WO2015162766A1

WO2015162766A1 - Storage system and semiconductor storage device

Info

Publication number: WO2015162766A1
Application number: PCT/JP2014/061611
Authority: WO
Inventors: 英通小関
Original assignee: 株式会社日立製作所
Priority date: 2014-04-24
Filing date: 2014-04-24
Publication date: 2015-10-29

Abstract

　A storage system is provided with: a plurality of semiconductor storage devices having a plurality of nonvolatile semiconductor memory chips, cache memory, and a processor; and a storage controller for controlling the plurality of semiconductor storage devices as a RAID group and storing user data and parity data in the plurality of semiconductor storage devices. The storage controller correlates one from among a plurality of sub-logical storage areas in each of the plurality of semiconductor storage devices to each of a plurality of extents one by one, and transmits configuration information indicating correspondence between the arrangement of user data and parity data in a logical address space and the plurality of extents and then a lookahead command indicating that the data be stored in the cache memory to each of the plurality of semiconductor storage devices. In accordance with the lookahead command received, the processor of the semiconductor storage devices reads the user data and parity data corresponding to the extent from the plurality of nonvolatile semiconductor memory chips on the basis of the configuration information and stores the read data in the cache memory.

Description

Storage system and semiconductor storage device

The present invention relates to a storage system and a semiconductor storage device.

The storage system has a large number of drives for storing data (HDD (Hard Disk Drive), SSD (Solid State Drive), etc.) and a storage controller for controlling the drive, and the computer has a large capacity data storage space. It is intended to provide.

In addition, the storage system can perform high-speed I / O processing by using a plurality of drives in a RAID (Redundant Array of Independent (or Independent) Disks) configuration. Further, the storage system is characterized in that parity data is generated based on a plurality of user data and stored in the drive, and has high fault tolerance against a drive failure.

One of the functions of the storage system is the capacity virtualization function. Capacity virtualization is a technique for providing a virtual capacity larger than the physical capacity of a storage system to a host computer as a virtual volume. In capacity virtualization, a plurality of drives in a storage system are first created to create a pool, and storage areas in the pool are managed in units of extents of a predetermined size. Then, in response to a write request from the host computer to the virtual volume, the storage system allocates an extent to the area in the virtual volume specified by the write request. Accordingly, the user can suppress the physical capacity of the drive to the minimum necessary in accordance with the usage application from time to time, thereby realizing cost reduction.

Also, in a storage system, data may be moved between drives for the purpose of appropriate data arrangement. Data movement is performed in extent units. For example, the storage system switches the drive to which the extent is placed according to the access characteristics per unit time (I / O frequency or the like) for each extent. For example, extent data having a high I / O frequency is preferentially stored in a high-performance drive. This reduces the amount of high-performance drive required and reduces the cost of the storage system when designing a configuration that meets system performance and capacity requirements compared to storing data on a single type of drive. Reduction is possible. Hereinafter, data movement processing in units of extents is referred to as extent rearrangement processing.

Here, in the extent relocation processing, the storage system changes the mapping of the extent and the virtual volume before the data movement to the mapping of the extent and the virtual volume after the data movement. Thereby, data movement can be performed without making the host computer recognize.

The extent rearrangement process is performed asynchronously with the host computer I / O process. Therefore, if the extent rearrangement takes time, the I / O processing performance of the host computer deteriorates. That is, extent rearrangement is performed to improve performance, but this processing also causes performance degradation. For this reason, it is better that the data movement time for extent rearrangement is short.

Here, for the purpose of improving response performance, data pre-fetching (prefetch) is a technique for copying data in advance from a low-speed storage medium to a high-speed storage medium such as DRAM in the drive.

In Patent Document 1, a host computer designates a logical address range to a storage device (for example, a flash memory card) and issues a prefetch command, and the flash memory card stores data corresponding to the logical address range in the flash memory A technique is disclosed in which the read response time to the host computer is shortened by copying from the memory to the DRAM.

US2004 / 0205301

However, the technique of Patent Document 1 has the following problems. The data movement between the drives performed by the storage controller cannot be recognized by the host computer and is executed asynchronously with the I / O processing by the host computer, so the host computer instructs the drive to read ahead. I cannot grasp the opportunity. In addition, since the host computer cannot recognize the data movement, it is impossible for the host computer to specify a logical address to be prefetched for a plurality of drives constituting the RAID group.

Since the storage system has a large number of drives, it is overhead to include the logical address range information in the prefetch command every time the storage controller sends the prefetch command to the drive.

Accordingly, an object of the present invention is to improve the response performance when the storage controller reads data from the drive by performing read-ahead in the drive while reducing the load on the storage controller that controls access to a large number of drives. is there.

A storage system according to an embodiment includes a plurality of semiconductor storage devices and a storage controller. Each of the plurality of semiconductor memory devices includes a plurality of nonvolatile semiconductor memory chips, a cache memory that temporarily stores data, and a processor that reads data from the plurality of nonvolatile semiconductor memory chips and stores the data in the cache memory. Have. The storage controller controls a plurality of semiconductor storage devices as a first RAID group, and stores a plurality of user data and parity data generated from the plurality of user data in the plurality of semiconductor storage devices. The processor provides storage areas of the plurality of nonvolatile semiconductor memory chips to the storage controller in association with the logical address space. The storage controller constitutes a plurality of extents that are logical storage areas. The storage controller divides each logical address space of the plurality of semiconductor storage devices into a plurality of sub logical storage areas, and each of the plurality of extents includes one from the plurality of sub logical storage areas of the plurality of semiconductor storage devices. Match each one. The storage controller transmits configuration information indicating the correspondence between the arrangement of user data and parity data in the logical address space and the plurality of extents to each of the plurality of semiconductor storage devices. The storage controller transmits to each of the plurality of semiconductor storage devices a prefetch command that designates the first extent of the plurality of extents and instructs to store data in the cache memory. The processor of the semiconductor storage device specifies user data and parity data corresponding to the first extent based on the configuration information in response to reception of the prefetch command, and the specified user data and parity data are stored in a plurality of nonvolatile semiconductor memories. Read from the chip and store in cache memory.

∙ Improved response performance when the storage controller reads data from the drive without imposing a load on the storage controller. As a result, the time required for data movement is shortened, and even if data movement is performed, a decrease in I / O processing performance required from the host computer is suppressed.

FIG. 10 is a schematic diagram illustrating extent migration processing in the storage system according to the first embodiment. It is a block diagram which shows the structural example of a storage system. It is a block diagram which shows the structural example of SSD (Solid State Drive). It is a schematic diagram explaining the example of the prefetch process in SSD. It is a schematic diagram explaining a virtual volume. It is a block diagram which shows the structural example of an extent. It is a block diagram which shows the example of the information hold | maintained at the memory which a storage controller has. The structural example of a drive management table is shown. 3 shows an example of the configuration of a RAID management table. The structural example of a pool management table is shown. The structural example of an extent management table is shown. The structural example of a virtual volume management table is shown. The structural example of a structure information management table is shown. It is a flowchart which shows the example of a task process in a storage controller. It is a flowchart which shows the example of a structure change process. It is a flowchart which shows the example of a write process. It is a flowchart which shows the example of a read process. It is a flowchart which shows the example of an extent movement process. It is a flowchart which shows the example of the I / O process of a drive controller. It is a flowchart which shows the example of the extent prefetch process in a drive controller. FIG. 10 is a schematic diagram for explaining a read process in an extent movement process in the second embodiment. 12 is a flowchart illustrating an example of extent movement processing according to the second embodiment. 12 is a flowchart illustrating an example of extent prefetch processing in the drive controller according to the second embodiment.

FIG. 1 is a schematic diagram for explaining extent migration processing in the storage system 10 according to the first embodiment.

The storage system 10 includes SSDs 701 to 704 and HDDs (Hard Disk Drives) 705 to 708, which are examples of drives (typically nonvolatile storage devices), and a storage controller 100 that controls these drives. In the storage controller 100, the SSDs 700 to 704 are combined to form a RAID 5 RAID group 500, and the HDDs 705 to 708 are combined to form a RAID 5 RAID group 600. The storage controller 100 manages the storage area of the RAID group 500 (600) by dividing it into partitions of a predetermined size. This partition is called an “extent”. Therefore, as shown in FIG. 1, the extent 70 (80) includes data held by each of the plurality of SSDs 701 to 704 (HDDs 705 to 708).

The storage system 10 is connected to the host computer 30 and provides the virtual volume 60 to the host computer 30. The virtual volume 60 is a virtual volume configured according to thin provisioning, for example. When the storage controller 100 receives a write request for the virtual volume 60 from the host computer 30, the storage controller 100 allocates the extent 70 in the RAID group 500 to the virtual extent 610 of the virtual volume 60, and writes the data accompanying the write request to the extent 70. In this embodiment, this state is referred to as the extent 70 being the reference destination of the virtual extent 610. The extent 70 includes a predetermined number of user data D1 to D12 and parity data P1 to P4.

The storage controller 100 executes extent movement processing for moving the data of the extent 70 in the RAID group 500 to the extent 80 in the RAID group 600 as necessary, for example. The RAID group 500 is composed of SSDs with short I / O response time but small capacity, and the RAID group 600 is composed of HDDs with long I / O response time but large capacity. In the first embodiment, it is assumed that the migration source RAID group 500 and the migration destination RAID group 600 have the same RAID configuration and the same extent size. However, the migration destination RAID group may be composed of a plurality of SSDs.

When the I / O frequency to the virtual extent 610 is low, the storage controller 100 moves the data of the extent 70 in the RAID group 500, which is the reference destination of the virtual extent 610, to the extent 80 in the RAID group 600, Extent movement processing for switching the reference destination of the virtual extent 610 to the extent 80 is executed.

On the other hand, when the access frequency to the virtual extent 610 is high, the storage controller 100 moves the data group held in the extent 80, which is the reference destination of the virtual extent 610, to the extent 70, and changes the reference destination of the virtual extent 610. An extent movement process for switching to the extent 70 is executed.

That is, the storage controller 100 executes extent migration processing in which a virtual extent having a high I / O frequency is referred to as a RAID group 500 configured with an SSD having a short I / O response time. Thereby, the response time with respect to the host computer 30 of the storage system 100 can be shortened. The extent movement processing may be called “extent relocation processing”.

The size of one extent is generally relatively large from several MB to several GB. Therefore, the extent movement process for moving large data is a process that is heavy on the storage controller 100 and the drive and requires time. The present embodiment is intended to shorten the time required for extent movement processing.

FIG. 2 is a schematic diagram for explaining an example of prefetching processing in the SSD 701. This process is the same for the SSDs 702 to 704.

The SSD 701 holds user data D1, D5, D9 and parity data P4. These data D1, D5, D9 and P4 belong to the extent 70. When moving the extent 70, the storage system 10 executes the following processing.

(S1) The storage controller 100 generates a prefetch command having an extent number that is information that can identify the extent 70 of the migration source, and transmits the prefetch command to the drive controller 710. Since the extent 70 is configured across a plurality of SSDs 701 to 704, the storage controller 100 transmits a prefetch command to each of the SSDs 701 to 704. When moving a plurality of extents, the storage controller 100 may generate a prefetch command having a plurality of extent numbers.

(S2) The drive controller 710 that has received the prefetch command refers to the configuration information management table 2100 (see FIG. 13) and identifies the data D1, D5, D9, and P4 included in the migration source extent 70.

(S3) The drive controller 710 transfers the identified data D1, D5, D9, and P4 from the flash memory chip (hereinafter “FM”) 780 to the cache area 2000 (see FIG. 4) of the memory 716.

(S4) The storage controller 100 transmits, for example, a read command for the user data D1 to the drive controller 710 in the extent movement process.

(S5) The drive controller 710 that has received the read command transmits the user data D1 in the cache area 2000 to the storage controller 100 because the user data D1 has already been stored in the cache area 2000 (because of a cache hit). The same applies to the other data D5, D9 and P4.

Storage controller 100 transmits a prefetch command to drive controller 710 as described above before executing the extent movement process. Thereafter, the storage controller 100 executes extent migration processing. As a result, the drive controller 710 can store data that will be read later in the cache area 2000 based on the prefetch command. That is, the drive controller 710 can return read data in a short time in response to a read command received thereafter. That is, according to the present embodiment, the time required for the extent movement processing can be shortened.

FIG. 3 is a block diagram illustrating a configuration example of the storage system 10.

A host computer 30 is connected to the storage system 10 via a SAN (Storage Area Network) 20. There may be a plurality of host computers 30.

The host computer 30 and the storage system 10 transmit and receive data via the SAN 20. The SAN 20 is, for example, a Fiber Channel, a SCSI (Small Computer System Interface), an iSCSI (Internet Small System System SI), or a USB (Universal Serial Bus A 94), or a USB (Universal Serial Bus S 94). The SAN 20 may be another communication network such as a LAN (Local Area Network).

The host computer 30 is a type of host device that uses the storage system 10. The host computer 30 is, for example, an application server or a database server. The host computer 30 transmits a read request or a write request to the storage system 10 to read data from the storage system 10 or write data to the storage system 10.

The host computer 30 may transmit a request regarding the storage configuration to the storage system 10 to cause the storage system 10 to configure a RAID group, change the configuration of the RAID group, or generate a virtual volume 80. Note that a request related to the storage configuration may be transmitted by a management computer (not shown) that manages the storage system 10 instead of the host computer 30.

The storage system 10 includes a storage controller 100, SSDs 701 to 704, and HDDs 705 to 708. The SSDs 701 to 704 and the HDDs 705 to 708 may be stored in the drive box 110.

The storage controller 100 has a function of controlling the SSDs 701 to 704 and the HDDs 705 to 708. The storage controller 100 includes a processor 104, a host I / F 101, a drive I / F 107, a memory 103, and an internal bus 102 that connects these

elements

104, 101, 107, and 103. Each of the processor 104, the memory 103, the host I / F 101, and the drive I / F 107 may be one, or two or more.

The processor 104 implements various functions of the storage system 10 by executing various operations according to an arithmetic circuit or a computer program and controlling the

other elements

101, 103, and 107.

The memory 103 stores computer programs and data used by the processor 104. The memory 103 includes, for example, a DDR3-DRAM, an MRAM (Magnetorative Random Access Memory), or a FeRAM (Ferroelectric Random Access Memory).

The host I / F 101 is an I / F device for connecting the storage controller 100 to the SAN 20.

The drive I / F 107 is an I / F device for connecting SSDs 701 to 704 and HDDs 705 to 708 as types of drives to the storage controller 100.

The internal bus 102 is a path for the

elements

104, 101, 107 and 103 to transmit data bidirectionally. The internal bus 102 is configured by, for example, PCI (Peripheral Component Interconnect). The internal bus 102 may be connected to devices (none of which are shown) that realize a switch function, a DMA transfer function, a RAID calculation function, and the like related to data transmission. These devices may be configured by ASICs (Application Specific Integrated Circuits).

FIG. 4 is a block diagram illustrating a configuration example of the SSD 701. Note that the SSDs 702 to 704 also have similar configuration examples.

The SSD 701 includes a drive controller 710 and a plurality of FM 780 that are a kind of nonvolatile semiconductor storage media connected to the drive controller 710.

FM780 is, for example, a NAND type or NOR type flash memory. The FM 780 reads and writes data in units of pages that are a section of the storage area. The FM 780 may be, for example, a phase change memory, an MRAM that is a magnetoresistive memory, a ReRAM (Resistance Random Access Memory) that is a resistance change type memory, or an FeRAM that is a ferroelectric memory.

The drive controller 710 includes a processor 713, a memory 716, a drive I / F 711, an FM controller 717, a compression / decompression circuit 718, and an internal bus 712 that connects these

elements

713, 716, 711, 717, and 718. Is done. Each of the processor 713, the memory 716, and the drive I / F 711 may be one, or two or more.

The processor 713 implements various operations according to an arithmetic circuit or a computer program, and controls

other elements

711, 716, 718, 717, thereby realizing various functions of the SSD.

The memory 716 stores computer programs and data used by the processor 713. The memory 716 is a storage medium that can read / write data at a higher speed than the FM 780, and is, for example, a DRAM. The memory 716 has a cache area 2000. In the cache area 2000, data to be written to the FM 780 and data read from the FM 780 are temporarily stored. The data read from the FM 780 includes data read in response to the pre-read command. The memory 716 includes a configuration information management table 2100. Details of the configuration information management table 2100 will be described later.

The drive I / F 711 is an I / F device for connecting the drive controller 710 to the drive I / F 107 of the storage system 10.

The compression / decompression circuit 718 is a circuit for compressing data to be written to the FM 780 or decompressing data read from the FM 780. The function of the compression / decompression circuit may be realized by the processor 713.

The FM controller 717 controls a plurality of FM 780 connected to the FM controller 717. For example, the FM controller 717 controls reading and writing of data with respect to the FM 780.

The internal bus 712 is a path for the

elements

713, 716, 711, 717, and 718 to transmit data bidirectionally. The internal bus 712 may be connected to a device (not shown) that implements a switch function related to data transmission. This device may be configured by ASICs.

The processor 713 provides the storage controller 100 with the physical storage areas of the plurality of FM 780s associated with a logical address space (corresponding to an LBA (Logical Block Address) described later). The physical storage area is composed of a plurality of blocks, and the blocks are data erasure units. The block is composed of a plurality of pages, and the page is a unit of data read / write. The memory 716 stores information (logical-physical mapping table) indicating the correspondence between the logical address space and the physical storage area (a plurality of pages). The storage controller 100 transmits a read command / write command by designating a logical address. When the processor 713 in the SSD 700 receives a read command or a write command, the processor 713 identifies a page on the physical storage area from a logical address specified by the command based on the logical-physical mapping table, and executes a read or write process.

The HDDs 705 to 708 include a drive controller 710 and a storage medium controlled by the drive controller 710, although the FM 780, the FM controller 717, and the like are different from the SSD 701 in FIG. Are the same and will not be described.

FIG. 5 is a schematic diagram illustrating virtual volumes and pools.

The storage controller 100 configures a pool 500 by collecting storage areas of a plurality of

RAID groups

200, 300, and 400. The pool 500 may be configured by different types of RAID groups. For example, the pool 500 includes a RAID group 200 configured by SSD, a RAID group 300 configured by SAS-HDD, and a RAID group 400 configured by SATA-HDD. For example, the I / O speed increases in the order of SATA-HDD, SAS-HDD, and SSD, but the cost per unit capacity (bit cost) increases in the order of SATA-HDD, SAS-HDD, and SSD.

In a RAID group, tiers (layers) are set based on the types of drives that make up the RAID group. For example, in order of increasing I / O speed, the RAID group 200 composed of SSDs is tier 0, the RAID group 300 composed of SAS-HDDs is tier 1, and the RAID group 400 composed of SATA-HDDs is Tier 2 is set.

The virtual volume 600 is a virtual logical volume that the host computer 30 recognizes in order to store user data. The virtual volume 600 has a virtual address (logical address constituting the virtual volume), and a virtual extent is configured by dividing the virtual address into a predetermined range. Note that the capacity defined as the capacity of the virtual volume 600 can be a storage capacity larger than the total capacity of the storage media included in the storage system 10. The virtual volume 600 is assumed to be composed of an arbitrary number of virtual extents 601 to 607.

The storage controller 100 allocates extents constituting the pool 500 to the virtual extents of the virtual volume 600. For example, when the storage controller 100 receives a write request for a certain virtual address range in the virtual volume 600, the storage controller 100 specifies a virtual extent (for example, a virtual extent 602) that includes the virtual address range. Then, the storage controller 100 allocates the extent 242 of the RAID group 200 that constitutes the pool 500 to the specified virtual extent 602, and stores data associated with the write request. For example, none of the extents is allocated to the virtual extent 605 that does not include the address for which the write request has been made.

In FIG. 5, one extent is allocated to one virtual extent, but a configuration may be used in which a plurality of extents are allocated to one virtual extent.

The storage controller 100 may execute extent movement processing between RAID groups with different tiers. For example, when the I / O frequency for the virtual extent 602 is low, the storage controller 100 executes an extent migration process for migrating the reference extent 242 of the virtual extent 602 to the RAID group 300 of the tier 1. This is because maintaining extents with a low I / O frequency in a Tier 0 RAID group with a high bit cost results in poor cost performance.

FIG. 6 is a block diagram showing a configuration example of extents.

The RAID group 200 has a “3D + 1P” configuration of RAID 5 based on the SSDs 701 to 704. In the case of RAID 5, user data and parity data are rotated to SSDs 701 to 704 and stored in order. For example, as shown in FIG. 6, user data D1 is stored in SSD 711, user data D2 is stored in SSD 712, user data D3 is stored in SSD 713, and parity data P1 generated from D1 to D3 is stored in SSD 714. The set of D1, D2, D3, and P1 is called a stripe column 901. The sizes of Dn and Pn (n is an integer of 1 or more and indicates the order of data) are all equal, and this size is referred to as “strip width”. For example, when the stripe width is “256 KB”, the sizes of D1, D2, D3, and P1 are “256 KB”, and the stripe column size is “256 KB × 4 = 1024 KB”.

In the second stripe row, D4 is stored in SSD 714, D5 is stored in SSD 711, D6 is stored in SSD 712, and P2 generated from D4 to D6 is stored in SSD 713. Compared to the first stripe row, the SSD storing parity data is different. Thus, in the extent 70, the user data D1 to D16 and the parity data P1 to P6 are rotated and stored in order for each stripe column 901. In RAID5, parity data is generated by an exclusive OR operation (XOR operation) of a plurality of user data in each stripe column. For example, P1 is generated by an exclusive OR operation of D1, D2, and D3. The storage controller 100 may include an exclusive OR circuit and generate parity data, or the processor 104 may generate parity data by a program. Alternatively, the drive may include an exclusive OR operation circuit to generate parity data.

* By generating parity data and storing it in the drive in this way, the data can be restored in the event of a drive failure. For example, when a failure occurs in the drive 701, D1 is restored by an exclusive OR operation of D2, D3, and P1.

The cache area 1000 of the memory 103 is composed of a plurality of segments 500 having a predetermined size. Therefore, the size of data that can be stored in the cache area at a time is equal to or smaller than the size of the segment 500. For example, when the size of the segment and the stripe width are the same, the storage controller 100 stores D1 to D16 and P1 to P6 included in the extent 70 in the segment 500 of the cache area in 24 times.

6, user data and parity data are stored in the SSD 701 in the order of D1, D5, D9, P4, D13, and D17. Therefore, the SSD 701 has regularity that one parity data is stored after three user data. When attention is paid only to the user data of the SSD 701, the data is stored in the order of D1, D5, D9, D13, and D17. Therefore, the SSD 701 has regularity that the order of user data increases by four (that is, by 1024 KB).

This regularity is the same for the other SSDs 702 to 704. However, the number of user data D until the parity data P first appears varies depending on the position (order) of the SSD in the RAID group 200. For example, the SSDs 701 to 704 have the following relationship.

The SSD 701 stores the first parity data P after the three user data D.
The SSD 702 stores the first parity data P after the two user data D.
The SSD 703 stores the first parity data P after one user data D.
The SSD 704 stores the first parity data P at the top.

Therefore, each of the plurality of drives receives “n” and “m” (n and m are integers of 1 or more) of “nD + mP” indicating the RAID configuration, the stripe width size, and the RAID group from the storage controller 100. By receiving the configuration information including the position of the drive, the size of the extent, and the extent number for uniquely identifying the extent, the extent number and the LBA ( A configuration information management table 2100 (see FIG. 13) for specifying the correspondence with the drive LBA can be generated.

FIG. 7 is a block diagram illustrating an example of information held in the memory 103 included in the storage controller 100.

The memory 103 included in the storage controller 100 has a table area 1001 and a cache area 1000.

In the table area 1001, a drive management table 1100, a RAID management table 1200, a pool management table 1300, an extent management table 1400, and a virtual volume management table 1500 are stored.

The drive management table 1100 holds information about drives. The RAID management table 1200 holds information related to RAID groups. The pool management table 1300 holds information regarding the pool. The extent management table 1400 holds information related to extents. The virtual volume management table 1500 holds information regarding virtual volumes. Details of these tables will be described later.

The cache area 1000 is composed of a plurality of segments 500. The storage controller 100 reads data from the drive with a size equal to or smaller than the segment, and stores (caches) the data in the segment 500 of the cache area 1000. The storage controller 100 divides the write data transmitted from the host computer 30 into a size equal to or smaller than the segment and stores (caches) it in the segment 500 of the cache area 1000.

FIG. 8 shows a configuration example of the drive management table 1100.

The drive management table 1100 has information regarding each drive controlled by the storage controller 100. The drive management table 1100 has a record having a drive number 1101, a drive type 1102, and a logical capacity 1103 as attributes.

The drive number 1101 is information for uniquely identifying the drive. The drive type 1102 is information indicating the type of drive corresponding to the drive number 1101. The drive type 1102 is, for example, SSD, SAS-HDD, SATA-HDD, or the like. The logical capacity 1103 is information representing the logical storage capacity of the drive.

8, a record 1110 indicates that the drive corresponding to the drive number 1101 “0” indicates that the drive type 1102 is “SSD” and the logical capacity 1103 is “800 GB”.

FIG. 9 shows a configuration example of the RAID management table 1200.

The RAID management table 1200 has information regarding RAID groups controlled by the storage controller 100. The RAID management table 1200 includes a record having a RAID number 1201, a drive type 1202, a RAID level 1203, a RAID configuration 1204, a RAID configuration drive 1205, and a tier 1208 as attributes.

The RAID number 1201 is information for uniquely identifying a RAID group.

The drive type 1202 is the same information as the drive type 1102 in FIG. Since the RAID group is composed of the same type of drive, one drive type 1202 is associated with one RAID number 1201.

The RAID level 1203 is information representing the level of the RAID group. Examples of the RAID level include “RAID1 + 0”, “RAID1”, “RAID3”, “RAID4”, “RAID5”, “RAID6”, and the like.

The RAID configuration 1204 is information representing the configuration of a RAID group. “MD + nP” indicates that there are m pieces of data in the stripe column and n pieces of parity. For example, when three drives store user data and one drive stores parity data for one stripe column, the RAID configuration is expressed as “3D + 1P”.

The RAID configuration drive 1205 is information representing the drive number of each drive constituting the RAID group.

The tier 1206 is information representing a tier set in the RAID group. For example, a tier “0” is assigned to a RAID group configured with an SSD having the fastest I / O speed, a tier “1” is assigned to a RAID group configured with a SAS-HDD having the next fastest I / O speed, Tier “2” is set in the RAID group configured by the SATA-HDD having a low I / O speed.

In FIG. 9, the record 1210 includes a RAID group corresponding to the RAID number 1201 “0” composed of “SSD” (1202) corresponding to the drive numbers “1”, “10”,. Represents. In the record 1210, the RAID group corresponding to the RAID number 1201 “0” is configured by “RAID5” (1203) of “3D + 1P” (1204), and the RAID group has a tier “0” (1206). Indicates that is set.

FIG. 10 shows a configuration example of the pool management table 1300.

The pool management table 1300 has information regarding the pool controlled by the storage controller 100. The pool management table 1300 includes a record having a pool number 1301, a RAID number 1302, a RAID remaining capacity 1303, and a pool remaining capacity 1304 as attributes.

The pool number 1301 is information for uniquely identifying the pool. The RAID number 1302 is the same information as the RAID number 1201 of FIG. When one pool is composed of a plurality of RAID groups, a plurality of RAID numbers are associated with one pool number.

The RAID remaining capacity 1303 is information indicating the remaining capacity (usable capacity) of the RAID group. The pool remaining capacity 1304 is information indicating the remaining capacity (usable capacity) of the pool. That is, the pool remaining capacity 1304 is a total value of the remaining capacity of the RAID groups included in the pool.

10, the record 1310 indicates that the pool corresponding to the pool number 1301 “0” is configured by the RAID group corresponding to the RAID numbers “0”, “1”,..., “N” (1302). To express. For example, the remaining capacity 1303 of the RAID group corresponding to the RAID number 1302 “0” is “123 GB”, and the remaining capacity 1304 of the pool corresponding to the pool number 1301 “0” is “3456 GB”.

FIG. 11 shows a configuration example of the extent management table 1400.

The extent management table 1400 has information regarding extents controlled by the storage controller 100. The extent management table 1400 has a record having extent number 1401, RAID number 1402, extent size 1403, stripe width 1404, status 1405, extent LBA 1406, drive number 1407, and drive LBA 1408 as attributes. Have

The extent number 1401 is information for uniquely identifying an extent. The RAID number 1402 is the same information as the RAID number 1201 of FIG. The extent size 1403 is information representing the extent size (for example, Byte). The stripe width 1404 is information representing the size of the extent stripe width (for example, Byte).

Status 1405 is information indicating the status of the extent. When the extent has already been allocated to the virtual extent of the virtual volume, the status 1405 records the virtual volume number to which the extent is allocated and the virtual extent number. When the extent is not allocated to the virtual extent, the status 1405 is “N / A”. “0, 3” of the status 1405 in FIG. 11 indicates that the virtual extent number “3” of the virtual volume number “0” is assigned.

The in-extent LBA 1406 is information representing an LBA indicating one section of the storage area configured by the RAID group. The drive number 1407 is the same information as the drive number 1102 in FIG. The in-drive LBA 1408 is information representing an LBA range indicating a section of the storage area of the drive.

In FIG. 11, a record 1410 has an extent corresponding to the extent number 1401 “0” belonging to the RAID group corresponding to the RAID number 1402 “1”, and the virtual extent number “3” (1405) of the virtual volume number “0”. Indicates that it is allocated to a virtual extent corresponding to. The record 1410 represents that the extent size 1403 is “1 GB” and the stripe width 1404 is “256 KB” in the RAID group corresponding to the RAID number 1402 “1”. In the record 1410, the extent corresponding to the extent number 1401 “0” is the LBA in the extent corresponding to the RAID number 1402 “1” LBA 1406 “D1”, “D2”, “D3”, “P1”,. It represents that it is comprised by. The record 1410 indicates that the range in the drive LBA 1408 “0 KB to 255 KB” of the drive corresponding to the drive number 1407 “0” is assigned to the in-extent LBA 1406 “D1”.

FIG. 12 shows a configuration example of the virtual volume management table 1500.

The virtual volume management table 1500 has information on the virtual volume 60 that the storage controller 100 provides to the host computer 30. The virtual volume management table 1500 includes a virtual volume number 1501, a virtual capacity 1502, a real used capacity 1503, a virtual address 1510, a virtual extent number 1504, a RAID number 1505, an extent number 1506, a tier 1507, and a read. It has a record with IOPS 1508 and write IOPS 1509 as attributes.

The virtual volume 600 has a virtual address (logical address constituting the virtual volume), and a virtual extent is configured by dividing the virtual address into a predetermined range.

The virtual volume number 1501 is information for uniquely identifying the virtual volume. The virtual capacity 1502 is information representing the entire capacity of the virtual volume. The actual used capacity 1503 is information indicating the capacity actually used in the virtual volume. The actual used capacity 1503 is equal to the capacity allocated from the pool to the virtual volume. The virtual address 1510 indicates an address that uniquely identifies an area in the virtual volume. The virtual extent number 1504 is information for uniquely identifying a virtual extent. As shown in FIG. 12, the virtual address of the virtual volume is divided for each predetermined virtual address range (1510) and associated with the virtual extent (1504).

The RAID number 1505, extent number 1506, and tier 1507 are the same information as the RAID number 1201, extent number 1401, and tier 1206 described above.

The read IOPS 1508 is information indicating the number of reads that have occurred per unit time for the extent. The write IOPS 1509 is information indicating the number of writes that have occurred in unit time for the extent. The read IOPS 1508 and the write IOPS 1509 may be collectively referred to as “statistical information”. That is, the statistical information is information representing the I / O frequency for the extent. The statistical information is used to determine an extent to be subjected to extent movement processing. For example, the storage controller 100 sets an extent corresponding to an extent number for which the read IOPS 1508 or the write IOPS 1509 is smaller than a certain threshold as a migration target to a RAID group that is a lower tier (that is, the I / O speed is low). . On the other hand, the storage controller 100 sets the extent corresponding to the extent number for which the read IOPS 1508 or the write IOPS 1509 is larger than a certain threshold as the transfer target to the RAID group that is the upper tier (that is, the I / O speed is high). To do.

12, a record 1520 indicates that the virtual capacity 1502 of the virtual volume corresponding to the virtual volume number 1501 “0” is “800 GB” and the actual used capacity 1503 is “150 GB”. In the record 1520, the virtual volume corresponding to the virtual volume number 1501 “0” is configured by virtual extents corresponding to the virtual extent numbers 1504 “0”, “1”, “2”,. For example, this indicates that the area of the virtual address 1510 “0 to 1024” of this virtual volume is associated with the virtual extent number 1504 “0”. In this record 1520, the extent corresponding to the extent number 1506 “0” of the RAID group corresponding to the RAID number 1505 “0” is allocated to the virtual extent corresponding to the virtual extent number 1504 “0”. Represents. This record 1520 indicates that the extent corresponding to the extent number 1506 “0” belongs to the RAID group of the tier 1507 “0”. This record 1520 indicates that the read IOPS 1508 in the extent corresponding to the extent number 1506 “0” is “100” and the write IOPS 1509 is “0”.

Also, the RAID number 1505, extent number 1506, tier 1507, read IOPS 1508, and write IOPS 1509 corresponding to a virtual extent to which no extent has been assigned is “N / A”.

FIG. 13 shows a configuration example of the configuration information management table 2100.

The configuration information management table 2100 stored in the memory of each drive has information indicating the relationship between each drive and the extent. The configuration information management table 2100 includes a record having the attributes of the RAID configuration 2101, the in-RAID drive position 2102, the stripe width 2103, the extent size 2104, the in-drive LBA 2105, the in-extent LBA 2106, and the extent number. .

The RAID configuration 2101, the stripe width 2103, the extent size 2104, the in-drive LBA 2105, the in-extent LBA 2108, and the extent number 2107 are the RAID configuration 1204, the stripe width 1404, the extent size 1403, the in-drive LBA 1408, the in-extent LBA 1406, and the extent. The information is the same as the number 1401.

The in-RAID drive position 2102 is information indicating the position (order) of the own drive in the RAID group. As illustrated in FIG. 6, the arrangement of user data and parity data for each drive differs depending on the drive position in the RAID group. The intra-RAID drive position 2102 is used to determine the arrangement of user data and parity data.

In FIG. 13, a record 2110 indicates that its own drive (that is, the drive having the configuration information management table 2100) belongs to the RAID group configured by “3D + 1P” (2101), and “first” in the RAID group. (2102). The record 2110 represents that the RAID group to which the drive belongs belongs to a stripe width 2103 “256 KB” and an extent size 2104 “1 GB”. The record 2110 indicates that, for example, the in-drive LBA 2105 “0 KB to 255 KB” area of the own drive is allocated to the in-extent LBA 2106 “D1” having the extent number 2107 “1”.

Since a plurality of drives are controlled as a RAID group by the storage controller 100, each drive controller 710 cannot normally recognize that it is controlled as a RAID group. That is, each drive controller 710 cannot normally discriminate between user data and parity data. Since the extent is also managed by the storage controller 100, each drive controller 710 cannot normally recognize the extent configuration. For this reason, in this embodiment, the drive controller 710 receives the configuration information including the RAID configuration 2101, the RAID drive position 2102, the stripe width 2103, and the extent size 2104 from the storage controller 100, so that the RAID and extent configuration can be obtained. Can be recognized. The drive controller 710 can calculate the correspondence relationship between the in-drive LBA 2105, the in-extent LBA 2106, and the extent number 2107 based on these pieces of

information

2101, 1022, 2103, and 2104. That is, the range of the in-drive LBA 2105 and the size of the in-extent LBA 2106 can be calculated from the stripe width 2103. Further, from the extent size 2104 and the stripe width 2103, the number of in-extent LBAs 2108 (and in-drive LBAs 2105) belonging to one extent can be calculated. Further, it can be determined from the RAID configuration 2101 and the intra-RAID drive position 2102 whether each data in the in-extent LBA 2106 is user data or parity data.

FIG. 14 is a flowchart showing an example of task processing in the storage controller 100. The storage controller 100 periodically executes this task process.

The storage controller 100 determines whether or not a request related to configuration setting (initial setting or configuration change) has been received from the host computer 30 (S100). The configuration change request is a request that the host computer 30 sends to the storage controller 100 when the RAID group or extent configuration is changed. Also, when a RAID group is newly added to the pool, a request for a configuration change is issued for initial setting.

If a request for configuration change has been received (S100: Yes), the storage controller 100 executes configuration change processing (S101). Details of the configuration change processing will be described later (see FIG. 15).

If no request for configuration change has been received (S100: No), the storage controller 100 determines whether a read or write request has been received from the host computer 30 (S102). If no request has been received (S102: No), the storage controller 100 proceeds to S106.

When a read or write request has been received (S102: Yes), the storage controller 100 determines whether this request is a read or a write (S103).

If this request is a read (S103: Read), the storage controller 100 executes a read process (S104), and proceeds to S106. Details of the read process will be described later (see FIG. 17).

If this request is a write (S103: write), the storage controller 100 executes a write process (S105) and proceeds to S106. Details of the write process will be described later (see FIG. 16).

In S106, the storage controller 100 determines whether or not extent rearrangement is necessary (S106).

When it is determined that the extent rearrangement is necessary (S106: Yes), the storage controller 100 executes the extent migration process (S107), and proceeds to S108. Details of the extent movement processing will be described later (see FIG. 18).

When it is determined that the extent rearrangement is unnecessary (S106: No), the storage controller 100 proceeds to S108.

In S108, the storage controller 100 determines whether or not a request for stopping the storage system 10 has been received (S108).

When the stop request has been received (S108: Yes), the storage controller 100 executes the stop process of the storage system 10 and ends the process (END). When the stop request has not been received (S108: No), the storage controller 100 returns to S100.

FIG. 15 is a flowchart showing an example of the configuration change process. This configuration change process corresponds to S101 in FIG.

The storage controller 100 generates a RAID group according to the content of the configuration change request and updates the RAID management table 1200 (S200). The configuration change request includes contents recorded in the RAID management table 1200 such as the RAID level of the RAID group after the change.

The storage controller 100 registers the generated RAID group in the pool and updates the pool management table 1300 (S201).

The storage controller 100 generates an extent for the generated RAID group and updates the extent management table 1400 (S202).

The storage controller 100 transmits a configuration change command including initial setting or changed configuration information to each drive constituting the RAID group (S203). The configuration information includes, for example, the RAID level of the RAID group after the change, the RAID configuration (nD + mP), the drive position in RAID of each drive, the stripe width, the extent size, and the like.

The drive controller 710 updates the configuration information management table 2100 based on the configuration information received from the storage controller 100 (S204).

The storage controller 100 creates a virtual volume as necessary when it is included in the request from the host computer 30, and updates the virtual volume management table 1500 (S205). When the host computer 30 is requesting generation of a virtual volume, the request includes at least the size of the virtual volume. Then, this process ends (END).

Through the above processing, each drive controller 710 can generate the configuration information management table 2100.

FIG. 16 is a flowchart showing an example of the write process. This write process corresponds to S105 in FIG.

The host computer 30 transmits a write request and write data to the storage controller 100 (S300).

When the storage controller 100 receives a write request from the host computer 30, the storage controller 100 refers to the virtual volume management table 1500 and the extent management table 1400, and the extent is included in the virtual extent including the write destination address of the virtual volume specified by the write request. It is determined whether or not it has been assigned (S301).

If no extent is assigned to the write-destination virtual extent (S301: No), the storage controller 100 assigns an extent to the virtual extent (S302), and proceeds to S303.

If the extent has already been allocated to the write-destination virtual extent (S301: Yes), the storage controller 100 proceeds directly to S303.

In S303, the storage controller 100 stores the write data in the cache area 1000 (S303), and transmits a write command and write data to the write destination drive (S304). Then, the storage controller 100 receives a write completion notification from the write destination drive (S305).

The storage controller 100 updates the write IOPS 1509 (S306). The storage controller 100 transmits a completion response to the write request to the host computer 30 (S307).

The host computer 30 receives a completion response to the write request from the storage controller 100 (S308), and ends the processing (END).

Through the above processing, the storage system 10 can store the write data transmitted from the host computer 30.

FIG. 17 is a flowchart showing an example of read processing. This read process corresponds to S104 in FIG.

The host computer 30 transmits a read request to the storage controller 100 (S400). When the storage controller 100 receives a read request from the host computer 30, the storage controller 100 specifies a virtual extent including the address of the virtual volume specified by the read request, specifies an extent allocated to the virtual extent, and drives that constitute the extent. The drive of the read destination is specified from among (S401).

The storage controller 100 transmits a read command to the specified drive (S402). Then, the storage controller 100 receives read data from the drive (S403).

The storage controller 100 stores the read data in the cache area 1000 (S404). The storage controller 100 updates the read IOPS 1508 (S405).

The storage controller 100 transmits a completion response and read data to the read request to the host computer 30 (S406). The host computer 30 receives a completion response and read data from the storage controller 100 (S407), and ends the processing (END).

Through the above processing, the storage system 10 can respond to the read data in response to the read request from the host computer 30.

FIG. 18 is a flowchart showing an example of extent movement processing. This extent movement process corresponds to S107 in FIG.

The storage controller 100 determines the source extent (S500). For example, the storage controller 100 checks the extent number 1506 of the virtual volume management table 1500 and refers to the tier 1507 and IOPS (read IOPS 1508 and write IOPS 1509) corresponding to the allocated extent number. The storage controller 100 determines whether the tier where the extent is arranged is appropriate based on the referenced information. The storage controller 100 determines to move to a higher tier when an extent having a relatively large IOPS is arranged in the lower tier. A threshold may be provided for each tier, such as placing it at tier 0 if the IOPS is greater than or equal to a predetermined value. As the IOPS, only the read IOPS 1508 or the write IOPS 1509 may be used, or the total value of the read IOPS 1508 and the write IOPS 1509 may be used. As a result, data with a high IOPS is stored in the SSD, and the I / O processing performance from the host computer 30 can be improved.

 Relocation of extents is performed periodically. The shorter this period, the better the extents are placed and the performance will be improved. However, if the data is frequently moved, the data movement itself becomes an overhead for the storage controller 100, and the performance of the storage system is degraded. In this embodiment, the storage controller 100 instructs the SSD to read ahead before moving the data, thereby suppressing the performance degradation.

The storage controller 100 determines the migration destination extent from among the statuses 1405 “unallocated” (S501). That is, in order to move data, an unallocated extent in which no data is stored is required. In order to maximize the performance of the storage system, it is desirable to use as much of the SSD tier (top tier) capacity as possible. For this reason, when an extent arranged in the HDD is arranged in the SSD, there is a possibility that an unallocated extent does not exist. In this case, the storage controller 100 selects the extent with the smallest IOPS from the SSD tier, and moves the data stored in the extent to the “unallocated” extent of the HDD tier. Then, the storage controller 100 updates the status 1405 of the SSD tier extent to “unallocated”. As described above, when the extent rearrangement is frequently performed, the number of extent rearrangements using the SSD as a migration source increases. Therefore, the effect of improving the response performance when the storage controller reads data from the SSD by executing prefetching in the SSD before the rearrangement of extents as in the present embodiment is significant.

In this example, even if the IOPS is equal to or greater than the threshold, if there is an extent with a larger IOPS, data is transferred from the SSD to the HDD. For this reason, it is difficult to measure IOPS inside the SSD and perform prefetching by prediction based on the measured IOPS. Therefore, the storage controller 100 transmits a prefetch command to each drive after determining whether or not extent rearrangement is necessary.

The storage controller 100 secures the number of segments 500 necessary for migration of data belonging to the extent in the cache area 1000 (S502).

The storage controller 100 transmits a prefetch command to each of the plurality of drives holding the migration source data (S503). The prefetch command is a command for instructing prefetching of data, and has one or more transfer source extent numbers to be prefetched. Here, the drive controller 710 prefetches data based on the extent number included in the prefetch command. Details of this processing will be described later (see FIG. 20).
The storage controller 100 notifies each drive of the extent number in advance, and each drive holds information on the correspondence between the extent number and the logical address provided by its own drive to the storage controller 100 (FIG. 13). ing. As a result, the storage controller 100 can cause the drive to perform prefetching simply by notifying each drive of the extent number to be prefetched using the prefetch command. That is, the load on the storage controller 100 for transmitting the prefetch command is small, and the I / O processing performance of the host computer 30 does not need to be reduced.

In this embodiment, since the migration source and destination have the same RAID level, RAID configuration, and stripe width, prefetching of both user data and parity data is designated. The extent includes one or more stripes including a plurality of user data and parity data generated from the plurality of user data. The value of the parity data is not updated unless any of a plurality of user data included in the stripe is updated. That is, it is not necessary to update parity data in data movement that maintains a stripe. Therefore, it is efficient for the storage controller 100 to read from the drive including parity data and move both user data and parity data.

The storage controller 100 transmits a read command for data (user data and parity data) belonging to the migration source extent to each of the plurality of drives constituting the migration source extent (S504). The read command includes at least information specifying an address range (first LBA and length) in which data to be read is stored.

The storage controller 100 reads data belonging to the migration source extent and stores it in the cache area 1000 (S505). The storage controller 100 transmits a write command to each of the plurality of drives constituting the extent in order to write data to the migration destination extent (S506). The write command includes information for designating the address range (first LBA and length) of the data write destination.

The storage controller 100 releases the segment of the cache area 100 (S507). The storage controller 100 releases the migration source extent (S508). Specifically, the storage controller 100 deletes data stored in the migration source extent. The storage controller 100 updates the virtual volume management table 1500 and the extent management table 1400 (S509).

With the above processing, the storage controller 100 can improve the response performance from the drive constituting the migration source extent without increasing the load on the storage controller 100, and shorten the time required for data migration. Can do.

FIG. 19 is a flowchart showing an example of I / O processing of the drive controller 710.

The drive controller 710 determines whether there is an unprocessed command (S600). If there is no unprocessed command (S600: No), the drive controller 710 ends the process (END).

When there is an unprocessed command (S600: Yes), the drive controller 710 selects a command to be processed and determines the content of the command (S601).

If the command is a prefetch command (S601: prefetch), the drive controller 710 executes extent prefetch processing (S602), and returns to S600. Details of the extent prefetching process will be described later (see FIG. 20).

If the command is a read or write command (S601: read or write), it is determined whether the command is read or write (S603).

When the command is a read command (S603: Read), the drive controller 710 determines whether or not the read target data is already stored in the cache area 2000 (whether or not the cache is hit) (S604).

When the read target data has been stored in the cache area 2000 (S604: Yes), the drive controller 710 transmits the data stored in the cache area 2000 to the storage controller (S606), and returns to S600.

When the read target data is not stored in the cache area 2000 (S604: No), the drive controller 710 reads data from the FM 780 and stores it in the cache area 2000 (S605). Then, the drive controller 710 transmits the data stored in the cache area 2000 to the storage controller 100 (S606), and returns to S600.

If the command is a write command (S603: write), the drive controller 710 stores the write data in the cache area 2000 (S607). Then, the drive controller 710 writes the write data stored in the cache area 2000 to the FM 780 (S608), and returns to S600.

Here, the read / write process when the SSD has a compression function will be described.

First, the case of lead will be described. The processor 713 reads the compressed data from the FM 780 and stores it in the cache area 2000. Then, the processor 713 reads the compressed data from the cache area 2000 and transfers it to the compression / decompression circuit 718. The compression / decompression circuit 718 decompresses the compressed data. The processor 713 stores the decompressed data in the cache area 2000.

Next, the case of light will be explained. After the write data is stored in the cache area 2000 in S607, the processor 713 reads the write data from the cache area 2000 and transfers it to the compression / decompression circuit 718. Then, the processor 713 transfers the compressed write data to the FM 780. Note that the processor 713 may temporarily store the write data in the FM 780 without compressing the data, and may compress the data in response to a reclamation process asynchronous with the write command.

Through the above processing, the drive controller 710 can store the data in the cache area 2000 in advance based on the prefetch command received from the storage controller 100. When the drive controller 710 receives a read command from the storage controller 100, if the data corresponding to the read command has already been stored in the cache area 2000, the drive controller 710 can transmit the read data to the storage controller 100 at high speed.

FIG. 20 is a flowchart showing an example of extent prefetch processing in the drive controller 710. This extent prefetching process corresponds to S602 in FIG.

In this embodiment, a case where data is processed inside the drive will be described as an example. Data may be processed inside the drive, such as data compression. When the drive is instructed to prefetch data by the storage controller 100, it is required to respond at high speed to the subsequent read command. For this reason, in order to increase the response speed to the read command, the drive pre-reads the processed data by applying an appropriate process.

The drive controller 710 identifies the LBA in the drive to be prefetched based on the extent number included in the prefetch command received from the storage controller 100 and the configuration information management table 2100 that it has (S700). In other words, the drive controller 710 refers to the configuration information management table 2100 and sets data corresponding to the range of the in-drive LBA corresponding to the extent number as a prefetch target. In this embodiment, since the prefetch command specifies both user data and parity data, the drive controller 710 sets both user data and parity data as prefetch targets.

The drive controller 710 transfers the data stored in the physical storage area corresponding to the pre-read target LBA in the FM 780 to the cache area 2000 (S701). At this time, the drive controller 710 determines whether or not the prefetch target data is compressed data (S702).

When the prefetch target data is compressed data (S702: Yes), the drive controller 710 decompresses the compressed data using the compression / decompression circuit 728 and stores it in the cache area 2000 (S703), and ends the processing (S703). END). Since this process is the same as the above-described read process, details are omitted. When the prefetch target data is not compressed data (S702: No), the drive controller 710 ends the process as it is (END).

Through the above processing, the drive controller 710 can store data to be read later in the cache area 2000 based on the pre-read command.

Example 1 has the following effects.

(1) If the storage controller 100 transmits a prefetch command to each of the drive controllers 710 of the plurality of drives that are the basis of the extent before executing the extent movement process (before transmitting the read command), the drive controller 710 Then, it is possible to pre-read data to be read in accordance with the read command in the extent movement process thereafter in its own cache area 2000. In other words, it is possible to cause each drive that is the basis of one extent to prefetch a plurality of data in the extent with one command. Thereby, the response performance of the storage controller 100 reading data from the drive is improved. Therefore, the time required for the extent movement process can be shortened.

(2) The drive controller 710 stores data in the cache area 2000 based on the prefetch command, so that useless data is not stored in the cache area 2000. For example, when IOPS is measured inside the drive and data having a large IOPS is prefetched, a load is applied to the drive controller 710. Therefore, according to the present embodiment, the processing load on the drive controller 710 can be reduced and the finite cache area 2000 can be used efficiently.

In this embodiment, a case will be described in which the storage controller selectively instructs the drive to prefetch data.

Since the storage system 10 has a RAID group, user data and parity data are stored in the drive. The data to be prefetched may be user data or both user data and parity data. However, since the drive cannot discriminate whether the stored data type is user data or parity data, appropriate read-ahead cannot be performed only with the range of logical addresses. If unnecessary read-ahead occurs, the drive performance degrades, leading to performance degradation as a storage system. However, it is overhead that the storage controller 100 designates the data type for each logical address for prefetching.

Therefore, in this embodiment, the storage controller 100 designates the data type in addition to the extent number and instructs the drive to prefetch.

As described above, the storage controller 100 generates parity data within the storage system 10. That is, the host computer 30 does not recognize parity data. In this embodiment, since the storage controller 100 instructs the drive to perform prefetching, it is possible to distinguish between user data and parity data and execute appropriate prefetching.

FIG. 21 is a schematic diagram for explaining the read process in the extent movement process in the second embodiment.

In the second embodiment as compared with the first embodiment, the storage controller 100 reads the user data D1, D2, and D3 from the SSD 701 to the cache area 1000 without reading the parity data P1. Then, the storage controller 100 generates parity data P1 from the read user data D1, D2, and D3 in the cache area 1000 (reference numeral 903).

According to the above configuration, since the parity data P1 is not read, the consumption of I / O resources between the storage controller 100 and the SSD 701 can be reduced. The above configuration is more effective when the extent migration source and migration destination RAID configurations (nD + mP configuration) are different. This is because if the extent migration source and the migration destination RAID configuration are different, the parity data of the migration source cannot be taken over, and it is necessary to regenerate the parity data. For this reason, it is not necessary to read the parity data, and the parity data pre-read by the source drive is wasted. For example, when the migration source RAID configuration is “3D + 1P” and the migration destination RAID configuration is “4D + 1P”, the storage controller 100 generates a new P1 ′ based on the read D1, D2, D3, and D5. To do.

FIG. 22 is a flowchart illustrating an example of extent movement processing (S107) according to the second embodiment. This extent movement process corresponds to S107 in FIG.

The storage controller 100 executes the same processing as S500 to S502 in FIG. 18 (S520 to S522).

The storage controller 100 transmits a prefetch command to the migration source drive (S523). This prefetch command designates the data type to be prefetched in addition to the extent number of the movement source. The data type is information indicating which of “user data and parity data” or “user data only” is prefetched. Here, the migration source drive prefetches either “user data and parity data” or “user data only” based on the extent number and data type included in the prefetch command. For example, when the RAID configuration of the migration source RAID group is the same as the RAID configuration of the migration destination RAID group, the storage controller 100 designates the data type of “user data and parity data”, and the migration source RAID group If the RAID configuration differs from the RAID configuration of the migration destination RAID group, a data type of “user data only” may be designated. Details of this processing will be described later (see FIG. 23).

Further, the storage controller 100 designates the data type of “user data only” even if the RAID configuration of the migration source RAID group and the RAID configuration of the migration destination RAID group are the same, and the storage controller 100 Parity data may be regenerated.

For example, when the load of the storage controller 100 (for example, the operating rate of the processor 104) is smaller than the first threshold and the load of the drive controller 710 (for example, the operating rate of the processor 713) is larger than the second threshold, A data type of “user data only” may be designated. As a result, the parity data read processing in the drive does not occur, and the load on the drive is reduced. For example, when the load of the storage controller 100 (for example, the operating rate of the processor 104) is equal to or higher than the first threshold and the load of the drive controller 710 (for example, the operating rate of the processor 713) is equal to or lower than the second threshold, The data type of “user data and parity data” may be designated.

For example, when the processing performance (for example, performance based on IOPS or access frequency) of the storage controller 100 is larger than the third threshold and the I / O performance of the drive is lower than the fourth threshold, “user data only” May be specified. In this case, the time for generating the parity data in the storage controller 100 can be shorter than the time for reading the parity data from the drive.

The storage controller 100 sends a read command for user data belonging to the migration source extent to the migration source drive (S524),

The storage controller 100 reads only the user data belonging to the migration source extent and stores it in the cache area 1000 (S525).

Storage controller 100 generates parity data from the user data after the user data in the stripe column is prepared in cache area 1000 (S526).

The storage controller 100 writes the user data and parity data to the extent of the movement destination (S527). The storage controller 100 executes processing similar to S507 to S509 in FIG. 18 (S528 to S530).

Through the above processing, the storage controller 100 can move the user data belonging to the migration source extent to the migration destination extent.

FIG. 23 is a flowchart illustrating an example of extent pre-read processing in the drive controller 710 regarding the second embodiment. This extent prefetching process corresponds to S602 in FIG.

The drive controller 710 identifies the LBA in the drive to be prefetched based on the extent number included in the prefetch command received from the storage controller 100 and the configuration information management table 2100 that it has (S720).

The drive controller 710 determines whether the prefetching method included in the prefetching command is “user data only” or “user data and parity data” (S721).

When the data type is “user data and parity data” (S721: user data and parity data), the drive controller 710 stores the user data and parity data stored in the pre-read target LBA in the FM 780 in the cache area 2000. (S723), and the process proceeds to S725.

When the data type is “user data only” (S721: user data only), the drive controller 710 transfers only the user data stored in the pre-read target LBA in the FM 780 to the cache area 2000 (S722). The process proceeds to S725.

The drive controller 710 executes the same processing as S702 to S703 in FIG. 20 (S725 to S726).

With the above processing, the drive controller 710 performs processing for prefetching only user data belonging to the extent and processing for prefetching all user data and parity data belonging to the extent based on the prefetching command transmitted from the storage controller 100. Can be switched.

Example 2 has the following effects in addition to the effects of Example 1.

(1) Since the storage controller 100 does not read parity data from the drive, it is possible to suppress consumption of I / O resources between the storage controller 100 and the drive.

(2) The drive controller 710 determines whether or not the parity data is to be prefetched based on the prefetch command, so that useless data is not stored in the cache area 2000. That is, the load on the processor 713 for data caching is reduced, and the performance degradation for prefetching is suppressed. Further, the use efficiency of the cache area 2000 of the memory 716 is improved, and more necessary data can be stored in the cache area 2000 having a limited capacity.

The above-described embodiments of the present invention are examples for explaining the present invention, and are not intended to limit the scope of the present invention only to those embodiments. Those skilled in the art can implement the present invention in various other modes without departing from the gist of the present invention.

The storage controller 100 may transmit a prefetch command including all the extent numbers of the migration source before transmitting the first read command associated with the extent migration processing. As a result, the storage controller 100 can reduce the number of times the command is issued. Alternatively, the storage controller 100 may include information related to the prefetch command in the first read command. In this case, the drive controller 710 performs prefetching based on information related to the prefetch command included in the first read command. The data specified by the first read command cannot be prefetched, but the number of commands issued can be further reduced. Alternatively, each time a predetermined number of read commands are transmitted, the storage controller 100 may transmit a prefetch command including the next one or two or more migration source extent numbers. Thereby, the cache area 2000 inside the drive can be used efficiently. This is because the capacity of the cache area 2000 is limited and is used in processes other than prefetching, and even if a large amount of prefetching is instructed at one time, the cache area 2000 may not be stored in the cache area.

The above-described configuration related to the prefetch command can be used for purposes other than extent movement processing between layers. For example, the configuration related to the prefetch command described above is applied to wear leveling processing between drives in extent units, which is executed for extending the life of SSD as described in the patent document “WO2013 / 118170”. May be.

In at least one of the first and second embodiments, the storage controller may set a data type indicating whether the write target data according to the write command is user data or parity data in the write command for the drive. . The drive may manage the data type set in the received write command in association with the write destination LBA (at least one of the in-extent LBA and the in-drive LBA). Thus, the drive can distinguish whether the stored data is user data or parity data for each LBA in the extent.

In the above description, there is a description of information in the expression “xxx table”, but the information may be expressed in any data structure. That is, “xxx table” can be referred to as “xxx information” to indicate that the information does not depend on the data structure.

10: Storage system 30: Host computer 100: Storage controller 710: Drive controller 70, 80: Extent 500, 600: RAID group 1000: Cache area 2000: Cache area 1100: Drive management table 1200: RAID management table 1300: Pool management table 1400: Extent management table 1500: Virtual volume management table 2100: Configuration information management table

Claims

A plurality of semiconductors each having a plurality of nonvolatile semiconductor memory chips, a cache memory that temporarily stores data, and a processor that reads data from the plurality of nonvolatile semiconductor memory chips and stores the data in the cache memory A storage device;
A storage controller that controls the plurality of semiconductor memory devices as a first RAID group and stores each of a plurality of user data and parity data generated from the plurality of user data in the plurality of semiconductor memory devices;
The processor provides storage areas of the plurality of nonvolatile semiconductor memory chips in association with a logical address space to the storage controller;
The storage controller
Configure multiple extents that are logical storage areas,
Dividing each logical address space of the plurality of semiconductor memory devices into a plurality of sub-logical storage areas;
Each of the plurality of extents is associated one by one from the plurality of sub-logical storage areas of each of the plurality of semiconductor storage devices,
The configuration information indicating the correspondence between the arrangement of user data and parity data in the logical address space and the plurality of extents is transmitted to each of the plurality of semiconductor storage devices,
A pre-read command is sent to each of the plurality of semiconductor memory devices to designate the first extent of the plurality of extents and to store data in the cache memory,
The processor of the semiconductor memory device identifies user data and parity data corresponding to the first extent based on the configuration information in response to receiving the prefetch command, and identifies the identified user data and parity data. A storage system that reads from the plurality of nonvolatile semiconductor memory chips and stores them in the cache memory.
The storage controller
Transmitting each of the plurality of semiconductor memory devices to the prefetch command instructing to store user data in the cache memory by designating a second extent of the plurality of extents;
The processor of the semiconductor memory device identifies user data corresponding to a specified extent based on the configuration information in response to receiving the prefetch command, and identifies the identified user data as the plurality of nonvolatile semiconductors The storage system according to claim 1, wherein the storage system is read from a memory chip and stored in the cache memory.
The storage controller
And controlling a second RAID group having a different RAID configuration from the first RAID group,
When the data stored in the second extent belonging to the first RAID group is moved to the second RAID group, the prefetch command instructing to store the user data in the cache memory is transmitted. The storage system according to claim 2.
The storage controller
The storage system according to claim 3, wherein the access frequency of each of the plurality of extents is measured, and the extent to which data is to be moved is determined based on the access frequency.
The processor determines whether the data is compressed when storing the data in the cache memory based on the prefetch command. If the data is compressed, the processor decompresses the data. The storage system according to claim 4, wherein the storage system is stored in the cache memory.
The configuration information includes a size of the extent, a size of a unit for generating the parity data, a RAID level of the RAID group, and a position of the semiconductor memory device in the RAID group. The storage system according to claim 5.
The storage system according to claim 6, wherein the cache memory is configured by a storage medium that can be accessed at a higher speed than the nonvolatile semiconductor memory chip.
A plurality of nonvolatile semiconductor memory chips, a cache memory that temporarily stores data, and a processor that reads data from the plurality of nonvolatile semiconductor memory chips and stores the data in the cache memory,
The processor provides a storage area of the plurality of nonvolatile semiconductor memory chips to an external device in association with a logical address space;
Configuration information indicating correspondence between the arrangement of user data and parity data in the logical address space and a plurality of sub logical storage areas obtained by dividing the logical address space managed by the external device is received from the external device. And
When receiving a prefetch command for instructing to store in the cache memory the data specifying the first sub logical storage area of the plurality of sub logical storage areas, the first sub logical storage area Corresponding user data and parity data are specified, and the specified user data and parity data are read from the plurality of nonvolatile semiconductor memory chips and stored in the cache memory.
If the prefetch command designates a second sub logical storage area and designates storing user data in the cache memory, the processor determines the designated second sub logical storage area based on the configuration information. 9. The semiconductor memory device according to claim 8, wherein user data corresponding to a logical storage area is specified, and the specified user data is read from the plurality of nonvolatile semiconductor memory chips and stored in the cache memory.
The processor determines whether the data is compressed when storing the data in the cache memory based on the prefetch command. If the data is compressed, the processor decompresses the data. The semiconductor memory device according to claim 9, wherein the semiconductor memory device is stored in the cache memory.
The configuration information includes a size of the extent, a size of the parity data, a RAID level of the RAID group, and a position of the semiconductor memory device in the RAID group. The semiconductor memory device described.
The semiconductor memory device according to claim 11, wherein the cache memory is configured by a storage medium that can be accessed at a higher speed than the nonvolatile semiconductor memory chip.