JP3621666B2 - Disk control system and control method thereof - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a disk control system for controlling a plurality of disk devices and a control method for controlling the system.
[0002]
[Prior art]
In recent years, RAID (Redundant Array of Independent Disks) technology has begun to be widely used in server computers from the viewpoint of improving the performance and reliability of disk devices.
[0003]
RAID is a technique for grouping a plurality of disk devices to realize a single logical disk. In a disk control system using RAID, performance and reliability can be improved by distributing data and redundant data for failure recovery to a plurality of disk devices.
[0004]
There are various RAID levels such as 0, 1, 4, 5, and 10, among which RAID 5 is widely used because of its good price-to-capacity ratio and excellent fault tolerance. RAID 5 is composed of a disk array in which a stripe group including a plurality of data stripes and a parity stripe is arranged across a plurality of disk devices. The RAID 5 disk array includes N + 1 (N is 2 or more) disk devices in which one disk device for parity storage is added to a plurality of disk devices for data storage, and the parity is N + 1 in units of stripe groups. (N is 2 or more) By distributing the data to the disk devices, concentration of access to the parity disk can be prevented.
[0005]
[Problems to be solved by the invention]
However, in a normal RAID 5 disk array, a plurality of block data having consecutive logical addresses are sequentially arranged in a plurality of disk devices in units of stripe groups. The logical addresses of the block data between them are discontinuous. For this reason, when a host requests to read data across the stripe group boundary, two or more read requests are always issued to the disk array regardless of the requested data size. . This is a factor that degrades the data read performance of the RAID5 disk array. Therefore, in order to realize sufficient read performance, it is necessary to consider not only parallel accessibility to a plurality of disk devices but also continuity of logical addresses when viewed by individual disk devices.
[0006]
The present invention has been made in consideration of such circumstances, and a disk control system capable of sufficiently reducing the number of accesses to a disk device by optimizing the data arrangement relationship between stripe groups and its An object is to provide a control method.
[0007]
[Means for Solving the Problems]
In order to solve the above-described problems, the present invention provides a plurality of disks constituting a disk array in units of stripe groups each including a plurality of block data having consecutive logical addresses and redundant data for recovering a failure of the block data. In a disk control system for distributing and storing data in a device, the largest logical address among the block data belonging to the preceding stripe group among the stripe groups allocated adjacent to the plurality of disk devices. The data arrangement of each of the plurality of stripe groups is arranged so that the block data and the block data having the smallest logical address among the block data belonging to the following stripe group are physically stored in the same disk device. The means to decide A disk access request from the bets, based on the data arrangement the determined, Control means for controlling access to the plurality of disk devices, and when an access request from the host requests a read of data in a range across two adjacent stripe groups, the read is requested Control means for issuing a read request for reading the data requested to be read in one disk access to a disk device in which data is physically stored continuously It is characterized by comprising.
[0008]
In this disk control system, among the adjacently assigned stripe groups, the block data having the highest logical address among the block data belonging to the preceding stripe group and the block data belonging to the succeeding stripe group are the most logical. Since the data arrangement of each of the plurality of stripe groups is determined so that block data with a small address is physically continuously stored in the same disk device, it is logically continuous across the boundary of adjacent stripe groups. The two block data are stored physically continuously in the same disk device. Therefore, while maintaining parallel access performance, it is possible to guarantee the continuity of logical addresses between stripe groups when viewed with individual disk devices. When reading of data in a range that extends across the boundary between two adjacent stripe groups is requested by an access request from the host, the disk device in which the data requested to be read is physically stored continuously By issuing a read request for reading the data requested to be read in one disk access, The number of accesses to the disk device for data reading can be sufficiently reduced, and the reading performance can be improved.
[0009]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be described with reference to the drawings.
FIG. 1 shows the configuration of a disk control system according to an embodiment of the present invention. This disk control system is used by being mounted on a computer system such as a server computer, for example, and comprises a RAID controller 12 and a disk array 13 as shown. The RAID controller 12 controls a plurality of disk devices (DISK # 1 to DISK # 3) constituting the disk array 13 in response to a disk access request from the host 11. The host 11 includes a CPU of a computer system, an operating system executed by the CPU, a program such as an application, and the like.
[0010]
The RAID controller 12 groups a plurality of disk devices (DISK # 1 to DISK # 3) and realizes a single logical disk by the plurality of disk devices (DISK # 1 to DISK # 3). In other words, due to the function of the RAID controller 12, a plurality of disk devices (DISK # 1 to DISK # 3) appear to the host 11 as a single large capacity logical disk. Here, it is assumed that grouping of a plurality of disk devices (DISK # 1 to DISK # 3) is performed based on the RAID5 level.
[0011]
The RAID 5 disk array 13 includes N + 1 disk units in which one (1) redundant disk device for storing redundant data (parity) for failure recovery is added to N (N is 2 or more) disk devices for data storage. Consists of disk devices. In the figure, a case where N = 2, that is, a case where a RAID 5 disk array 13 is configured by three disk devices DISK # 1 to DISK # 3 is illustrated. Each of the disk devices (DISK # 1 to DISK # 3) includes a magnetic disk drive device (hard disk drive HDD) having an interface such as IDE or SCSI.
[0012]
The RAID controller 12 groups the disk devices DISK # 1 to DISK # 3 by striping with parity, and distributes and stores data and parity in the disk devices DISK # 1 to DISK # 3. Striping is realized by transversely arranging a plurality of stripe groups S1, S2, S3, S4,... With parity in the disk devices DISK # 1 to DISK # 3 as shown.
[0013]
Each of the stripe groups S1, S2, S3, S4,... For storing two block data having consecutive logical addresses in two different disk devices and for recovering the failure of these data stripes Redundant data (parity) for storing data in the remaining one disk device. The unit of data or parity assigned to each of the disk devices DISK # 1 to DISK # 3 is a stripe, and consists of two data stripes for data storage and one redundant stripe (parity stripe) for parity storage corresponding thereto. A set of stripes is a stripe group.
[0014]
The data arrangement of each of the plurality of stripe groups S1, S2, S3, S4,... Is determined and managed by the RAID controller 12. As a result, data having consecutive logical addresses can be distributed and stored in three disk devices DISK # 1 to DISK # 3 in units of stripe groups S1, S2, S3, S4,. Can be sequentially moved between the three disk devices DISK # 1 to DISK # 3 in units of stripe groups.
[0015]
Therefore, in the RAID 5 disk array 13, the unit in which the disk device to which the parity stripe is allocated is changed can be defined as a stripe group.
[0016]
As shown in the figure, the RAID controller 12 is provided with a control unit 121, a data arrangement management unit 122, and a data buffer 123. The control unit 121 processes an access request from the host 11 to the disk array 13 according to the data arrangement of each of the stripe groups S1, S2, S3, S4,... Determined by the data arrangement management unit 122. From the host 11, the disk array 13 appears as one logical disk drive. The control unit 121 determines which physical address of which disk device in the disk array 13 the logical address specified in the access request from the host 11 corresponds to, and makes a physical access request to each corresponding disk device. (Data read request / data write request) is issued. When the access request from the host 11 is a write, the already written data and parity are also read in order to recalculate a new parity. The data buffer 123 is used to temporarily hold write data from the host 11, temporarily hold data read from the individual disk devices DISK # 1 to DISK # 3, and the like.
[0017]
Further, the control unit 121 includes 1) a function of restoring data on a failed disk device using parity when a certain disk device fails, and 2) a case where the failed disk device is replaced with a new disk device. In addition, a function for restoring the contents to be stored in the new disk device using the data and parity of the remaining other disk devices is also provided.
[0018]
The data arrangement management unit 122 determines and manages the data arrangement of each of the plurality of stripe groups S1, S2, S3, S4,. This manages how block data and parity are arranged in the three disk devices DISK # 1 to DISK # 3. Specifically, the data arrangement management unit 122
1) An arithmetic expression for calculating the physical storage position (physical disk numbers # 1 to # 3 and the physical address on the disk device specified by the physical disk number) of the corresponding block data from the value of the logical address Or
2) a mapping table indicating the correspondence between logical addresses and physical storage locations (physical disk numbers # 1 to # 3 and physical addresses on the disk device specified by the physical disk numbers);
Etc.
[0019]
An example of the calculation formula (RAID block calculation formula) 1) used in this embodiment will be described. For simplicity, when the unit of block data is a sector, the row number (RowNum) corresponding to the sector number (SectorNum) to be accessed is determined from the sector number (SectorNum) and the number of HDDs (HDDNum) as follows. And column number (ColumnNum). Here, the row number (RowNum) indicates the number of the stripe group from the top, and the column number (ColumnNum) indicates the number of the disk device from the left. In addition, for each stripe group (row), a disk device (column number) to which block data having the smallest sector number in the stripe group is assigned is defined as a start column number (StartColumnNum).
[0020]
(1) When the disk is composed of an odd number
RowNum = (SectorNum-1) / (HDDNum-1) +1
StartColumnNum = (HDDNum− (RowNum% (HDDNum−1)))% HDDNum + 1
ColumnNum = (StartColumnNum + SectorNum + 1)% (HDDNum + 1)
Here,% indicates a remainder (mod). The column number of parity in each row is the column number next to the column number corresponding to the maximum sector number in the row (if the column number corresponding to the maximum sector number is equal to the number of HDD components, the parity column number is The number is 1).
[0021]
(2) When the disk is composed of an even number
RowNum = (SectorNum-1) / (HDDNum-1) +1
StartColumnNum = (HDDNum− (RowNum% (HDDNum−1)))% (HDDNum−2) +1
ColumnNum = (StartColumnNum + SectorNum + 1 + (RowNum / Constant))% (HDDNum + 1)
Constant: An appropriate constant is used.
[0022]
In the case of an even number, in the above formula, the parity is placed only in the even-numbered column. Therefore, by using an appropriate numerical value such as RowNum, a numerical value is added to ColumnNum so that the column in which parity is placed is not an even number.
[0023]
That is, in the present embodiment, basically, logical data is sequentially assigned to each of the stripe groups S1, S2, S3, S4,... So that a continuous logical address is assigned to each of the two data stripes in each stripe group. Assign an address. In this case, block data (data stripes) having consecutive logical addresses are the same between adjacent stripe groups, that is, between stripe groups S1 and S2, between S2 and S3, between S3 and S4, and so on. A data arrangement that is physically continuously stored in the disk device is used. An example of this data arrangement is shown in FIG.
[0024]
In FIG. 1, D1 to D8 indicate block data, respectively, and the order of subscripts 1 to 8 of D1 to D8 corresponds to the order of logical addresses (sector number in the above formula). P1 to P4 are parities generated from block data in the corresponding stripe groups S1 to S4.
[0025]
That is, in the first stripe group S1, data stripes corresponding to block data D1 and D2 having consecutive logical addresses are assigned to the disk devices DISK # 1 and # 2, respectively, and parity generated from the block data D1 and D2 is used. A parity stripe for storing P1 is assigned to the disk device DISK # 3. In the second stripe group S2, two block data D3 and D4 subsequent to the block data D1 and D2 of the stripe group S1 are stored. In this case, the most logical address in the preceding stripe group S1 is stored. The next block data D3 following the large block data D2 is physically and continuously stored in the same disk device as the disk device in which the block data D2 is stored (here, the disk device DISK # 2). That is, in the second stripe group S2, data stripes corresponding to the block data D3 and D4 are respectively assigned to the disk devices DISK # 2 and # 3, and the parity P2 generated from the block data D3 and D4 is stored. Therefore, a parity stripe is assigned to the disk device DISK # 1. Here, the reason why the data stripe corresponding to the block data D4 is allocated to the disk device DISK # 3 is to make the disk device storing the parity stripe different between the stripe groups S1 and S2.
[0026]
In the third stripe group S3, two block data D5 and D6 subsequent to the block data D3 and D4 of the stripe group S2 are stored. Also in this case, the next block data D5 following the block data D4 having the largest logical address in the preceding stripe group S2 is the same as the disk device (here, the disk device DISK # 3) in which the block data D4 is stored. It is physically stored in the disk device. That is, in the third stripe group S3, data stripes corresponding to the block data D5 and D6 are respectively assigned to the disk devices DISK # 3 and # 1, and the parity P3 generated from the block data D5 and D6 is stored. Therefore, a parity stripe is assigned to the disk device DISK # 2. As a result, between the stripe groups S2 and S3, similarly to the relationship between S1 and S2 described above, the block data having the largest logical address in the preceding stripe group and the most logical in the following stripe group. The block data having a small address is stored in the same disk device, and the disk devices storing the parity stripes have different relationships.
[0027]
Further, also in the fourth stripe group S4, the disk device (here, the data stripe of the next block data D7 following the block data D6 having the largest logical address in the preceding stripe group S3 is stored). The disk device DISK # 1) is physically and continuously arranged on the same disk device, and the data stripe corresponding to the block data D8 is arranged on the disk device DISK # 2 and is generated from the block data D7 and D8. Parity stripes for storing the parity P4 to be stored are arranged in the disk device DISK # 3.
[0028]
By adopting such a data arrangement, the portions across adjacent stripe groups, for example, block data D2 and D3, block data D4 and D5, and block data D6 and D7 are all stored in the same disk device. Therefore, logically continuous data is physically stored continuously on one physical disk. Therefore, even when data in a range that spans the boundary between adjacent stripe groups is requested by the host 11, the read requested data is stored in the same disk device. The data can be read only by issuing a request. This is shown in FIG.
[0029]
FIG. 2 shows a state of disk access processing corresponding to a case where reading of block data D1 to D5 is requested by a read request from the host 11. Among the block data D1 to D5, logically continuous block data D2 and D3 extending across the boundary between the stripe groups S1 and S2 are physically continuously stored in the disk device DISK # 2, and Since the logically continuous block data D4 and D5 straddling the boundary of S3 are physically continuously stored in the disk device DISK # 3, the block data D2 and D3 are stored in the disk device DISK # 2 once. It can be read out only by issuing a read request. Similarly, the block data D4 and D5 can be read out only by issuing a single read request to the disk device DISK # 3.
[0030]
That is, in this embodiment, when the host 11 requests reading of the block data D1 to D5, the following three read requests are generated, and disk access corresponding to each of the read requests is executed. Become.
[0031]
Read request # 1: Read request for reading block data D1 stored in the head data stripe from the disk device DISK # 1 (physical disk number = # 1, head address = head data stripe, data size = 1) block)
Read request # 2: Read request for reading the block data D2 and D3 respectively stored in the head and second data stripes from the disk device DISK # 2 (physical disk number = # 2, head address = head data) Stripe, data size = 2 blocks)
Read request # 3: Read request for reading block data D4 and D5 stored in the second and third data stripes from the disk device DISK # 3 (physical disk number = # 3, head address = 2) Data stripe, data size = 2 blocks)
Since these read requests # 1 to # 3 are issued in parallel to the disk devices DISK # 1 to # 3, a disk access for reading the block data D1 from the disk device DISK # 1 and a block from the disk device DISK # 2 are performed. Disk access for reading data D2 and D3 and disk access for reading block data D4 and D5 from disk device DISK # 3 are executed in parallel. In this case, the data read from the disk device DISK # 2 is in the logical address order (D2, D3), and the data read from the disk device DISK # 3 is also in the logical address order (D4, D5). A data string to be returned to the host 11 as read data by simply storing the data read from the disk devices DISK # 1 to # 3 in the data buffer 123 in the order of read requests # 1 to # 3. That is, block data D1 to D5 can be obtained.
[0032]
FIG. 3 shows a more specific example of the data arrangement of each of the stripe groups S1, S2, S3, S4,... Used in the present embodiment.
[0033]
As shown in the figure, in the stripe groups S1, S2, S3, S4,..., Each data block is composed of a plurality of sub-block data. Here, for simplicity, an example is shown in which one block data is composed of four sub-block data. The size of one sub-block data corresponds to, for example, a sector size or a cluster size. The host 11 can specify the logical address in units of sub-block data.
[0034]
In FIG. 3, d1 to d32 indicate subblock data, respectively, and the order of subscripts 1 to 32 of d1 to d32 corresponds to the order of logical addresses. For example, the block data D1 is composed of sub-block data d1 to d4, and these sub-block data d1 to d4 are physically stored in the head data stripe of the disk device DISK # 1. The block data D2 is composed of sub-block data d5 to d8, and these sub-block data d5 to d8 are physically continuously stored in the head data stripe of the disk device DISK # 2. In this case, the first parity stripe of the disk device DISK # 3 includes the parity p1 generated from the subblock data d1 and d5, the parity p2 generated from the subblock data d2 and d6, and the subblock data d3 and d7. And the parity p4 generated from the sub-block data d4 and d8 are stored as shown in the figure.
[0035]
The block data D3 in the stripe group S2 is composed of sub-block data d9 to d12, and these sub-block data d9 to d12 are physically continuously stored in the second data stripe of the disk device DISK # 2. ing. As a result, the sub-block data d5 to d12 having consecutive logical addresses are physically and continuously stored in the same disk device DISK # 2.
[0036]
Similarly, the sub-block data d13 to d16 constituting the block data D4 in the stripe group S2 and the sub-block data d17 to d20 constituting the block data D5 in the stripe group S3 are physically stored in the same disk device DISK # 3. The sub-block data d21 to d24 constituting the block data D6 in the stripe group S3 and the sub-block data d25 to d28 constituting the block data D7 in the stripe group S4 are stored in the same disk device. It is stored physically continuously in DISK # 1.
[0037]
Therefore, data in a range extending over the boundary between adjacent stripe groups, for example, data in the range of sub-block data d5 to d12, data in the range of sub-block data d13 to d20, and data in the range of sub-block data d21 to d28 All data can be read out only by issuing a single read request to the corresponding disk device. This is shown in FIG.
[0038]
FIG. 4 shows a state of read access processing when reading of the sub-block data d2 to d18 is requested by a read request from the host 11.
[0039]
Of the sub-block data d2 to d18, the sub-block data d5 to d12, which are data strings straddling the boundary between the stripe groups S1 and S2, are physically stored continuously in the disk device DISK # 2, and the stripe group S2 Since the sub-block data d13 to d18, which is a data string extending over the boundary between S3 and S3, are physically continuously stored in the disk device DISK # 3, the sub-block data d5 to d12 is once for the disk device # 2. Similarly, the sub-block data d13 to d18 can be read out only by issuing a single read request to the disk device DISK # 3. That is, in this embodiment, when the host 11 requests reading of d2 to d18, the following three read requests are generated, and disk access corresponding to each of the read requests is executed.
[0040]
Read request # 1: Read request for reading the second to fourth stored sub-block data d2 to d4 in the head data stripe from the disk device DISK # 1 (physical disk number = # 1, head Address = second subblock in the first data stripe, data size = 3 subblocks)
Read request # 2: Read request for reading the sub data blocks d5 to d8 and d9 to d12 respectively stored in the first and second data stripes from the disk device DISK # 2 (physical disk number = # 2, (Start address = start subblock in start data stripe, data size = 8 subblocks)
Read request # 3: Read request for reading the sub data blocks d13 to d16 and d17 to d18 respectively stored in the second and third data stripes from the disk device DISK # 3 (physical disk number = # 3 , Start address = second data stripe, data size = 6 sub-blocks)
Next, the procedure of disk access processing executed by the controller 121 of the RAID controller 12 will be described with reference to the flowchart of FIG.
[0041]
First, the control unit 121 receives a read request from the host 11 for the disk array 13 (step S101). This read request includes the head logical address and the data size, and this specifies the range to be read. The control unit 121 sets the read target range specified by the read request from the host 11 for each stripe group based on the data arrangement of the stripe groups S1, S2, S3, S4,. Then, a read request for the corresponding physical disk (disk device) is generated for each stripe (step S102). In this read request generation process, for each stripe group, the physical disk (disk device) storing the data in the read target range and its physical storage location are detected.
[0042]
For example, assuming the example shown in FIG. 2, the block data D1 and D2 belonging to the stripe group S1 in the read target ranges D1 to D5 are stored in the first data stripes of the disk devices DISK # 1 and # 2, respectively. Read request (A) for reading the block data D1 from the first data stripe of the disk device DISK # 1, and a read request for reading the block data D2 from the first data stripe of the disk device DISK # 2 ( B) is generated. Further, it is detected that the block data D3 and D4 belonging to the stripe group S2 are stored in the second data stripes of the disk devices DISK # 2 and # 3, respectively, and the second data stripe of the disk device DISK # 2 is detected. A read request (C) for reading the block data D3 from the disk data and a read request (D) for reading the block data D4 from the second data stripe of the disk device DISK # 3 are generated. Further, it is detected that the block data D5 belonging to the stripe group S3 is stored in the third data stripe of the disk device DISK # 3, and the block data D5 is read from the third data stripe of the disk device DISK # 3. A read request (E) is generated.
[0043]
Next, the control unit 121 logically continues each of the read target data from these generated read requests, and is physically continuously stored in the same physical disk (disk device). It is checked whether there is a combination of read requests for which the read data is to be read (step S103). When such a combination of read requests exists, the control unit 121 performs an integration process for combining the read requests into one, and stores the data to be accessed for each of the read requests in a single disk. A new read request for reading by access is generated (step S104). For example, considering the read requests (A to E) described above, a read request (B) for reading the block data D2 from the disk device DISK # 2 and a read request (for reading the block data D3 from the disk device DISK # 2) ( C) are combined into one read request, and a new read request (X) is generated for reading the data string (D2, D3) for two blocks from the head data stripe of the disk device DISK # 2. Also, the read request (D) for reading the block data D4 from the disk device DISK # 3 and the read request (E) for reading the block data D5 from the disk device DISK # 3 are combined into one read request. A new read request (Y) for reading a data string (D4, D5) for two blocks from the second data stripe of the device DISK # 3 is generated.
[0044]
Then, the control unit 121 issues each final read request obtained through the processing of steps S103 and S104 to the corresponding disk device, and starts read access (step S105). In the above example, there are three read requests issued: a read request (A), a read request (X), and a read request (Y). These read requests correspond to the read request # 1, the read request # 2, and the read request # 3 described in FIG.
[0045]
Thereafter, the control unit 121 waits for a response from each of the read-accessed disk devices, and after all the read accesses have been completed, the read data read by the read access is collected from the data buffer 123 to the host 11. Return (step S106). Since the read data is held in the data buffer 123 in the order of read request as described above, the requested data string can be returned to the host 11 without performing processing such as rearrangement.
[0046]
Note that the data arrangement of the present embodiment is applicable not only to a RAID 5 level in which a disk device storing parity is moved in units of stripe groups, but also to a RAID 4 level disk array in which the parity is fixedly arranged in a specific disk device. I can do it. An example of application to a RAID 4 level disk array will be described below as a second embodiment of the present invention.
[0047]
(Second Embodiment)
FIG. 6 shows an example of data arrangement when a RAID 4 level disk array is configured by three disk devices DISK # 1 to DISK # 3. The disk devices DISK # 1 and # 2 are data storage disks, and the disk device DISK # 3 is a parity storage disk.
[0048]
That is, in the leading stripe group S1, data stripes corresponding to block data D1, D2 having consecutive logical addresses are assigned to the disk devices DISK # 1, # 2, respectively, and parity generated from the block data D1, D2 is used. A parity stripe for storing P1 is assigned to the disk device DISK # 3. In the second stripe group S2, two block data D3 and D4 subsequent to the block data D1 and D2 of the stripe group S1 are stored. In this case, the most logical address in the preceding stripe group S1 is stored. The next block data D3 following the large block data D2 is physically continuously stored in the same disk device as the disk device in which the block data D2 is stored (here, the disk device DISK # 2). That is, in the second stripe group S2, data stripes corresponding to the block data D3 and D4 are respectively assigned to the disk devices DISK # 2 and # 1, and the parity P2 generated from the block data D3 and D4 is stored. The disk device to which the parity stripe is assigned is the disk device DISK # 3, as is the stripe group S1.
[0049]
In the third stripe group S3, two block data D5 and D6 subsequent to the block data D3 and D4 of the stripe group S2 are stored. Also in this case, the next block data D5 following the block data D4 having the largest logical address in the preceding stripe group S2 is the same as the disk device (here, the disk device DISK # 1) in which the block data D4 is stored. It is physically stored in the disk device. That is, in the third stripe group S3, the data stripes corresponding to the block data D5 and D6 are allocated to the disk devices DISK # 1 and # 2, respectively, and the parity P3 generated from the block data D5 and D6 is stored. The parity stripe for this is assigned to the disk device DISK # 3. Further, also in the fourth stripe group S4, the disk device (here, the data stripe of the next block data D7 following the block data D6 having the largest logical address in the preceding stripe group S3 is stored) The disk device DISK # 2) is physically and continuously arranged on the same disk device, and the data stripe corresponding to the block data D8 is arranged on the disk device DISK # 1, and from the block data D7 and D8, A parity stripe for storing the generated parity P4 is arranged in the disk device DISK # 3.
[0050]
By adopting such a data arrangement, as in the case of RAID 5 described with reference to FIG. 2, portions across adjacent stripe groups, for example, block data D2 and D3, block data D4 and D5, block data D6 and D7 Are stored in the same disk device, and logically continuous data is physically stored continuously on one physical disk.
[0051]
FIG. 7 shows a state of disk access processing corresponding to a case where reading of block data D1 to D5 is requested by a read request from the host 11. Among the block data D1 to D5, logically continuous block data D2 and D3 extending across the boundary between the stripe groups S1 and S2 are physically continuously stored in the disk device DISK # 2, and Since the logically continuous block data D4 and D5 straddling the boundary of S3 are physically continuously stored in the disk device DISK # 1, the block data D2 and D3 are stored once in the disk device DISK # 2. It can be read out only by issuing a read request. Similarly, the block data D4 and D5 can be read out only by issuing a single read request to the disk device DISK # 1. That is, in the second embodiment, when reading of the block data D1 to D5 is requested from the host 11, the following three read requests are generated, and disk access corresponding to each of the read requests is executed. It will be.
[0052]
Read request # 1: Read request for reading block data D1 stored in the head data stripe from the disk device DISK # 1 (physical disk number = # 1, head address = head data stripe, data size = 1) block)
Read request # 2: Read request for reading the block data D2 and D3 respectively stored in the head and second data stripes from the disk device DISK # 2 (physical disk number = # 2, head address = head data) Stripe, data size = 2 blocks)
Read request # 3: Read request for reading the block data D4 and D5 respectively stored in the second and third data stripes from the disk device DISK # 1 (physical disk number = # 1, head address = 2) Data stripe, data size = 2 blocks)
The read requests # 1 and # 2 are issued in parallel to the disk devices DISK # 1 and # 3, but the read request # 3 is issued after completion of the read access corresponding to the read request # 1. .
[0053]
FIG. 8 shows a more specific example of the data arrangement of each of the stripe groups S1, S2, S3, S4,... Corresponding to FIG.
[0054]
In FIG. 8, d1 to d32 indicate sub-block data, respectively, and the order of subscripts 1 to 32 of d1 to d32 corresponds to the order of logical addresses. For example, the block data D1 is composed of sub-block data d1 to d4, and these sub-block data d1 to d4 are physically stored in the head data stripe of the disk device DISK # 1. The block data D2 is composed of sub-block data d5 to d8, and these sub-block data d5 to d8 are physically continuously stored in the head data stripe of the disk device DISK # 2. In this case, the first parity stripe of the disk device DISK # 3 includes the parity p1 generated from the subblock data d1 and d5, the parity p2 generated from the subblock data d2 and d6, and the subblock data d3 and d7. And the parity p4 generated from the sub-block data d4 and d8 are stored as shown in the figure.
[0055]
The block data D3 in the stripe group S2 is composed of sub-block data d9 to d12, and these sub-block data d9 to d12 are physically continuously stored in the second data stripe of the disk device DISK # 2. ing. As a result, the sub-block data d5 to d12 having consecutive logical addresses are physically and continuously stored in the same disk device DISK # 2.
[0056]
Similarly, the sub-block data d13 to d16 constituting the block data D4 in the stripe group S2 and the sub-block data d17 to d20 constituting the block data D5 in the stripe group S3 are physically stored in the same disk device DISK # 1. The sub-block data d21 to d24 constituting the block data D6 in the stripe group S3 and the sub-block data d25 to d28 constituting the block data D7 in the stripe group S4 are stored in the same disk device. It is stored physically continuously in DISK # 2.
[0057]
FIG. 9 shows the state of read access processing when reading of the sub-block data d2 to d18 is requested by the read request from the host 11. Of the sub-block data d2 to d18, the sub-block data d5 to d12, which are data strings straddling the boundary between the stripe groups S1 and S2, are physically stored continuously in the disk device DISK # 2, and the stripe group S2 Since the sub-block data d13 to d18, which is a data string extending over the boundary between S3 and S3, are physically continuously stored in the disk device DISK # 1, the sub-block data d5 to d12 is stored in the disk device DISK # 2. Similarly, the sub-block data d13 to d18 can be read out only by issuing a single read request to the disk device DISK # 1. That is, in the second embodiment, when reading of d2 to d18 is requested from the host 11, the following three read requests are generated, and disk access corresponding to each of the read requests is executed. Become.
[0058]
Read request # 1: Read request for reading the second to fourth stored sub-block data d2 to d4 in the head data stripe from the disk device DISK # 1 (physical disk number = # 1, head Address = second subblock in the first data stripe, data size = 3 subblocks)
Read request # 2: Read request for reading the sub data blocks d5 to d8 and d9 to d12 respectively stored in the first and second data stripes from the disk device DISK # 2 (physical disk number = # 2, (Start address = start subblock in start data stripe, data size = 8 subblocks)
Read request # 3: Read request for reading the sub data blocks d13 to d16 and d17 to d18 respectively stored in the second and third data stripes from the disk device DISK # 1 (physical disk number = # 1) , Start address = second data stripe, data size = 6 sub-blocks)
The procedure of disk access processing executed by the controller 121 of the RAID controller 12 is the same as that described with reference to FIG.
[0059]
In each of the above embodiments, the description has been made on the premise of so-called hardware RAID in which a plurality of disk devices DISK # 1 to DISK # 3 are grouped by the RAID controller 12, but the data arrangement and read described in each embodiment. The processing for grouping the requests includes the functions of the control unit 121 and the data arrangement management unit 122 of the RAID controller 12 even for software RAID that groups a plurality of disk devices DISK # 1 to DISK # 3 in software by an operating system or the like. It can be applied by mounting on an operating system or the like.
[0060]
Further, the data arrangement described in each embodiment is effective not only at the time of data reading but also at the time of data writing in that the number of I / Os to the disk array 13 can be reduced.
[0061]
In addition, the present invention is not limited to the above-described embodiments, and various modifications can be made without departing from the scope of the invention in the implementation stage. Further, the above embodiments include inventions at various stages, and various inventions can be extracted by appropriately combining a plurality of disclosed constituent elements. For example, even if some constituent requirements are deleted from all the constituent requirements shown in the embodiment, the problem described in the column of the problem to be solved by the invention can be solved, and the effect described in the column of the effect of the invention Can be obtained as an invention.
[0062]
【The invention's effect】
As described above, according to the present invention, it is possible to sufficiently reduce the number of accesses to the disk device by optimizing the data arrangement relationship between stripe groups.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of a computer system using a disk control system according to a first embodiment of the present invention.
FIG. 2 is a view showing a state of data arrangement and disk access processing for a disk array in the same embodiment;
FIG. 3 is a view showing a specific example of data arrangement with respect to the disk array in the same embodiment;
4 is a diagram showing a state of disk access processing when the data arrangement of FIG. 3 is used.
FIG. 5 is an exemplary flowchart showing the procedure of disk access processing in the embodiment;
FIG. 6 is a diagram showing data arrangement for a disk array in a second embodiment of the present invention.
7 is a view showing a state of disk access processing corresponding to the data arrangement in FIG. 7;
FIG. 8 is a view showing a specific example of data arrangement with respect to the disk array in the second embodiment.
FIG. 9 is a view showing a state of disk access processing corresponding to the data arrangement of FIG. 8;
[Explanation of symbols]
11 ... Host
12 ... RAID controller
13 ... Disk array
121 ... Control unit
122... Data arrangement management unit
123: Data buffer
DISK # 1 to # 3 ... Disk device

Claims (10)

A disk control system in which data is distributed and stored in a plurality of disk devices constituting a disk array in units of stripe groups each including a plurality of block data having consecutive logical addresses and redundant data for recovering the failure of the block data. In
Among the stripe groups allocated adjacent to each other for the plurality of disk devices, the block data having the highest logical address among the block data belonging to the preceding stripe group and the block data belonging to the subsequent stripe group Means for determining the data arrangement of each of the plurality of stripe groups so that block data having the smallest logical address is physically continuously stored in the same disk device;
Control means for controlling access to the plurality of disk devices based on a disk access request from the host and the determined data arrangement, and a boundary between two stripe groups adjacent to each other by the access request from the host When data reading within a range is requested, the data requested to be read is accessed once for the disk device in which the data requested to be read is physically continuously stored. And a control means for issuing a read request for reading with the disk control system. The means for determining the data arrangement is:
Among the stripe groups allocated adjacent to each other, the block data having the largest logical address among the block data belonging to the preceding stripe group and the block data having the smallest logical address among the block data belonging to the following stripe group Are stored in the same disk device in a continuous manner and the disk device in which the redundant data is stored is moved to another disk device in units of the stripe group. 2. The disk control system according to claim 1, wherein the data arrangement is determined. Each block data is composed of a plurality of sub-block data having consecutive logical addresses, and the plurality of sub-block data constituting each block data is physically continuously stored in the same disk device. 2. The disk control system according to claim 1, wherein: The control means includes
When reading of a range of data across a plurality of adjacent stripe groups is requested by an access request from the host, the range specified by the read request is divided into stripe groups, and the access target is set for each stripe group. Means for generating a read request to be issued to each disk device;
When the logical address is continuous and the data stored physically continuously in the same disk device is detected, the generated data is read out from the disk device with one disk access. Means for integrating read requests relating to data in which each of the access target data is logically continuous among the read requests and physically stored in the same disk device into a single read request The disk control system according to claim 1, further comprising: In a disk control system that controls a disk array in response to an access request from a host,
N + 1 (N ≧ 2) disk devices constituting the disk array;
N data stripes for storing data having consecutive logical addresses in N disk devices different in block units and redundant data corresponding to the contents of the N data stripes are stored in one disk device. A plurality of stripe groups each composed of redundant stripes for the above are allocated to the N + 1 disk units, and among the stripe groups allocated adjacent to each other, the block data belonging to the preceding stripe group is the most. Among data stripes for storing block data with a large logical address and block data belonging to the following stripe group The data stripes for storing the block data with the smallest logical address are physically allocated continuously to the same disk unit, and the disk units to which the redundant stripes are allocated are the N + 1 units in units of the stripe group. Means for determining a data arrangement of each of the plurality of stripe groups allocated to the N + 1 disk devices so as to be circulated between the disk devices;
Control means for controlling access to the disk array on the basis of a disk access request from the host and the determined data arrangement, and a range spanning the boundary between two adjacent stripe groups by the access request from the host To read the data requested to be read in one disk access to the disk device in which the data requested to be read is physically continuously stored. And a control means for issuing a read request for the disk control system. Control for distributing and storing data in a plurality of disk units constituting a disk array in units of stripe groups each including a plurality of block data having consecutive logical addresses and redundant data for recovering the failure of those block data A method,
Among the stripe groups allocated adjacent to each other for the plurality of disk devices, the block data having the highest logical address among the block data belonging to the preceding stripe group and the block data belonging to the subsequent stripe group Determining the data arrangement of each of the plurality of stripe groups so that the block data with the smallest logical address is physically continuously stored in the same disk device;
A step of controlling access to the plurality of disk devices based on a disk access request from a host and the determined data arrangement, wherein the access request from the host causes a boundary between two stripe groups to be adjacent to each other; When reading of data in a spanning range is requested, the data requested to be read is transferred to the disk device in which the requested data is physically continuously stored by one disk access. And a step of issuing a read request for reading. Determining the data placement comprises:
Among the stripe groups allocated adjacent to each other, the block data having the largest logical address among the block data belonging to the preceding stripe group and the block data having the smallest logical address among the block data belonging to the following stripe group Are stored in the same disk device in a continuous manner and the disk device in which the redundant data is stored is moved to another disk device in units of the stripe group. The control method according to claim 6, wherein the data arrangement is determined. Each block data is composed of a plurality of sub-block data having consecutive logical addresses, and the plurality of sub-block data constituting each block data is physically continuously stored in the same disk device. The control method according to claim 6. Controlling the access comprises:
When reading of a range of data across a plurality of adjacent stripe groups is requested by an access request from the host, the range specified by the read request is divided for each stripe group, and the access target is set for each stripe group. Generating a read request to be issued to each disk device;
When the logical address is continuous and the data stored physically continuously in the same disk device is detected, the generated data is read out from the disk device with one disk access. In the read request, each data to be accessed is logically continuous and is physically stored in the same disk device. 7. The control method according to claim 6, further comprising the step of integrating related read requests into a single read request. A control method for controlling a disk array including N + 1 (N ≧ 2) disk devices in response to an access request from a host,
N data stripes for storing data having consecutive logical addresses in N disk devices different in block units and redundant data corresponding to the contents of the N data stripes are stored in one disk device. A plurality of stripe groups each composed of redundant stripes for the above are allocated to the N + 1 disk units, and among the stripe groups allocated adjacent to each other, the block data belonging to the preceding stripe group is the most. A data stripe for storing block data having a large logical address and a data stripe for storing block data having the smallest logical address among the block data belonging to the following stripe group are physically stored in the same disk device. Continuous Each of the plurality of stripe groups assigned to the N + 1 disk devices is circulated between the N + 1 disk devices in units of the stripe group. Determining data placement;
A step of controlling access to the disk array based on a disk access request from the host and the determined data arrangement, wherein a range across a boundary between two adjacent stripe groups by the access request from the host; When a data read is requested, the data requested to be read is read in one disk access to the disk device in which the data requested to be read is physically continuously stored. And a step of issuing a read request.

Images