KR102389929B1 - Storage Device Based on RAID - Google Patents

Abstract

The present disclosure discloses a RAID-based storage device.
The present embodiment includes a plurality of disks, wherein each of the plurality of disks is divided into a plurality of local groups, a local parity exists for each of the plurality of local groups, and a global parity exists for each stripe of each of the plurality of disks. It provides a RAID-based storage device, characterized in that.

Description

Storage Device Based on RAID}

This embodiment relates to a RAID-based storage device.

The content described below merely provides background information related to the present embodiment and does not constitute the prior art.

In general, the widely used RAID (Redundant Array of Independent Disks or Redundant Array of Inexpensive Disks) 6 technology reads data to all other disks in order to restore data when a disk error occurs. However, the RAID 6 technology has a problem in that data read performance is reduced by half due to an operation for data restoration.

As a technique for improving data read performance degradation when data is restored in RAID 6, there is a Local Reconstruction Code (LRC) technique. The LRC method is a technology for distributed storage. When data is restored using local parity (P parity) and global parity (Q parity), data read from disks divided into local groups is performed for fast It is a technology that supports data restoration.

Parity arrangement information (data layout) of the RAID 6 and LRC methods is as shown in FIG. 1A . The six 'D data' of RAID 6 shown in FIG. 1A are protected with 'P parity (local parity)' and 'Q parity (global parity)'. In other words, 'parity' in RAID means temporary information for data restoration. 'Parity' allows data to be read from each individual disk, and multiple data reads or writes are possible at the same time. In the case of RAID 6, two parities (P parity, Q parity) are used as an expansion concept of RAID 5.

On the other hand, in the LRC scheme, as shown in FIG. 1A, 'P parity (local parity)' is divided into two groups. That is, the three ‘D1 data’ are protected with ‘P1 parity (local parity)’, and the remaining three ‘D2 data’ are protected with ‘P2 parity (local parity)’. The six ‘D1, D2 data’ are protected with ‘Q parity (global parity)’. The number of 'P parity (local parity)' and 'Q parity (global parity)' may be set to one or more depending on data restoration performance and data reliability.

As shown in FIG. 1B , when an error occurs in the first disk, the number of data read operations generated to restore the corresponding data when a data read request is made to the corresponding disk is indicated. RAID 6 reads a total of 6 data. On the other hand, the LRC method performs restoration by reading only three pieces of data in one group.

However, in a RAID array in which I/O performance of all disks is the same, data is equally distributed among all data disks for fast I/O performance. In other words, all data disks receive the same number of I/O requests on average, and in this case, the above feature does not help with fast data recovery.

For example, in RAID 6 of FIG. 1C, due to an error occurring in the data disk, data read is performed twice as much as in normal for all data disks. Due to an error on the data disk, the data read performance of the overall RAID is reduced by half. For the sake of understanding, a random read is assumed. It is assumed that merging of data read requests does not occur in random read.

In the LRC method, due to an error occurring in the data disk, only the two data disks of group 1 perform twice as much data reading compared to normal. In the LRC method, the group 2 data disk is the same as the normal situation. However, a bottleneck occurs in the two data disks of group 1, and the utilization of the three data disks of group 2 is reduced to half. The data read performance of the general RAID of the LRC method is reduced by half. As a result, the LRC method also has the same bottleneck problem as RAID 6.

Conventional storage devices use data protection techniques to prevent data loss. RAID technology that utilizes additional parity disks used by common storage devices is a widely used data protection technique. It is divided into a plurality of RAID technologies according to the number of parities and parity arrangement information (data layout), and each has different characteristics.

In 'RAID 6', which is generally used, if a disk error occurs, data read performance is reduced by half. The LRC method for improving data read performance in ‘RAID 6’ also does not improve data restoration performance in a RAID array where the I/O performance of all disks is the same. In other words, the LRC method has a problem in that data restoration performance is not improved because a bottleneck occurs by using only some disks for data restoration when a disk error occurs due to the fixed local group determination method.

In this embodiment, in order to prevent sudden data I/O performance degradation (bottleneck) that occurs when a disk error occurs when data is read or written on a RAID array, data of the same local group is transferred to a plurality of different disks. An object of the present invention is to provide a distributed, storage device.

According to one aspect of the present embodiment, a plurality of disks are included, wherein each of the plurality of disks is divided into a plurality of local groups, a local parity exists for each of the plurality of local groups, and a global It provides a RAID-based storage device, characterized in that parity exists.

According to another aspect of the present embodiment, in the above-described RAID-based storage device, the RAID controller further comprising a RAID controller that reads or writes data from each of the disks on which data of the same local group is distributed and stored, RAID-based storage devices are provided.

As described above, according to this embodiment, when data is read or written on a RAID array, in order to prevent a sudden drop in data I/O performance (bottleneck) that occurs when a disk error occurs, data in the same local group is different from each other. It has the effect of enabling data reading or writing to be performed by fast data restoration in a distributed structure distributed over a plurality of disks.

According to this embodiment, there is an effect of eliminating a bottleneck and improving data restoration performance by using a disk distribution structure to restore data when a disk error occurs.

According to the present embodiment, the RAID array supports data restoration in the event of a failure of a storage device (eg, a hard disk (HDD: Hard Disk Drive), a solid state disk (SSD: Solid State Drive), etc.) for the purpose of increasing the reliability of the array. In using the parity information, the parity information has an effect of improving the reliability of the array as well as the data restoration performance.

According to the present embodiment, since it is possible to control a trade-off between reliability and data restoration performance in the RAID array, it is possible to secure high flexibility in RAID array management. In particular, the all-flash array (AFA) has the effect of improving data restoration performance as the performance of the solid state disk (SSD) increases.

According to this embodiment, a plurality of disks are divided into a plurality of local groups, and fast data restoration is possible using a distributed structure in which data of the same local group is distributed to a plurality of different disks. In addition, according to this embodiment, it can be applied to any type of all-flash array (AFA) product, and there is an effect that can be implemented independently of RAID.

1A to 1C are diagrams for explaining a conventional RAID technology.
2 is a block diagram schematically illustrating a storage device for fast data restoration according to the present embodiment.
3 is a diagram illustrating a disk distribution structure according to the present embodiment.
4 is a diagram for explaining fast data restoration when an error occurs in the disk distribution structure according to the present embodiment.
5 is a diagram showing another embodiment of the disk distribution structure according to the present embodiment.
6 is a flowchart illustrating a data reading method for fast data restoration according to the present embodiment.
7 is a flowchart illustrating a data writing method for fast data restoration according to the present embodiment.

Hereinafter, this embodiment will be described in detail with reference to the accompanying drawings.

2 is a block diagram schematically illustrating a storage device for fast data restoration according to the present embodiment.

The storage device 200 according to the present embodiment includes an application 210 , a RAID controller 220 , and a storage 230 . Components included in the storage device 200 are not necessarily limited thereto.

Each component included in the storage device 200 may be connected to a communication path that connects a software module or a hardware module inside the device to organically operate with each other. These components communicate using one or more communication buses or signal lines.

Each component of the storage device 200 shown in FIG. 2 means a unit for processing at least one function or operation, and may be implemented as a software module, a hardware module, or a combination of software and hardware.

The storage device 200 according to the present embodiment means a storage device operating based on RAID. RAID is a multiple-array independent disk, and it is a technology that divides and stores some redundant data on a plurality of disks (hard disk (HDD) or solid state disk (SSD)). There are various methods of dividing data in RAID, The methods are called 'levels', and depending on the level, the reliability of the storage device is increased or the overall performance is improved.RAID binds a plurality of disks to one logical disk and operates as a single logical disk, and there are hardware and software methods. The hardware method makes a plurality of disks look like a single disk to the operating system The software method is mainly implemented in the operating system and makes the disk appear to the user as a single disk.

The storage device 200 according to the present embodiment basically operates 'RAID 6', but is not limited thereto, and other RAID levels may be applied within the scope of the RAID technology. 'RAID 6' is a method that includes a striped set (at least 4 disks) in which parity (error detection function) is distributed.

The storage device 200 according to the present embodiment enables fast data restoration using a disk distribution structure even if a disk error occurs in a RAID array situation such as an all-flash array (AFA). The storage device 200 according to the present embodiment can eliminate a bottleneck and improve data restoration performance by using a disk distribution structure to restore data when a disk error occurs.

The storage device 200 according to the present embodiment is also applicable to an All Flash Array (AFA).

An all-flash array (AFA) supports fast I/O (Input/Output) performance by combining a plurality of flash-based disks (SSDs) in an array form. All-flash array (AFA) is a technology used as primary storage based on fast I/O performance. All-flash arrays (AFAs) are being actively developed led by major storage vendors, and data protection is supported by utilizing proprietary RAID (Redundant Array of Inexpensive Disks) technology.

Common RAID technologies focus on the trade-off between space overhead and data reliability. On the other hand, all-flash arrays (AFAs) are storage focused on fast I/O performance. All-Flash Array (AFA) is focusing on RAID technology for fast data restoration in order to prevent rapid data I/O performance degradation that occurs in the event of a disk failure.

The application 210 requests data input/output to the RAID controller 220 . The application 210 may be installed in the operating system of the external terminal. The external terminal refers to an electronic device that performs data communication via a network according to a user's key manipulation. In other words, the application 210 is installed in the external terminal and requests data read or write to the RAID controller 220 . The data input/output requested by the application 210 is mediated by the RAID controller 220 . The application 210 transmits at least one data read request to the RAID controller 220 based on at least one data position (Position). The application 210 transmits at least one data write request based on at least one data location to the RAID controller 220 .

The external terminal is provided with a memory for storing a program or protocol for communicating with the RAID controller 220 via a network, and a microprocessor for operating and controlling the program by executing the program. The external terminal is preferably a personal computer (PC: Personal Computer) and a laptop (Laptop), but is not necessarily limited thereto, and includes a smart phone, a tablet, a personal digital assistant (PDA), a game. It may be an electronic device such as a console, a portable multimedia player (PMP), a wireless communication terminal, a TV, or a media player.

An external terminal includes (i) a communication device such as a communication modem for performing communication with various devices or wired and wireless networks, (ii) a memory for storing various programs and data, and (iii) a microcomputer for operation and control by executing the program Various devices including a processor and the like. According to at least one embodiment, the memory is a computer such as random access memory (RAM), read only memory (ROM), flash memory, optical disk, magnetic disk, solid state disk (SSD), etc. It may be a readable recording/storage medium. According to at least one embodiment, a microprocessor may be programmed to selectively perform one or more of the operations and functions described herein. According to at least one embodiment, the microprocessor may be fully or partially implemented as hardware such as an Application Specific Integrated Circuit (ASIC) having a specific configuration.

Related data and a program are stored in a memory, and the processor reads and processes the related data from the memory. Although one processor may perform each of the above functions, the processor may be implemented so that a plurality of processors may divide and process the process. The processor may be implemented in a general-purpose processor, but may also be implemented as a chip manufactured separately to perform the function.

The RAID controller 220 processes corresponding data input/output using a plurality of disks according to predefined parity arrangement information (data layout). The RAID controller 220 calculates and stores parity according to parity arrangement information (data layout), and performs a function of restoring data using parity when an error occurs in a disk.

Hereinafter, a procedure for the RAID controller 220 according to the present embodiment to read data on the RAID array will be described.

The RAID controller 220 receives at least one data read request based on at least one data position from the application 210 . The RAID controller 220 uses a logical block address or an offset included in the data read request to check parity arrangement information (data layout) according to the data location on the disk distribution structure.

The RAID controller 220 identifies a disk number (eg, #1, #2, #3... #9) for processing at least one data read request on the disk distribution structure. RAID controller 220 is a stripe number (eg, #0, #1, #2, #3) in the disk corresponding to the disk number (eg, #1, #2, #3... #9) (Disk Number) ) (Stripe Number). The RAID controller 220 is a specific disk (eg, #1) to which one data read is to be performed in a data read request of at least one of the disk numbers (eg, #1, #2, #3... #9). Check whether an error has occurred.

When an error occurs in a specific disk (eg, #1), the RAID controller 220 stores data for a stripe related to reading one data in a specific disk (eg, #1) while storing a plurality of disks different from the specific disk (eg, #1). Parity restoration related data (data of the same local groups and local parity (P parity) on a plurality of different disks) is read from

In other words, when an error occurs in a specific disk (eg, #1), the RAID controller 220 checks parity arrangement information (Data Layout) according to the data location on the disk distribution structure. The RAID controller 220 identifies the same local groups as the specific disk for each stripe related to reading one data in the specific disk (eg, #1) with reference to parity arrangement information (data layout). The RAID controller 220 reads data and local parity of the same local groups on a plurality of different disks as parity restoration related data.

The RAID controller 220 restores data to be read in a stripe related to one data read by using parity restoration related data. In other words, the restoration process of the RAID controller 220 will be described. The RAID controller 220 restores data to be read in a stripe related to one data read by using data of the same local group and local parity. The RAID controller 220 restores the data to be read in the stripe and returns it in response to the data read request.

Hereinafter, a procedure for the RAID controller 220 to write data on the RAID array will be described.

The RAID controller 220 receives at least one data write request based on at least one data location from the application 210 . The RAID controller 220 identifies a disk number (eg, #1, #2, #3... #9) for processing at least one data write request on the disk distribution structure. The RAID controller 220 reads the stripe number (eg, #0, #1, #2, #3) in the disk corresponding to the checked disk number (eg, #1, #2, #3... #9). Check it.

The RAID controller 220 checks the local parity for each of the same local groups in the disk corresponding to the disk number, respectively, and checks the global parity for each stripe in the disk corresponding to the disk number, respectively. The RAID controller 220 is configured to write data for a stripe related to one data write in a disk corresponding to a disk number (eg, #1, #2, #3... #9) and a local data write to perform one data write. Check parity and global parity.

The RAID controller 220 is a specific disk (eg, #1) to which one data write is to be performed in a data write request of at least one of the disk numbers (eg, #1, #2, #3... #9). Check whether an error has occurred. In other words, the RAID controller 220 checks whether an error occurs in a specific disk (eg, #1) including data for a stripe related to writing one data, local parity or global parity to which one data write is performed. .

When an error occurs in a specific disk (eg, #1), the RAID controller 220 prevents data writing to a stripe related to writing one data in a specific disk (eg, #1).

A disk distribution structure according to the present embodiment will be described. The disk distribution structure refers to a DLRC (Distributed Local Reconstruction Code) method. In the disk distribution structure, a plurality of disks (hard disk (HDD) or solid state disk (SSD)) is divided into a plurality of local groups. In the disk distribution structure, local parity exists for each local group. In the disk distribution structure, global parity exists for each stripe in a plurality of disks. The disk distribution structure has a distribution structure in which data of the same local group is distributed to a plurality of different disks.

Local parity (P parity) and global parity (Q parity) in the disk distribution structure have a distributed structure in which a plurality of disks are distributed and stored. Local parity (P parity) and global parity (Q parity) in the disk distribution structure may be separately stored on each disk. In the disk distribution structure, only one parity of local parity (P parity) and global parity (Q parity) is distributed among a plurality of different disks, and the remaining parity may be separately stored on one disk. .

In the disk distribution structure, data of the same local group is distributed and stored on a plurality of different disks. When restoring data to be read in a stripe related to reading one data, data is read according to the local group distributed and stored for each individual disk. The number of executions is reduced (distributed) so that the bottleneck is alleviated.

The storage 230 means a plurality of physical disks. The storage 230 may be implemented as a plurality of hard disks (HDD), a plurality of solid state disks (SSD), or a combination of a hard disk (HDD) and a solid state disk (SSD). The storage 230 stores data under the control of the RAID controller 220 .

3 is a diagram illustrating a disk distribution structure according to the present embodiment.

The general LRC method does not improve data restoration performance in a RAID array where the I/O performance of all disks is the same. In other words, due to the fixed local group determination method, the LRC method uses only some disks to restore data when a disk error occurs, so that a bottleneck occurs and data restoration performance is not improved.

Compared to the LRC method, the disk distribution structure of this embodiment uses the distributed local group determination method, and utilizes the entire disk for data restoration when a disk error occurs, thereby alleviating/offending the bottleneck and improving data restoration performance.

3 illustrates a difference between parity arrangement information (data layout) of the DLRC method and the LRC method according to the present embodiment. The difference between the DLRC method and the LRC method is that, in the LRC method, all stripes (one data unit) in one disk are fixed to one local group. On the other hand, in the case of the DLRC method, the stripes within one disk are distributed to each local group. Due to the above-described differences, the DLRC method has an advantage in terms of data recovery when an error occurs in the disk.

The disk distribution structure according to the present embodiment is a DLRC scheme as shown in FIG. 3, and a plurality of disks (hard disks (HDDs) or solid state disks (SSDs)) are divided into a plurality of local groups. In the disk distribution structure, local parity exists for each local group. In the disk distribution structure, global parity exists for each stripe in a plurality of disks. The disk distribution structure has a distribution structure in which data of the same local group is distributed to a plurality of different disks. Local parity (P parity) and global parity (Q parity) in the disk distribution structure may be separately stored in each disk as shown in FIG. 3 .

4 is a diagram for explaining fast data restoration when an error occurs in the disk distribution structure according to the present embodiment.

As shown in FIG. 4 , the number of data reads performed by other disks in order to restore data when an error occurs in the first disk is shown based on the parity arrangement information (data layout) of the LRC method and the DLRC method. For the sake of understanding, it is assumed that a random read is used. In case of random read, it is assumed that read request merge does not occur. On average, data read requests requested due to a disk error occur equally as '#0, #1, #2, #3' in 'Stripe Number Mod 4', and on average, ' The number of data reads performed by different disks to restore 4' data is as shown in FIG. 4 .

In the case of the LRC method, disk read performance is concentrated on three disks in the same local group, and this causes the bottleneck of ‘two’ data disks, which degrades data restoration performance. On the other hand, in the case of the DLRC method, the same local group is distributed to different '6' disks, so compared to the LRC method, data read performance of half of each disk occurs relatively compared to the LRC method. For this reason, in the case of the DLRC method, the bottleneck is alleviated/cancelled, and data restoration is performed quickly.

For example, if data read requests to all disks are '4' each when an error occurs in data, in the case of the LRC method, the second and third disks must complete each '8' data read request to request a full data read. However, in the DLRC method, the second, fourth, fifth, and sixth disks can complete the entire data read request only by performing each 'sixth' data read.

The DLRC scheme of this embodiment has a major feature that stripes in one disk are distributed to each local group. FIG. 3 is an embodiment for showing such a characteristic, and in the actual implementation of the invention, various types of parity arrangement information (data layout) according to the number of disks, local parity (P parity), global parity (Q parity) location change, etc. ) can be implemented.

5 is a diagram showing another embodiment of the disk distribution structure according to the present embodiment.

5 shows two examples (performance, reliability) of various types of parity arrangement information (data layout) configurable in a DLRC scheme. The DLRC-P method is parity arrangement information (data layout) focused on performance, in which 'P1 parity (local parity)', 'P2 parity (local parity)', and 'Q parity (global parity)' are adjacent within each disk. This indicates that the stripe in which parities are stored is circulated for each disk. In this case, since all parity is cycled, there is no degradation in data write performance, and when a disk error occurs, data read requests to other disks are equally generated at half level, which is the fastest data restoration possible.

The DLRC-R method focuses on data reliability. 'Q parity (global parity)' is fixed within one disk, 'P1 parity (local parity)' is 'P2 parity (local parity)' in the first three disks. parity)' is cycled through the last three disks. Since 'Q parity (global parity)' is fixed within one disk, a bottleneck occurs when writing data and performance is degraded. can restore up to three disks depending on the location of the error on the disk.

For example, in the storage device 200 according to the present embodiment, even if errors occur in the first, fourth, and fifth disks, restoration is possible. In addition, the storage device 200 according to the present embodiment includes various embodiments in which the number of disks increases, the number of local parities (P1, P2 parities) increases, and the number of global parities (Q parity) increases.

The disk distribution structure according to the present embodiment is a DLRC method as shown in FIG. 5, and a plurality of disks (hard disks (HDDs) or solid state disks (SSDs)) are divided into a plurality of local groups. In the disk distribution structure, local parity exists for each local group. In the disk distribution structure, global parity exists for each stripe in a plurality of disks. The disk distribution structure has a distribution structure in which data of the same local group is distributed to a plurality of different disks.

When the disk distribution structure focuses on performance, as shown in FIG. 5 , local parity (P parity) and global parity (Q parity) in the disk distribution structure are distributed and stored in a plurality of different disks.

When the disk distribution structure focuses on data reliability, as shown in FIG. 5, only one parity of local parity (P parity) and global parity (Q parity) in the disk distribution structure is divided into a plurality of disks that are different from each other. It has a distributed distribution structure, and the remaining parity may be separately stored on one disk.

6 is a flowchart illustrating a data reading method for fast data restoration according to the present embodiment.

The application 210 generates a data read request based on a location (or logical block address) and an offset (S610).

The RAID controller 220 receives at least one data read request based on at least one data location from the application 210 . The RAID controller 220 uses a logical block address or an offset included in the data read request to check parity arrangement information (data layout) according to the data location on the disk distribution structure.

The RAID controller 220 identifies a disk number (eg, #1, #2, #3... #9) for processing at least one data read request on the disk distribution structure. The RAID controller 220 checks the stripe number (eg, #0, #1, #2, #3) in the disk corresponding to the disk number (eg, #1, #2, #3... #9) (S620).

In step S620, the RAID controller 220 checks parity arrangement information (data layout) according to the data location included in the data read request on the disk distribution structure. The disk distribution structure has a distribution structure in which data of the same local group is distributed to a plurality of different disks.

The RAID controller 220 is a specific disk (eg, #1) to which one data read is to be performed in a data read request of at least one of the disk numbers (eg, #1, #2, #3... #9). It is checked whether an error has occurred (S630).

When an error occurs in a specific disk (eg, #1) in step S630, the RAID controller 220 stores data for a stripe related to reading one data within a specific disk (eg, #1) while storing data that is different from the specific disk. Parity restoration related data (data of the same local groups and local parity (P parity) on a plurality of different disks) are read from the plurality of disks (S640).

In step S640, when an error occurs in a specific disk (eg, #1), the RAID controller 220 checks parity arrangement information (data layout) according to the data location on the disk distribution structure. The RAID controller 220 identifies the same local groups as the specific disk for each stripe related to reading one data in the specific disk (eg, #1) with reference to parity arrangement information (data layout). The RAID controller 220 reads data and local parity of the same local groups on a plurality of different disks as parity restoration related data.

The RAID controller 220 restores data to be read in a stripe related to one data read by using the parity restoration related data (S650). In step S650, the RAID controller 220 restores data to be read in a stripe related to one data read by using data of the same local group and local parity.

The disk distribution structure is a DLRC method. Since the number of disks in the same local group is small and the same local group of the disk in error is distributed to all disks, it is intended to perform read within a stripe related to one data read quickly without a bottleneck. data can be restored.

When an error occurs in a specific disk (eg, #1) in step S630, the RAID controller 220 reads data to be read in a stripe related to one data read (S660). The RAID controller 220 restores the data to be read in the stripe and returns it in response to the data read request (S670).

Although it is described that steps S610 to S670 are sequentially executed in FIG. 6 , the present invention is not limited thereto. In other words, since it may be applicable to changing and executing the steps described in FIG. 6 or executing one or more steps in parallel, FIG. 6 is not limited to a chronological order.

As described above, the data reading method for fast data restoration according to the present embodiment illustrated in FIG. 6 may be implemented as a program and recorded in a computer-readable recording medium. A computer-readable recording medium in which a program for implementing the data reading method for fast data restoration according to the present embodiment is recorded and includes all types of recording devices in which data readable by a computer system is stored.

7 is a flowchart illustrating a data writing method for fast data restoration according to the present embodiment.

The application 210 generates a data write request based on a location (or logical block address) and an offset (S710).

The RAID controller 220 receives at least one data write request based on at least one data location from the application 210 . The RAID controller 220 uses the logical block address or offset included in the data write request to check parity arrangement information (data layout) according to the data location on the disk distribution structure.

The RAID controller 220 identifies a disk number (eg, #1, #2, #3... #9) for processing at least one data write request on the disk distribution structure. The RAID controller 220 checks the stripe number (eg, #0, #1, #2, #3) in the disk corresponding to the disk number (eg, #1, #2, #3... #9) (S720).

The RAID controller 220 corresponds to the stripe number (eg, #0, #1, #2, #3) in the disk corresponding to the disk number (eg, #1, #2, #3... #9). Data is read (S730). The reason why the RAID controller 220 performs step S730 is to calculate global parity.

The RAID controller 220 checks the local parity for each of the same local groups in the disk corresponding to the disk number, respectively, and checks the global parity for each stripe in the disk corresponding to the disk number (S740).

The RAID controller 220 is configured to write data for a stripe related to one data write in a disk corresponding to a disk number (eg, #1, #2, #3... #9) and a local data write to perform one data write. Parity and global parity are checked (S750).

The RAID controller 220 is a specific disk (eg, #1) to which one data write is to be performed in a data write request of at least one of the disk numbers (eg, #1, #2, #3... #9). It is checked whether an error has occurred (S760). In step S760, the RAID controller 220 checks whether an error occurs in a specific disk (eg, #1) including data for a stripe related to writing one data, local parity or global parity to which one data writing is to be performed. .

If an error does not occur in a specific disk (eg, #1) in step S760, the RAID controller 220 writes data for a stripe related to writing one data, and writes one data to a local Parity or global parity is updated (S770).

When an error occurs in a specific disk (eg, #1) in step S760, the RAID controller 220 prevents data writing to a stripe related to writing one data in a specific disk (eg, #1) is not performed (S780) . In step S780, when an error occurs in a specific disk (eg, #1), the RAID controller 220 returns as response data to the data write request without performing (ignoring) writing data.

Although it is described that steps S710 to S780 are sequentially executed in FIG. 7 , the present invention is not limited thereto. In other words, since it may be applicable to changing and executing the steps described in FIG. 7 or executing one or more steps in parallel, FIG. 7 is not limited to a chronological order.

As described above, the data writing method for fast data restoration according to the present embodiment illustrated in FIG. 7 may be implemented as a program and recorded in a computer-readable recording medium. A computer-readable recording medium in which a program for implementing the data writing method for fast data restoration according to the present embodiment is recorded and includes all types of recording devices in which data readable by a computer system is stored.

The above description is merely illustrative of the technical idea of this embodiment, and a person skilled in the art to which this embodiment belongs may make various modifications and variations without departing from the essential characteristics of the present embodiment. Accordingly, the present embodiments are intended to explain rather than limit the technical spirit of the present embodiment, and the scope of the technical spirit of the present embodiment is not limited by these embodiments. The protection scope of this embodiment should be interpreted by the following claims, and all technical ideas within the scope equivalent thereto should be interpreted as being included in the scope of the present embodiment.

As described above, the present embodiment is applied to the RAID field and is a useful invention that prevents a sudden drop in data input/output performance (bottleneck) that occurs when an error occurs in a disk.

200: storage device 210: application
220: RAID controller 230: storage

Claims (9)

comprising a plurality of disks,
Each of the plurality of disks is divided into a plurality of local groups, a local parity exists for each of the plurality of local groups, and a global parity exists for each stripe of each of the plurality of disks,
Each of the local parity and the global parity is distributed and stored in the plurality of disks,
The local parity and the global parity are stored in the same order for each disk, one each in a plurality of adjacent stripes within the same disk, and the plurality of stripes in which the local parity and the global parity are stored determines the level of the stripe for each disk. stored for circulation in different ways
characterized in that, RAID-based storage device. delete delete According to claim 1,
Any one of the local parity and the global parity is stored in any one of the plurality of disks, and the other one is distributed and stored in the other disk, RAID-based storage device. 5. The method of claim 4,
The global parity is stored in any one of the plurality of disks, and the local parity is stored in the remaining disks of the plurality of disks so that a stripe in which the local parity is stored is circulated, RAID-based storage Device. According to claim 1,
A RAID-based storage device, characterized in that the stripes of each of the plurality of disks are distributed in each of the plurality of local groups. According to claim 1,
A RAID-based storage device, characterized in that data of the same local group is distributed and stored in a plurality of different disks among the plurality of disks. 8. The method of claim 7,
A RAID controller that reads data or writes data from each disk in which the data of the same local group is distributed and stored
Further comprising a, RAID-based storage device. 9. The method of claim 8,
The RAID controller,
A RAID-based storage device, comprising receiving an input/output request from an application and processing the input/output request from the plurality of disks using predefined parity configuration information.

Images