disk device 41B, the track 113B, which is an adjacent track outside the track 111B, becomes the next track of the track 111B.

  3A shows the disk management table 35, and FIG. 3B shows the HSD management table 36.

  In FIG. 3A, the disk management table 35 stores the disk type and head position information at the time of the last access for each physical disk in the disk devices 41 and 42. The physical disks of the disk devices 41 and 42 are represented by physical disk numbers (ID). For example, the physical disk number of the physical disk 101A of the disk device 41A is “1”. The disk type is a distinction as to whether the disk device 41 or 42 to which the physical disk belongs is “mirror primary”, “mirror secondary”, or “hot spare disk (data buffer)”. The final access head position is a head position after the last (immediate) access in the disk device 41 or 42 is executed. For example, assume that the final access head position of the disk device 41A is represented as P (a).

  For example, in the physical disk 101A of the disk device 41A whose physical disk number is “1”, the final access head position is the sector 115A of the track 111A in FIG. Therefore, the logical address value of the sector 115A is stored in the disk management table 35 as the final access head position P (a). The final access head position P (b) of the physical disk 101B is the logical address value of the sector 115B of the track 111B in FIG. The final access head position P (c) of the physical disk 101C is the logical address value of the sector 115C of the track 111C in FIG.

  The final access head position P (a) is calculated by the following process, for example. The same applies to the final access head position P (b) of the disk device 41B and the final access head position P (c) of the disk device 42.

  For example, to simplify the calculation, the physical disk is divided into, for example, two equal parts in the circumferential direction, and the outer half of the circumference is set as the first area and the inner half of the circumference is set as the second area. The number of tracks on the physical disk 101A is assumed to be 100. Accordingly, the first area is from the outermost periphery da (first track) to the track (50th track) immediately before the midpoint db of the physical disk 101A. The second area is from the track immediately after the midpoint db of the physical disk 101A (the 51st track) to the innermost circumference dc (the 100th track). The sector density X in the first area is 1000 sectors / track, and the sector density Y in the second area is 500 sectors / track. In the following example, logical addresses are assigned in order from the outer periphery.

  In such an example, the final access head position P (a) in the first area is obtained by P (a) = LBA1 ÷ X. Here, LBA1 is the value of the logical block address LBA (Logical Block Address) of the last access in the first area. In this example, it is assumed that the LBA is calculated in units of sectors. The LBA value is assumed to be that the first sector, which is the first sector in the first track, which is the first track, is “address 0”, and the address increases by one each time one sector is added.

  For example, when LBA1 = 100002, the final access head position P (a) is 10002 (address / sector) ÷ 1000 (sector / track) = second sector of the 10th track.

  The final access head position P (a) in the second area is obtained by P (a) = (LBA2-db * X) / Y + db. Here, LBA2 is the value of the logical block address LBA of the last access in the second area.

  For example, when LBA1 = address 60014, the final access head position P (a) is (LBA2-50 × 1000) ÷ Y + 50 = (60014−50 × 1000) (address / sector) ÷ 500 (sector / track) +50 ( Track) = the 14th sector of the 70th track.

  In FIG. 3B, the HSD management table 36 stores the head storage position, data length, and regular storage position for each of a plurality of data stored in the disk device 42. The head storage position is represented by, for example, a physical address (or logical address) in the disk device 42. The regular storage position is a logical address (hereinafter referred to as a regular address) when the data is stored in the disk devices 41A and 41B that form a mirroring pair. The regular address includes a physical disk number (ID) in the disk devices 41A and 41B and a logical address in the physical disk.

  The plurality of data is registered in the HSD management table 36 in the order in which the data is stored in the disk device 42. For example, when looking at certain data, the head storage position is SA (d), the data length is L (g), the normal storage position is physical disk number 1, and the storage position is P (j). In this case, the head storage position SA (e) of the next data is the address next to the address obtained by adding the data length L (g) to the head storage position SA (d) of the immediately preceding data. Therefore, in the disk device 42, the writing process is always started from the position next to the final access head position in the access for the immediately preceding writing process.

  FIG. 4 is a diagram showing a writing process in the disk device 42.

  In FIG. 4, it is assumed that the disk device 42 stores a plurality of data in blocks Ba, Bb, Bc, Bd, and Be. For example, it is assumed that the data of the block Ba is the data of the head storage position SA (d) in FIG. In this case, in the track 111C of the physical disk 101C of the disk device 42, data of the data length L (g) is stored in the block Ba. The block Ba is stored in, for example, a plurality of sectors, and the top logical address is SA (d). The normal storage position of this data is the storage position P (j) in the disk device 41A whose physical disk number is 1. The same applies to the block Bb and the like.

  For example, when the data of the block Bb is written after the data writing process of the block Ba, the head 103C hardly needs to move. Therefore, the access time for the writing process in the disk device 42 can be substantially ignored.

  Further, for example, it is assumed that a period (empty time) in which there is no access to the disk device 41A or 41B storing the data of the block Ba occurs after the data writing process of the block Be. Using this idle time, the disk control unit 34 reads the data of the block Ba from the disk device 42 and writes it to the regular storage location in the disk device 41A or 41B. In this case, as a result of the reading process, in the disk device 42, the position of the head 103C is the last position of the data of the block Ba.

  Therefore, the disk control unit 34 returns the head 103C to the state immediately before the writing process to the regular storage position after the writing process to the regular storage position. That is, the position of the head 103C is the last position of the data of the block Be. As a result, when the next data of the data of the block Bb is written, the head 103C hardly needs to move. Therefore, also in this case, the access time for the writing process in the disk device 42 can be ignored in effect.

  As described above, the disk device 42 operates as a data buffer that performs writing processing at high speed. As a result, the writing process can be performed at a higher speed than writing the data to be mirrored on both the disk devices 41A and 41B. As a result, the access processing load of the disk array device 2 can be distributed.

  FIG. 5 is a diagram showing a process of reading the mirrored data from the disk device 41A and the disk device 41B.

  For example, the data to be read is present in the sector 125A as shown in FIG. 5A in the disk device 41A, and the sector as shown in FIG. 5B in the disk device 41B. Suppose that it exists at 125B.

  In this case, the disk control unit 34 reads the data from either the disk device 41A or the disk device 41B and transfers it to the cache memory 33. After this transfer, the controller unit 3 transfers the data from the cache memory 33 to the HBA 11 of the host device 1.

  At this time, the disk control unit 34 determines, based on the disk management table 35, whether to read the data from the disk device 41A or the disk device 41B. That is, the disk device 41A or 41B that minimizes the time until the head 103A or 103B moves to the data block, that is, the head positioning time (seek time + search time) is selected.

  Specifically, the disk control unit 34 reads the final access head position P (a) of the head 103A and the final access head position P (b) of the head 103B with reference to the disk management table 35. Then, the disk controller 4 calculates the positioning time of the head 103A from the final access head position P (a) of the head 103A and the logical address value of the sector 125A in which the data exists. Similarly, the positioning time is calculated for the head 103B in the disk device 41B.

  For example, in the case of the sector 125A, the positioning time is about (head movement for one track) + (1/3 rotation). On the other hand, in the case of the sector 125B, the positioning time is about (head movement for 98 tracks) + (1/4 rotation). Therefore, the movement time of the sector 125A is shorter.

  Based on the calculation result, the disk control unit 34 selects the sector 125A of the physical disk 101A (disk device 41A) having a shorter moving time of the head 103A. Therefore, the data is read from the disk device 41A.

  As described above, the mirrored data can be read from one of the disk devices 41A and 41B at a higher speed. As a result, the access processing load of the disk array device 2 can be distributed.

  The above is the processing when the data does not exist in the cache memory 33. When the data exists in the cache memory 33, the data is read from the cache memory 33 and transferred to the host device 1.

  As described above, the disk control unit 34 performs write processing and read processing on the disk device 41A, the disk device 41B, and the disk device 42. That is, the disk control unit 34, when both the disk device 41A and the disk device 41B are not in use in the writing process, the disk device 41A or the disk device 41B, the disk device 42, To write the data. This writing process is referred to as a first writing process.

  In addition, when one of the disk device 41A and the disk device 41B is in use during the writing process, the disk control unit 34 assigns the disk device 41A and the disk device 41B to the disk device that is not in use and the disk device 42. Write the data. This writing process is referred to as a second writing process. Here, “one in use” means a case where one of the devices is in a reading process, and no other writing process is executed during a certain writing process.

  The disk control unit 34 writes the data written in the disk device 42 in the first or second writing process into one of the disk device 41 or the disk device 42. This writing process is called a writing process to a regular address. The disk device written at this time is a disk device in which the data has not been written in the first or second writing process. The writing process to the regular address is executed during a period in which other writing processes and other reading processes are not performed on the disk.

  Thereby, finally, in the writing process, the data is written by the disk control unit 34 to the same logical address (regular address) in the disk device 41A and the disk device 41B. After the writing process to the regular address, the information on the data in the HSD management table 36 is deleted by the disk control unit 34.

  The disk control unit 34 reads the data from the disk having the short access time of either the first disk or the second disk in the reading process of the data written by the writing process.

  Based on the information stored in the disk management table 35, the disk control unit 34 positions the head that moves from the disk device 41A and the disk device 41B to the physical access position corresponding to the logical address to be written when a data write request is made. The disc having the shortest time (“seek time + search time”) is determined and selected.

  Then, the disk control unit 34 writes data to the disk thus selected. At the same time, the disk control unit 34 temporarily writes the mirroring target data to a predetermined logical address of the disk device 42 in which the head position is maintained and the head moving time is short, in a predetermined order. Include. This writing information is recorded and stored in the HSD management table 36. After the writing, the head position of the disk device 42 is controlled to be held at the physical position of the logical address.

  As a result, one write data is temporarily written to the disk device 42 as compared with the operation of simultaneously writing the same data to both disks to the disk device 41A and the disk device 41B during mirroring. Thus, the head positioning time can be shortened, and the occupation time for the temporary load process of the write process of the disk array apparatus 2 can be shortened.

  Further, the data temporarily written in the disk device 42 in this way is based on the HSD management table 36 during the free time when the original mirrored disk is not accessed. Processing data is transferred from the disk device 42 and written to a predetermined logical address of the disk. As a result, the load of access processing of the disk array device 2 can be distributed.

  Further, at the time of a data read request from the host device 1, the controller unit 3 responds to the logical address of the block of the read request data from the disk device 41A or the disk device 41B based on the information stored in the disk management table 35. The disk having the shortest positioning time (seek time + search time) of the head moving to the physical position to be determined is selected and selected.

  After selecting the disk, the controller unit 3 reads data from the selected disk, and controls to hold the head position at the physical position of the logical address after the completion of the reading. The logical address in this state is notified from the disk control unit 34 to the disk management table 35, and the information is stored in the disk management table 35.

  As described above, in the mirrored disk array device 2, when a data read request is made, it is possible to determine and select a disk that can shorten the head positioning time among the disk devices 41A and 41B. The read processing time can be shortened.

  Hereinafter, the writing process and the reading process in the disk array apparatus of FIG. 1 will be described with reference to FIGS.

  6 shows the flow of the first write processing in the disk array device 2, and FIG. 7 shows the flow of the second write processing in the disk array device 2. FIG. 8 shows the flow of read processing in the disk array device 2.

  FIG. 6 is an explanatory diagram showing the flow of the first write processing (first write processing) when neither the disk device 41A nor the disk device 41B is in use.

  In the controller unit 3, the CPU 32 notified of the write request from the channel control unit 31 checks whether or not there is a free space in the cache memory 33. When there is a free space on the cache memory 33, the CPU 32 issues an instruction to write data onto the cache memory 33 from the channel control unit 31, and the data is written onto the cache memory 33 according to this instruction (TR11). The cache memory 33 notifies the disk control unit 34 that there is a write request. Upon receiving this notification, the disk control unit 34 determines and selects a disk device to which data is to be written from the disk storage unit 4. That is, as described above, the disk control unit 34 refers to the disk management table 35 and selects the disk device 41A or the disk device 41B which has a shorter access time. In this case, for example, the disk device 41A is selected.

  Next, the disk controller 34 issues a write data transfer instruction to the cache memory 33, and thereafter receives the data (TR12). The disk control unit 34 first writes the transferred data to the disk device 41A (TR13).

  After this writing is completed, the disk control unit 34 performs a process of writing the same data as the data written to the disk device 41A to the disk device 42 (TR14). At this time, the disk device 42 is sequentially used from the outer peripheral direction of the disk track as described above.

  The disk control unit 34 writes the data to the disk device 42 and, at the same time, stores the head storage position and the like in the disk device 42 for the written data in the HSD management table 36. Thereafter, the disk control unit 34 writes data from the disk device 42 to the disk device 41B based on the HSD management table 36 when the disk device 41B is free (TR15).

  On the other hand, in the controller unit 3, when there is no space on the cache memory 33, the same processing as the processing TR12 to TR15 is performed on the disk device 41A or the disk device 42 in order to make space on the cache memory 33. . That is, the cache memory 33 writes the data to the disk device 41 for data selected by, for example, LRU (Least Recently Used) control. As a result, an empty area is created on the cache memory 33. Therefore, processes TR11 to TR15 are executed using this free area.

  Also in this write process, the disk control unit 34 that has received the write request notification from the cache memory 33 determines the disk device to which data is to be written from the disk storage unit 4 and refers to the disk management table 35. The one with the shorter access time is selected from the disk devices 41A and 41B.

  In the writing process to the mirrored disk as described above, the head of the disk device 42 is controlled by the disk control unit 34 so that the moving time of the head is the shortest (or a time close thereto). As a result, it is possible to distribute the load of the write process rather than simultaneously performing the write process on the normal mirroring primary and secondary disks, and to shorten the waiting time for the completion of the write process from the write request on the host device 1 side.

  As a specific example, the access time Tacs in a hard disk (Hard Disk Drive: HDD) is obtained by the following equation. That is, Tacs = Tseek + Trot + Ttrs. The access time Tacs is the time from the data processing request until the data reading or writing process is completed.

  Here, the seek time Tseek (average positioning time) is a time required to move the magnetic head to a track storing data in order to read data. The seek time Tseek is on the order of several milliseconds in a standard HDD, for example, 8 milliseconds.

  The search time Trot (average rotation waiting time) is a time until a sector in which data is stored reaches the head. The search time Trot is normally obtained from the number of revolutions RPM (rpm) per minute. That is, Trot = 1 / RPM × 60 (seconds) / 2. For example, when the rotation speed RPM is 10,000 (rpm), the search time Trot = 60 (seconds) / 10000 (rotation) / 2 = 3 (milliseconds).

  The data transfer time Ttrs is a time for reading or writing data of a target sector. That is, Ttrs = 60 (seconds) / RPM × 1 / (average number of sectors per track). For example, when the average number of sectors per track is 1000 sectors, the data transfer time Ttrs = 60 (seconds) / 10000 (rpm) × 1/1000 = 0.006 (milliseconds).

  According to the above example, the access time Tacs = 8 (milliseconds) +3 (milliseconds) +0.006 (milliseconds). From the above, it can be seen that most of the access time Tacs depends on the seek time and the search time, and the data transfer time can be almost ignored.

  As described above, in the case of the first write process, the data to be written to one of the mirrored disks is determined from the storage information of the current write access position and the last access head position in the disk device 41A and the disk device 41B. To do.

  Since the logical address assignment in the outer circumferential direction and the inner circumferential direction is different between the disk device 41A and the disk device 41B, by selecting the disk with the shortest head position, the access time is averaged as Tseek (average positioning time) / It can be shortened to 2. For example, in the calculation example of the access time Tacs described above, the access time can be reduced by an average of Tseek / 2 = 8 (milliseconds) / 2 = 4 (milliseconds).

  Further, according to the disk array device 2, when the write request processing from the host device 1 is performed, the data to be written to the other disk of mirroring is written to the disk device 42, so that the position of the holding head position of the disk device 42 is changed. Since the logical address is written, the aforementioned (a) seek time and (b) search time can be eliminated. Accordingly, in the above-described calculation example of the access time Tacs, the access time can be reduced by Tseek + Trot = 8 (milliseconds) +3 (milliseconds) = 11 (milliseconds).

  As a result, compared with the operation of writing the same data to both disks simultaneously in the disk device 41A and the disk device 41B during mirroring, one write data is temporarily written to the disk device 42. Thus, the occupation time for the temporary load process of the write process of the disk array device 2 can be shortened.

  FIG. 7 is an explanatory diagram of the writing process (second writing process) when the disk device 41A is performing the reading process during the writing process.

  In the second writing process, the processes TR21 to TR25 are basically executed in the same manner as the processes TR11 to TR15 in the first writing process. In this case, in the process TR23, as described above, since the disk device 41A is in use, the data is written to the disk device 41B.

  As described above, in the calculation example of the access time Tacs described above, the access time can be reduced on average by Tseek + Trot = 8 (milliseconds) +3 (milliseconds) = 11 (milliseconds).

  Even if the access time in the disk device 41A is shorter than the access time in the disk device 41B, the disk device 41A is not selected. On average, the writing process can be completed more quickly by writing to the disk device 41B than waiting until the disk device 41A in use becomes available. Accordingly, the writing process is not performed after the disk device 41A being used becomes available, and the access time is not calculated and compared for the selection. Thereby, the processing burden on the disk control unit 4 can be reduced.

  FIG. 8 is a diagram for explaining the operation of read processing when a disk read processing request is made.

  When a data read request is generated from the host device 1 to the disk array device 2, in the controller unit 3, is the CPU 32 notified of the read request from the channel control unit 31 receiving the read request, free from the cache memory 33? Check for no. If it exists (cache hit), the CPU 32 issues an instruction to read data from the cache memory 33 to the channel control unit 31, and the data is read to the channel control unit 31 by this instruction. This data is transferred from the channel control unit 31 to the HBA 11 of the host device 1.

  On the other hand, when there is no data in the cache memory 33 (no cache hit) in the controller unit 3, the CPU 32 issues an instruction to read data to the disk control unit 34. In response to this, the disk control unit 34 selects either the disk device 41A or the disk device 41B, and reads data from the selected disk device (for example, the disk device 41A) (TR31). Next, data is transferred from the disk control unit 34 to the cache memory 33 (TR32), and then transferred from the cache memory 33 to the host device 1 via the channel control unit 31 (TR33).

  Here, in the process TR31, the disk device 41 is selected as follows. For example, the disk control unit 34 refers to the HSD management table 36 and checks whether or not the data is stored in the disk device 42. If not stored in the disk device 42, the disk control unit 34 selects either the disk device 4141A or the disk device 41B with the shorter read processing time, as described above with reference to FIG. On the other hand, when stored in the disk device 42, the disk control unit 34 selects the other disk device 41A or 41B that forms a mirroring pair with the disk device based on the regular address stored in the HSD management table 36. select.

  In this example, the disk device 42 is not selected in the process TR31. This is because when the disk device 42 is selected, the head moves due to the read processing, which becomes an obstacle to shortening the write processing time.

  As described above, the access time can be shortened to Tseek (average positioning time) / 2 on average. For example, in the calculation example of the access time Tacs described above, the access time can be reduced by an average of Tseek / 2 = 8 (milliseconds) / 2 = 4 (milliseconds).

It is a figure which shows an example of the embodiment of a disk array apparatus and its control apparatus. It is a figure which shows the relationship between the head position of a disk, a track | truck, and a logical address. It is a figure which shows a disk management table and an HSD management table. It is a figure which shows the write processing of a hot spare disk. It is a figure which shows the reading process of a mirroring disk. It is a figure which shows the flow of the 1st write-in process in a disk array apparatus. It is a figure which shows the flow of the 2nd write-in process in a disk array apparatus. It is a figure which shows the flow of the read-out process in a disk array apparatus.

Explanation of symbols

1 Host device 2 Disk array device 3 Controller unit 4 Disk storage unit 11 Host bus adapter (HBA)
31 channel controller 32 CPU
33 Cache memory 34 Disk controller 35 Disk management table 36 Hot spare disk data management table (HSD management table)
41, 42 discs

Claims (5)

A first disk assigned logical addresses in order from the outside of the disk;
A second disk that is in a mirroring pair with the first disk and is assigned a logical address in order from the inside of the disk;
A third disk for writing data to be written to the first and second disks in a predetermined order;
When both the first and second disks are not in use, the data is written to either the first disk or the second disk, which has a shorter access time, and the third disk, And a control means for writing the data to the third disk and the third disk that are not in use when one of the second disks is in use. Disk array device to be used. The control means writes the data written on the third disk to the first and second disks, on which the data has not been written in the writing process, to the other write on the disk. 2. The disk array device according to claim 1, wherein writing is performed during a period in which processing and other reading processing are not performed. 2. The disk according to claim 1, wherein the control unit reads the data from a disk having a short access time of either the first disk or the second disk in a reading process of the data written by the writing process. Array device. A first disk assigned a logical address in order from the outside of the disk, a second disk having a mirroring pair with the first disk, and assigned a logical address in order from the inside of the disk; A control device for a disk array device comprising a third disk for writing data to be written to the first and second disks in a predetermined order;
When both the first and second disks are not in use, the data is written to either the first disk or the second disk, which has a shorter access time, and the third disk, When one of the second disks is in use, the data is written to the non-use disk and the third disk of the first and second disks. Control device. A first disk assigned a logical address in order from the outside of the disk, a second disk having a mirroring pair with the first disk, and assigned a logical address in order from the inside of the disk; A control method for a disk array device comprising a third disk for writing data to be written to a first disk and a second disk in a predetermined order,
When both the first and second disks are not in use, the data is written to either the first disk or the second disk, which has a shorter access time, and the third disk, When one of the second disks is in use, the data is written to the non-use disk and the third disk of the first and second disks. Control method.

Images