JP2017054303A - Disk array device, disk array system, control method, and control program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To achieve an efficiency included by a read-modify-write without initialization of the entire stripes in either of the stripe not in the written state or the stripe in the already written state.SOLUTION: A disk array device includes: a plurality of disk devices containing a data stripe which is a stripe for writing data; a disk device containing a parity stripe being a stripe for writing a parity to a plurality of data stripes; and control means for, when data is not written in a stripe to be updated and the parity stripe, obtaining a writing command for any of the plurality of data stripes, correcting the predetermined data by data to be updated, writing the data after correction in a stripe to be updated, and generating a parity based on the predetermined parity, the predetermined data, and the data to be updated, and writing the generated parity in a parity stripe.SELECTED DRAWING: Figure 6

Description

  The present invention relates to a disk array device, a disk array system, a control method, and a control program, and more particularly to a disk array device, a disk array system, a control method, and a control program that store data and its parity.

  Disk devices such as HDDs (Hardware Disk Drives) and SSDs (Solid State Drives) used in the server main unit are built and operated with RAID (Redundant Arrays of Independent Disks) to ensure the safety of data against failures. Often done.

  When a large number of disk devices are connected, a disk array device having a RAID 5 or RAID 6 configuration uses a technique called Read Modify Write (hereinafter RMW) in order to speed up access processing.

  Although there are several types of RAID, RAID 5 will be described as an example. When writing data using RAID 5, the controller of the disk array device divides the data, generates parity from the divided data, and writes these to each disk device. Data is written to the disk unit with a specific size. This writing unit is called a stripe.

  A RAID 5 disk array device composed of n (n is an integer of 3 or more) disk devices writes data to n−1 disk devices and writes parity to the remaining one disk device.

  FIG. 1 is a conceptual diagram of a RAID 5 disk array device (inside the broken line) composed of three disk devices. A controller (not shown) of the disk array device distributes and writes data to the disk device 1 and the disk device 2. The controller divides the data into data 1 and data 2 and writes data 1 to the disk device 1 and data 2 to the disk device 2. Further, the controller writes the parity generated by calculation from data 1 and data 2 to the disk device 3.

  Data 1, data 2, and parity can be restored from the other two even if one of them is lost due to a failure of the disk device.

  FIG. 2 is a conceptual diagram of a RAID 5 disk array device (inside the broken line) composed of four disk devices. A controller (not shown) of the disk array device writes data to three units (disk devices 1 to 3) and writes parity to the remaining unit (disk device 4).

  Here, when overwriting only the data 1 of the disk device 1, the controller calculates the parity from the data to be overwritten on the disk device 1 and the data 2 and data 3 already written in the disk devices 2 and 3. The parity is written to the disk device 4. In this case, the controller executes the following four readings and writings in total.

  A) Write data 1 to the disk unit 1.

  B) Data 2 of the disk device 2 is read.

  C) Data 3 of the disk device 3 is read.

  D) Parity is newly generated from the read data 2 and 3 and the written data 1 and written to the disk device 4.

  Next, RMW will be described. FIG. 3 is a conceptual diagram for explaining the RMW in a RAID 5 disk array device composed of four disk devices. Here, the parity generated from the data 1, data 2, and data 3 stored in the three disk devices (disk devices 1 to 3) is written in the remaining one disk device (disk device 4). And Then, it is assumed that only data 1 is overwritten.

  Here, when executing RMW, a controller performs the following procedures.

  A) Read data 1 of the disk unit 1.

  B) Read the parity of the disk unit 4.

  C) Write data to the disk unit 1.

  D) A new parity is generated from the read data 1, parity, and data to be written to the disk device 1 and written to the disk device 4.

  When the parity is generated from the stored data (data 1 to 3), the controller generates the parity without reading the data 2 and the data 3. When RMW is executed with four disk devices, reading and writing are performed four times. This number of times is the same as the number of times when parity is generated without executing RMW.

  FIG. 4 is a conceptual diagram of a RAID 5 disk array device (inside the broken line) composed of five disk devices. Here, the parity generated from the data 1, data 2, data 3, and data 4 stored in the four disk devices (disk devices 1 to 4) is written to the remaining one disk device (disk device 5). Suppose you are. Then, it is assumed that only data 1 is overwritten.

  When only the data 1 of the disk device 1 is overwritten without using the RMW, the controller performs parity from the data to be overwritten on the disk device 1 and the data 2, data 3, and data 4 written in the disk devices 2 to 4. And the parity is written to the disk device 5. In this case, the controller executes the following five readings and writings in total.

  A) Write data 1 to the disk unit 1.

  B) Data 2 of the disk device 2 is read.

  C) Data 3 of the disk device 3 is read.

  D) Data 4 of the disk device 4 is read.

  D) A new parity is generated from the read data 2, 3, 4 and the written data, and is written to the disk device 5.

  FIG. 5 is a conceptual diagram for explaining the RMW in a RAID 5 disk array device composed of five disk devices. Here, the parity generated from the data 1, data 2, data 3, and data 4 stored in the four disk devices (disk devices 1 to 4) is written to the remaining one disk device (disk device 5). Suppose you are. Then, it is assumed that only data 1 is overwritten.

  Here, when executing RMW, a controller performs the following procedures.

  A) Read data 1 of the disk unit 1.

  B) Read the parity of the disk unit 4.

  C) Write data to the disk unit 1.

  D) A new parity is generated from the read data 1, parity, and data to be written to the disk device 1, and is written to the disk device 5.

  When the parity is generated from the data (data 1 to 4) stored in this manner, the controller generates the parity without reading the data 2, the data 3, and the data 4. When RMW is executed with five disk devices, reading and writing are performed four times. This number is one less than when parity is generated without executing RMW.

  Thus, when there are five or more disk devices and there are few places to be written, the use of RMW may improve the input / output efficiency.

  As described above, when the controller uses RMW, the number of times of reading and writing does not change even if the number of disk devices constituting the RAID increases. For this reason, when the number of disk devices constituting a RAID is 5 or more, it may be more efficient to use RMW for writing. In that case, the controller uses RMW.

  However, in order for the disk array device to use RMW, it is necessary to create parity based on the data written in the disk device. That is, the consistency of RAID needs to match. For this reason, the disk array device performs the entire initialization work at the time of system setup, and sets all data and parity to a specific value, for example, 0, thereby matching consistency. However, in recent years, with the increase in capacity of the disk device, there is a problem that this initialization time increases and it takes time for system setup.

  There is also a disk array device that performs initialization work in the background during operation. This operation is called Back Ground Initialization (hereinafter BGI). In this case, when the BGI is executed, the RMW cannot be used until the completion, and there is a problem that the system performance is remarkably deteriorated by the BGI processing.

  Patent Document 1 discloses a disk array device that executes RMW while performing stripe formatting processing in the disk device. When there is a write request during the formatting process, this apparatus performs writing by RMW if the target stripe column has been formatted, but performs writing in units of format columns if it has not been formatted. Here, the stripe column is a group of stripes including a stripe for writing data on a plurality of disk devices and a stripe for writing parity on another disk device.

  Patent Document 2 discloses a disk array device in which a parity block includes an identifier indicating a position in a specific parity group protected by parity data. This disk array device uses this information to reduce the number of disk accesses during writing.

JP 2003-241904 A JP 07-306759 A

  The disk array device of Patent Document 1 writes data to all stripes in a stripe column when data is to be written to any stripe in a non-formatted stripe column. That is, in this case, the disk array device must write to all the disk devices constituting the RAID in the device, and cannot enjoy the efficiency of the RMW.

  An object of the present invention is to provide a disk array device, a disk array system, a control method, and a control program that solve the above-described problems.

  The control device according to one embodiment of the present invention includes a plurality of disk devices that include data stripes that are data write stripes, and a disk device that includes parity stripes that are parity write stripes for the plurality of data stripes; When a write command to one of the plurality of data stripes is acquired and no data is written in the update target stripe and the parity stripe, the predetermined data is corrected with the update target data, and the corrected data is updated with the update target Control means for writing to the stripe, generating parity based on the predetermined parity, the predetermined data, and the update target data, and writing the generated parity to the parity stripe.

  The method according to an embodiment of the present invention includes a plurality of disk devices that include data stripes that are data write stripes, and a plurality of disk devices that include parity stripes that are parity write stripes for the plurality of data stripes. If a write command to one of the data stripes is acquired and no data is written to the update target stripe and the parity stripe, the predetermined data is corrected with the update target data, and the corrected data is changed to the update target stripe. Writing, parity is generated based on the predetermined parity, the predetermined data, and the update target data, and the generated parity is written to the parity stripe.

  A program according to an embodiment of the present invention is a computer including a plurality of disk devices that include data stripes that are data write stripes and a disk device that includes parity stripes that are parity write stripes for the plurality of data stripes. If a write command to any of the plurality of data stripes is acquired and no data is written in the update target stripe and the parity stripe, the predetermined data is corrected with the update target data, and the corrected data is Write to the update target stripe, generate parity based on the predetermined parity, the predetermined data, and the update target data, and execute a process of writing the generated parity to the parity stripe.

  The disk array device according to the present invention achieves the efficiency of the RMW for writing to a stripe that has not been written and writing to a stripe that has already been written without initializing all stripes.

FIG. 1 is a conceptual diagram of a RAID 5 disk array device (inside the broken line) composed of three disk devices. FIG. 2 is a conceptual diagram of a RAID 5 disk array device (inside the broken line) composed of four disk devices. FIG. 3 is a conceptual diagram for explaining the RMW in a RAID 5 disk array device composed of four disk devices. FIG. 4 is a conceptual diagram of a RAID 5 disk array device (inside the broken line) composed of five disk devices. FIG. 5 is a conceptual diagram for explaining the RMW in a RAID 5 disk array device composed of five disk devices. FIG. 6 is a configuration diagram of the disk array system 80 according to the first embodiment. FIG. 7 is a conceptual diagram illustrating a stripe row. FIG. 8 is a diagram showing information stored in the disk device 63. FIG. 9 is a configuration diagram of the computer device 40. FIG. 10 is a conceptual diagram for explaining an outline of the operation of the control unit 62. FIG. 11 is an operation flowchart when the control unit 62 writes data. FIG. 12 is a configuration diagram of a disk array device 60 according to the second exemplary embodiment.

<First Embodiment>
<Configuration>
FIG. 6 is a configuration diagram of the disk array system 80 according to the present embodiment. The disk array system 80 includes a disk array device 60 and a host device 70 connected to each other. The disk array system 80 may include a plurality of disk array devices 60 and a plurality of host devices 70. Further, the disk array device 60 may incorporate the host device 70 as a component of the device.

  The host device 70 is, for example, an information processing device, and writes and reads data to and from the disk array device 60.

  The disk array device 60 includes a control unit 62 and n (n is an integer of 3 or more) disk devices 63.

  The control unit 62 receives an input / output command from the host device 70 and executes reading / writing on the disk device 63. In addition, the control unit 62 controls the n disk devices 63 so as to configure RAID, for example, RAID 5 and 6.

  Each disk device 63 includes one or more stripes 64 of a predetermined size, which is a unit for writing and reading. The stripes 64 of each disk device 63 constitute a stripe row. FIG. 7 is a conceptual diagram illustrating a stripe row. In FIG. 7, each cell in the table represents a stripe 64, each row represents a stripe column, and each column represents a disk device 63.

  A stripe column is a group of stripes 64. More specifically, the stripe row includes data write stripes 64 (hereinafter referred to as data stripes 64) on n-1 disk devices 63 and parity write stripes 64 (hereinafter referred to as data stripes 64) on another disk device 63. , Parity stripe 64). Here, the parity is redundant data created from data stored in n−1 data stripes 64.

  When the disk array device 60 includes a plurality of stripe columns, the parity stripe 64 of each stripe column may be stored in the same disk device 63, for example, the disk device n, as shown in FIG. It may be stored in a different disk device 63. In the example of FIG. 3, the parity stripe 64 of the stripe column 1 is stored in the disk device n, but the parity stripe 64 of the stripe column 2 is stored in the disk device n-1.

  FIG. 8 is a diagram showing information stored in the disk device 63. Each disk device 63 stores a flag 65 in association with each stripe 64 and each stripe 64. For example, the flag 65 is turned on when the stripe 64 has never been written, and turned off when the stripe 64 has been written. The flag 65 may be read into the internal memory of the disk array device 60 when the disk array device 60 is started up or when the disk device 63 is first accessed.

  The control unit 62 of the disk array device 60 can be realized by a logic circuit including firmware. The disk device 63 can be realized by an HDD or an SSD.

  Further, the disk array device 60 can be realized by the computer device 40 having the program 43.

  FIG. 9 is a configuration diagram of the computer device 40. The computer device 40 includes a processor 41, a main storage unit 42, and an auxiliary storage device 44 that are connected to each other via a bus 45. Here, for example, the main storage unit 42 is a semiconductor memory, and the auxiliary storage device 44 is an HDD or an SDD. The main storage unit 42 stores a program 43.

  The program 43 is executed by the processor 41, thereby causing the processor 41 to function as the control unit 62. The auxiliary storage device 44 can be used as the disk device 63 as it is.

  When the disk array device 60 is realized by the computer device 40, the processor 41 may execute a business processing program that performs input / output to / from the disk device 63. In this case, the disk array device 60 incorporates the host device 70 in the device.

<Operation>
FIG. 10 is a conceptual diagram for explaining an outline of the operation of the control unit 62. When the controller 62 determines that it is efficient to execute the RMW, the controller 62 reads the flag 65 associated with the data stripe 64 to be read when reading data on the disk device 63. If the flag 65 is off, the control unit 62 reads data actually written in the disk device 63. However, if the flag 65 is on, the control unit 62 considers that 0 is written without actually reading the data written in the disk device 63 (A in the figure)). Further, the control unit 62 reads the flag 65 associated with the parity stripe 64 of the stripe column to which the data stripe 64 to be read belongs. If the flag 65 is off, the control unit 62 reads the parity actually written in the disk device 63. However, if the flag 65 is on, the control unit 62 considers that 0 has been written without actually reading the parity written in the disk device 63 (A in the figure)).

  When writing data to the disk device 63, the control unit 62 turns it off if the flag 65 associated with the written data stripe 64 is on (c in the figure). Further, when the parity is written in the disk device 63, the control unit 62 turns it off if the flag 65 associated with the written parity stripe 64 is on (D in the figure).

  The flag 65 associated with the data stripe 64 once written with data remains off thereafter. The flag 65 is reset to ON when the disk array device 60 is reset or when there is an instruction from the host device 70.

  By this method, the control unit 62 can perform the efficient data update equivalent to the RMW even for the stripe 64 to which writing is performed for the first time, without performing the entire surface initialization or the BGI.

  Note that due to a failure such as a sudden power interruption, the stripe 64 and the flag 65 may not be updated, and the stripe 64 and the flag 65 may be inconsistent. This is a case where the work is interrupted without completing a series of work from a) to d) in the figure. As a countermeasure, the control unit 62 starts the work from a) to d) again after the trouble is removed and the power supply becomes normal, and performs the work up to d).

  FIG. 11 is an operation flowchart when the control unit 62 writes data. This operation is started when the host device 70 issues a write command to the disk array device 60. The write command of the host device 70 designates the address of the data stripe 64 to be written and the write data.

  First, the control unit 62 receives a write command from the host device 70 (S1), and determines whether it is more efficient to use the RMW (S2). The control unit 62 may make this determination by a known method.

  When it is determined that it is more efficient not to use the RMW (NO in S2), the control unit 62 performs normal writing (S11). However, the controller 62 checks and updates the flag 65 after writing (S12). That is, when the flag 65 of the stripe 64 to which writing has been performed is on, the control unit 62 updates it to off.

  If it is determined that using RMW is more efficient (YES in S2), the control unit 62 checks the flag 65 of the data stripe 64 to be updated and the flag 65 of the parity stripe 64 (S5). That is, the control unit 62 reads the flag 65 associated with the write target data stripe 64 specified by the write command, and the flag 65 associated with the parity stripe 64 belonging to the same stripe column as the data stripe 64. Inspect.

  When both the flags 65 are OFF (OFF / OFF in S5), the control unit 62 executes RMW. The controller 62 may execute this RMW by a known method.

  That is, the control unit 62 first reads information written in the disk device 63 from both the data stripe 64 and the parity stripe 64 (S9). Then, the control unit 62 generates update data and update parity from the read data, the read parity, and the write data specified by the write command (S10). Thereafter, the control unit 62 writes the generated update data in the data stripe 64 and the update parity in the parity stripe 64 (S5).

  After executing the RMW, the control unit 62 turns off the flag 65 of the written data stripe 64 and parity stripe 64 (S6). That is, when the flag 65 of the stripe 64 to which writing has been performed is on, the control unit 62 updates it to off. Here, the control unit 62 ends the processing of the received write command.

  When both flags 65 are on (on / on in S5), the control unit 62 does not read the information written in the disk device 63 from either the data stripe 64 or the parity stripe 64. . The control unit 62 generates update data and update parity from the predetermined data, the predetermined parity, and the write data designated by the write command (S10), and turns off the above-described write (S5) and flag 65 (S6). To finish the process.

  The predetermined data and the predetermined parity are, for example, zero value data having the size of the stripe 64. The predetermined data and the predetermined parity are not limited to these values as long as the data and the parity are consistent. That is, it is only necessary that the value of the predetermined parity be the same as the parity generated when the predetermined data is stored in each data stripe 64 of the stripe column.

  When the flag 65 of the data stripe 64 is ON and the flag 65 of the parity stripe 64 is OFF (ON / OFF in S5), the control unit 62 reads the parity from the parity stripe 64 (S7). On the other hand, the control unit 62 does not read information written in the disk device 63 from the data stripe 64. This is because data has been written to the other data stripes 64 in the stripe column but has not yet been written to the data stripe 64 to be written this time.

  Thereafter, the control unit 62 generates update data and update parity from the predetermined data, the read parity, and the write data designated by the write command (S10), and the process of turning off the writing (S5) and the flag 65 described above. (S6) is performed and the process is terminated.

<Effect>
In the disk array device 60 according to the present embodiment, the RMW has both the write to the stripe 64 that has not been written and the write to the stripe 64 that has already been written without performing initialization of all the stripes 64. Achieve efficiency. That is, the control unit 62 can omit reading of the data stripe 64 that is not an update target.

  Therefore, the disk array device 60 according to the present embodiment can greatly reduce the time spent for system setup without performing BGI.

  The reason is that the control unit 62 operates as if the predetermined data and the predetermined parity consistent with the predetermined data are stored in the unwritten stripe 64.

<Second Embodiment>
FIG. 12 is a configuration diagram of the disk array device 60 according to the present embodiment.

  The disk array device 60 includes a plurality of disk devices 63 including data stripes 64 that are data write stripes 64. The disk array device 60 also includes a disk device 63 including a parity stripe 64 that is a parity writing stripe 64 for the plurality of data stripes 64.

  Furthermore, the disk array device 60 also includes a control unit 62. The control unit 62 acquires a write command to any of the plurality of data stripes 64, and corrects the predetermined data with the update target data if no data is written in the update target data stripe 64 and the parity stripe 64. Thereafter, the control unit 62 writes the corrected data to the update target data stripe 64, generates parity based on the predetermined parity, the predetermined data, and the update target data, and writes the generated parity to the parity stripe 64.

  In the disk array device 60 according to the present embodiment, the RMW has both the write to the stripe 64 that has not been written and the write to the stripe 64 that has already been written without performing initialization of all the stripes 64. Achieve efficiency. That is, the control unit 62 can omit reading of the data stripe 64 that is not an update target.

  Therefore, the disk array device 60 according to the present embodiment can greatly reduce the time spent for system setup without performing BGI.

  The reason is that the control unit 62 operates as if the predetermined data and the predetermined parity consistent with the predetermined data are stored in the unwritten stripe 64.

  While the present invention has been described with reference to the embodiments, the present invention is not limited to the above embodiments. Various changes that can be understood by those skilled in the art can be made to the configuration and details of the present invention within the scope of the present invention.

40 Computer Device 41 Processor 42 Main Storage Unit 43 Program 44 Auxiliary Storage Device 45 Bus 60 Disk Array Device 62 Control Unit 63 Disk Device 64 Stripe 64 Data Stripe 64 Parity Stripe 65 Flag 70 Host Device 80 Disk Array System

Claims (10)

A plurality of disk devices including a data stripe that is a data writing stripe;
A disk device including a parity stripe that is a stripe for parity writing with respect to the plurality of data stripes;
When a write command to one of the plurality of data stripes is acquired and no data is written in the update target stripe and the parity stripe, the predetermined data is corrected with the update target data, and the corrected data is updated with the update target A disk array device comprising: control means for writing to the stripe, generating parity based on the predetermined parity, the predetermined data, and the update target data, and writing the generated parity to the parity stripe.   If no data is written in the update target stripe and parity is written in the parity stripe, the control unit corrects predetermined data with the update target data and converts the corrected data into the update target stripe. The disk array device according to claim 1, wherein the disk array device is written and parity is read from the parity stripe, parity is generated based on the read parity, the predetermined data, and update target data, and the generated parity is written to the parity stripe. Each disk unit includes a flag indicating whether or not there is a write associated with a stripe in its own unit,
3. The disk array device according to claim 1, wherein the control unit determines whether the stripe is written by checking the flag associated with the write target stripe. 4.   The disk array device according to any one of claims 1 to 3, wherein the predetermined data and the predetermined parity data are zero data. The disk array device according to any one of claims 1 to 4, and
A disk array system including a host processor connected to the disk array device and reading / writing data from / to the disk array device. A write command to any of the plurality of data stripes in a plurality of disk devices including a data stripe that is a data writing stripe and a disk device including a parity stripe that is a parity writing stripe for the plurality of data stripes. Acquired,
If no data is written in the update target stripe and the parity stripe, the predetermined data is corrected with the update target data, the corrected data is written in the update target stripe, and the predetermined parity, the predetermined data, and the update target are written. A method of generating parity based on data and writing the generated parity in the parity stripe.   If no data is written in the update target stripe and parity is written in the parity stripe, the predetermined data is corrected with the update target data, the corrected data is written in the update target stripe, and The method according to claim 6, wherein parity is read from a parity stripe, parity is generated based on the read parity, the predetermined data, and update target data, and the generated parity is written to the parity stripe. Each disk unit includes a flag indicating whether or not there is a write associated with a stripe in its own unit,
The method according to any one of claims 6 to 7, wherein the presence or absence of writing of the stripe is determined by checking the flag associated with the write target stripe. A plurality of disk devices including a data stripe that is a data writing stripe;
In a computer comprising a disk device including a parity stripe that is a parity write stripe for the plurality of data stripes,
When a write command to one of the plurality of data stripes is acquired and no data is written in the update target stripe and the parity stripe, the predetermined data is corrected with the update target data, and the corrected data is updated with the update target A program that writes to a stripe, generates parity based on predetermined parity, the predetermined data, and the update target data, and executes processing for writing the generated parity to the parity stripe.   If no data is written in the update target stripe and parity is written in the parity stripe, the computer corrects the predetermined data with the update target data and writes the corrected data in the update target stripe. The program according to claim 9, wherein parity is read from the parity stripe, parity is generated based on the read parity, the predetermined data, and update target data, and a process of writing the generated parity to the parity stripe is executed.

Images