JP2011204285A - Storage system and control method of the same - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a storage system which solves a bottleneck in the storage system and efficiently uses a band provided by hardware, and in which high reliability can be obtained.SOLUTION: The storage system is a storage system provided with a storage device which stores data, a controller which controls an input and an output of data to the storage device, and an interface which connects the storage device with the controller. The storage device has a plurality of physical ports connected with the interface. The controller provides a plurality of storage domains by dividing the storage domain of the storage device into the plurality of storage domains logically, and assigns one or a plurality of physical ports to the storage domain which is divided logically.

Description

  The present invention relates to a storage system, and more particularly to a technique for increasing the speed of a storage system and improving reliability.

  In storage systems (disk array devices), the capacity of groups (RAID groups, storage pools) composed of storage modules increases as the capacity of storage modules such as magnetic disk drives (HDDs) and semiconductor storage devices (SSDs) increases. is doing.

  On the other hand, in many cases, the upper limit of the capacity of a logical volume (LU) configured on a group configured by storage modules is 2 TB. The upper limit of this capacity is because the LBA designation area that is a parameter of the READ (10) command and the WRITE (10) command in the SCSI protocol that is normally used in the block access type storage apparatus is 4 bytes. Variable length READ (16) command and WRITE (16) command are defined as commands that expand the LBA area, but many storage systems do not support these commands.

  The disadvantage of including multiple LUs in the same group is access to the same storage device even if it is logically access to different LUs. A bottleneck will occur.

  Currently, magnetic disk drives mainly used as storage devices are on the order of milliseconds, and have a waiting time due to the rotation of the disk and a seek time of the head. For this reason, the performance due to mechanical factors has not caught up with the performance improvement of the interface and controller. In other words, mechanical factors are performance bottlenecks.

  On the other hand, a semiconductor memory device (SSD) can read and write data at high speed because there is no mechanical waiting time. However, since the semiconductor memory device is designed with an emphasis on compatibility with a magnetic disk drive, the read / write performance of the semiconductor cannot be fully utilized. For example, there is a SAS controller with 8 ports mounted, and one SAS interface can transfer data at 6 Gbps, so the total bandwidth is 8 × 6 Gbit / s = 48 Gbit / s. Since SAS has a function of bundling a plurality of links and transferring data (referred to as a wide port or a wide link), it is possible to make full use of this bandwidth.

As a method of transferring data by bundling a plurality of physical links, Patent Document 1 describes a storage system that controls how many physical links of wide links are assigned to which SAS target device by a SAS initiator device. Has been. Patent Document 2 describes a technique for transferring data using a wide port in which a plurality of physical links are collected.
JP 2007-259933 A US Patent Application Publication No. 2008/0104264 JP 2005-165443 A

  The SAS interface magnetic disk drive has a two-port connector, and since these two ports are used redundantly, the data transfer performance of one magnetic disk drive is 6 Gbit / s. This transfer rate is sufficient for a magnetic disk drive, but not sufficient for a semiconductor memory device. That is, since the semiconductor memory device can read and write data at high speed, the configuration of the interface may become a bottleneck.

  However, in the technique described in Patent Document 1 described above, there are two target SAS ports, an active port and a redundant port.

  Storage logical partitioning (LPAR) that logically divides the resources of the storage device such as the processing capacity and storage capacity of the controller and improves the performance of the storage device and provides each divided logical partition to the user independently There is technology. In such operation, in order to provide each user with high performance, it is necessary to eliminate a bottleneck in the logical partition.

  In this regard, the technology described in Patent Document 1 described above does not consider logical division on the target side.

  Furthermore, in a storage system in which a plurality of storage devices having different reliability and performance are mixed, the performance and reliability that become the bottleneck become the performance and reliability of the storage system. (Verification) also needs to be considered.

  Specifically, in a conventional storage system, after writing data, the written data is read, and the write data and the read data are compared for each byte. Alternatively, the verification code of the write data is compared with the verification code of the read data. However, the conventional method described above consumes a transfer band for reading data, and consumes the processing capacity of the processor for comparing data, and thus has a problem of hindering performance improvement. This deterioration in performance becomes a serious problem in a storage device capable of high-speed processing such as a semiconductor storage device.

  An object of the present invention is to provide a storage system that eliminates bottlenecks in a storage system, efficiently uses a bandwidth provided by hardware, and obtains high reliability.

A typical example of the present invention is as follows. That is, a storage system including a storage device that stores data, a controller that controls input and output of data to and from the storage device, and an interface that connects the storage device and the controller, wherein the storage device includes: The controller includes a plurality of physical ports connected to the interface, and the controller logically divides a storage area of the storage device into a plurality of storage areas, and logically divides one or a plurality of the physical ports. It is characterized by allocating to a storage area.
Another example of the present invention is a storage device that stores data, a cache memory that temporarily stores data written to the storage device, a controller that controls input / output of data to the storage device, A storage system comprising an interface for connecting the storage device, the cache memory, and the controller, wherein the controller writes data for each data write block when transferring data from the cache memory to the storage device. A verification code for ensuring the validity of the generated data is generated, the data with the generated verification code added is transferred to the storage device, and added to the data written in a plurality of blocks of the storage device Append to the data of one or more blocks from the first block in the verification code Without using a verification code, characterized by verifying the data written in the storage device.

  According to the embodiment of the present invention, the performance of the storage system can be improved.

It is explanatory drawing which shows the outline | summary of the 1st Embodiment of this invention. It is explanatory drawing which shows the outline | summary of the 1st Embodiment of this invention. 1 is a front view of a storage system according to a first embodiment of this invention. It is a figure which shows the internal structure of the state which opened the front door of the controller housing | casing of the 1st Embodiment of this invention. It is a figure which shows the internal structure of the state which opened the back door of the controller housing | casing of the 1st Embodiment of this invention. 1 is a block diagram illustrating a configuration of a storage system according to a first embodiment of this invention. 1 is a front view of a semiconductor memory device according to a first embodiment of the present invention. 1 is a perspective view showing a structure of a case of a semiconductor memory device according to a first embodiment of the present invention. 1 is a perspective view showing an internal structure of a semiconductor memory device according to a first embodiment of the present invention. 1 is a block diagram showing a configuration of a semiconductor memory device (base substrate) according to a first embodiment of the present invention. It is a figure which shows the example of arrangement | positioning of the interface connector of the memory | storage device of the 1st Embodiment of this invention. It is a figure which shows the example of arrangement | positioning of the interface connector of the memory | storage device of the 1st Embodiment of this invention. It is a figure which shows the example of arrangement | positioning of the interface connector of the memory | storage device of the 1st Embodiment of this invention. It is explanatory drawing which shows the example of the logical division of the storage area of the 1st Embodiment of this invention, and the allocation of the path | pass (bandwidth) of the logically divided storage area. It is explanatory drawing which shows the example of the logical division of the storage area of the 1st Embodiment of this invention, and the allocation of the path | pass (bandwidth) of the logically divided storage area. It is explanatory drawing which shows the example of the logical division of the storage area of the 1st Embodiment of this invention, and the allocation of the path | pass (bandwidth) of the logically divided storage area. It is explanatory drawing which shows the example of the logical division of the storage area of the 1st Embodiment of this invention, and the allocation of the path | pass (bandwidth) of the logically divided storage area. It is explanatory drawing which shows the example of the logical division of the storage area of the 1st Embodiment of this invention, and the allocation of the path | pass (bandwidth) of the logically divided storage area. It is explanatory drawing which shows the example of the logical division of the storage area of the 1st Embodiment of this invention, and the allocation of the path | pass (bandwidth) of the logically divided storage area. FIG. 9B is an explanatory diagram of configuration information for realizing the logical division and path allocation shown in FIGS. 9A to 9F. It is a figure explaining the example of the link set in the storage system of the 1st Embodiment of this invention. It is a figure explaining the example of the link set in the storage system of the 1st Embodiment of this invention. It is a figure explaining the example of a link set in the storage system of the 1st Embodiment of this invention. It is a figure explaining the example of the link set in the storage system of the 1st Embodiment of this invention. It is a figure explaining the example of the link set in the storage system of the 1st Embodiment of this invention. It is explanatory drawing which shows the structure of the management table with which the protocol control part of the disk control part of the 1st Embodiment of this invention is provided. It is explanatory drawing which shows the structure of the management table with which the protocol control part of the SAS expander of the 1st Embodiment of this invention is equipped. It is a flowchart of the process performed in the processor and disk control part of the controller housing | casing of the 1st Embodiment of this invention. It is a flowchart of the process performed in the SAS expander of the controller housing | casing of the 1st Embodiment of this invention. It is a flowchart of the path | pass change process at the time of the failure generation of the 1st Embodiment of this invention. It is a figure explaining the outline | summary of the 2nd Embodiment of this invention. It is a figure explaining the outline | summary of the 2nd Embodiment of this invention. It is explanatory drawing which shows the error detection model of the 2nd Embodiment of this invention. It is a flowchart of the data verification process of the 2nd Embodiment of this invention.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

<Embodiment 1>
1A and 1B are explanatory diagrams showing an overview of the first embodiment of the present invention.

  In general, the storage device 22 includes one or more physical ports for inputting and outputting data. For example, as illustrated in FIG. 1A, a storage device including a SAS interface includes two physical ports. One of these ports is used for the active system and the other is used for the standby system. Then, the active and standby paths for transferring data to and from the disk control unit are set using each port. For this reason, even if the storage area of the storage device is logically divided, the divided storage area is not associated with the port.

  On the other hand, as shown in FIG. 1B, since the storage device of the first embodiment includes eight physical ports, four ports can be used for the active system, and the other four ports can be used. Can be used for standby systems. Then, a path for transferring data to and from the disk control unit is set using each port. In this case, each of the four ports can be assigned to each logically divided storage area, and data can be transferred independently to each storage area. In addition, four ports divided in time can be assigned to one storage area, and data can be transferred at high speed.

  The storage device 22 according to the first embodiment includes an LED 223 that displays the load of the physical port. Furthermore, an interface (SD memory card slot 222) for attaching an auxiliary storage device (for example, SD memory card 1000) that provides a storage area used as a replacement area is provided. For this reason, even when the spare area is insufficient, the storage device can be extended by setting the storage area of the SD memory card 1000 as the spare area.

  FIG. 2 is a front view of the storage system according to the first embodiment of this invention.

  FIG. 2 shows one controller housing 1 and three drive housings 2. The controller housing 1 and the drive housing 2 and the drive housing 2 are connected by a SAS interface.

  The controller casing 1 and the drive casing 2 have the same appearance, but the appearances of the two may be different.

  FIG. 3A is a diagram illustrating an internal structure in a state where the front door of the controller housing 1 according to the first embodiment of the present invention is opened.

  The front side of the controller housing 1 is provided with a slot in which the semiconductor storage device 22 (or magnetic disk drive 21) is attached. In addition, a power supply unit 1010 is provided at the lower part on the front side of the controller housing 1. The semiconductor storage device 22 attached to the controller housing 1 is connected by a backplane (not shown).

  FIG. 3B is a diagram illustrating the internal structure of the controller housing 1 according to the first embodiment of the present invention in a state where the back door is opened.

  A channel control unit 100 and a disk control unit 200 are provided in the lower part on the back side of the controller housing 1. A controller unit 1020 including a processor 400, a memory 430, and a shared memory 450 is provided on the channel control unit 100 and the disk control unit 200. Note that a portion other than the channel control unit 100, the disk control unit 200, and the controller 1020 may be provided with a slot in which the semiconductor memory device 22 (or the magnetic disk drive 21) can be attached.

  The structure on the front side of the drive casing 2 is the same as the configuration on the front side of the controller casing 1 shown in FIG. 3A. Further, the structure on the back side of the drive housing 2 is provided with a slot in which the semiconductor memory device 22 (or the magnetic disk drive 21) is attached, similarly to the structure on the front side of the controller housing 1 shown in FIG. 3A.

  FIG. 4 is a block diagram showing a configuration of the storage system according to the first embodiment of this invention.

  The storage system (controller housing 1) of this embodiment includes a plurality of channel controllers 100, a plurality of disk controllers 200, a cache memory 300, a processor 400, a memory 430, and a shared memory 450. The channel controller 100, disk controller 200, cache memory 300, processor 400, memory 430, and shared memory 450 are connected by a data controller 500. As shown in the figure, the storage system 1 normally includes a plurality of channel controllers 100 and a plurality of disk controllers 200, but each of these components may include only one.

  The cache memory 300, the processor 400, the memory 430, the shared memory 450, and the data controller 500 constitute a controller unit 1020.

  The channel control unit 100 is a physical interface that is a driver that inputs and outputs signals according to the communication protocol with the host computer 10, and a protocol control that controls input and output of signals according to the communication protocol with the host computer 10. And a transfer control unit that controls an address of data transferred from the channel control unit 100 to the cache memory 300. The channel control unit 100 is connected to the host computer 10 by, for example, a SAN 11 that communicates using a fiber channel protocol. The channel control unit 100 and the host computer 10 may be connected by Ethernet (registered trademark). In this case, the channel control unit 100 and the host computer 10 communicate with each other using the iSCSI protocol.

  The disk control unit 200 is a physical interface that is a driver that inputs and outputs a signal according to a communication protocol between the storage devices 21 and 22, and a signal input / output according to a communication protocol between the storage devices 21 and 22. A protocol control unit for controlling and a transfer control unit for controlling an address of data transferred from the disk control unit 200 are provided. In the storage system according to the present embodiment, the disk controller 200 includes a 4-port SAS (Serial Attached SCSI) interface. The connection between the disk control unit 200 and the storage devices 21 and 22 may be another connection method as long as a multi-port connection that transfers data using a plurality of ports as one link is possible.

  In the storage system 1 of the present embodiment, the magnetic disk drive 21 and the semiconductor storage device 22 are used as the storage devices 21 and 22. However, in addition to these, an optical disk, a magnetic tape, or the like may be used. 4 shows the storage devices 21 and 22, the storage system may be provided with the storage devices 21 and 22 or may be connected to another subsystem having the storage devices 21 and 22 ( It may be in the form of a general NAS or VTL).

  The cache memory 300 temporarily stores data requested to be written by the host computer 10 and temporarily stores data read from the storage devices 21 and 22.

  The processor 400 controls the operation of the storage system (for example, an instruction to transfer data from the channel control unit 100 to the cache memory 300).

  The memory 430 stores a program executed by the processor 400 and data required when the program is executed.

  The shared memory 450 stores data (for example, storage system configuration information) that is commonly used by each unit.

  The data controller 500 includes a switch (for example, a crossbar switch) that connects the channel control unit 100, the disk control unit 200, the cache unit 300, the processor 400, the memory 430, and the shared memory 450. The data controller 500 constitutes an internal network connecting the channel control unit 100, the disk control unit 200, the cache unit 300, the processor 400, the memory 430, and the shared memory 450. Data is transferred between these units by the internal network.

  The storage system may include an interface (SVP) to which a maintenance terminal is connected.

  When the channel control unit 100 receives a data write request from the host computer 10, the channel control unit 100 converts the received write request into a protocol in the storage system, analyzes the address of the write request converted by the protocol, and displays the analysis result. Sent to the processor 400. The processor 400 determines a transfer destination of the write request (data) based on the analysis result of the address, and instructs the data controller 500 of the transfer destination. The data controller 500 transfers data by switching the switch so that the data is transferred to the transfer destination instructed by the processor 400.

  A write request (data) from the host computer 10 is transferred from the channel control unit 100 to the cache memory 300 and temporarily stored in the cache memory 300. When the data is stored in the cache memory 300 (even before the data is stored in the storage devices 21 and 22), a report of the completion of writing is returned to the host computer 10, so the response to the host computer 10 Can be improved.

  Data stored in the cache memory 300 is transferred to the disk controller 200. The disk control unit 200 converts the transferred data into a protocol (for example, SAS protocol) for transferring to the storage devices 21 and 22. Data obtained by converting the protocol is written in the storage devices 21 and 22.

  The storage system 1 of the present embodiment includes a SAS expander 600, a magnetic disk drive 21, and a semiconductor storage device 22 in addition to the controller unit.

  The SAS expander 600 includes a switch (for example, a crossbar switch) that connects the disk control unit 200, the magnetic disk drive 21, and the semiconductor storage device 22, and the disk control unit 200 and the magnetic disk drive 21 (or the semiconductor storage device 22). Transfer data to and from. The SAS expander 600 connects to the storage devices 21 and 22 using the number of ports corresponding to the storage device to which it is connected. For example, the SAS expander 600 and the semiconductor memory device 22 are connected by 4 ports for the active system or 4 ports for the standby system (8 ports in total).

  In addition, the number of ports connected between the SAS expander 600 and the magnetic disk drive 21 differs depending on the interface provided in the magnetic disk drive 21. For example, the magnetic disk drive 21 having the SAS interface and the SAS expander 600 are connected by one port for each of the active system and the standby system. In addition, the magnetic disk drive 21 having a SATA (Serial Advanced Technology Attachment) interface and the SAS expander 600 are connected by one port of the active system or one port of the standby system. Since the SATA magnetic disk drive 21 has only one port, a port applicator 610 for switching the number of ports is provided at the interface of the magnetic disk drive 21, and one port for the active system or one port for the standby system (two ports in total) Is connected to the SATA magnetic disk drive 21.

  The SAS expander 600 is connected to the SAS expander 600 of another housing (for example, the drive housing 2) by a SAS interface.

  The magnetic disk drive 21 records data on the magnetic disk. The magnetic disk drive 21 is connected to the disk control unit 200 via the SAS expander 600 by a SAS interface.

  The semiconductor memory device 22 is a so-called SSD (Solid State Disk) including a non-volatile memory (for example, a flash memory), and is connected to the disk controller 200 via a SAS interface. The memory elements constituting the semiconductor memory device 22 are not limited to flash memories, and may be semiconductor elements such as M-RAM, P-RAM, STT-RAM, racetrack memory, and the like.

  The drive chassis 2 of the storage system according to the present embodiment includes a SAS expander 600 and a magnetic disk drive 21. The drive housing 2 illustrated in FIG. 4 includes only the magnetic disk drive 21, but may include a semiconductor storage device 22 in addition to the magnetic disk drive 21.

  FIG. 5 is a front view of the semiconductor memory device 22 according to the first embodiment of this invention.

  The semiconductor storage device 22 includes a latch release lever 221, an SD memory card slot 222, and a status display LED 223 on the front surface thereof. The semiconductor memory device 22 includes a front substrate, a base substrate, an additional memory substrate 1 and an additional memory substrate 2 therein (see FIG. 6).

  The latch release lever 221 is operated when the semiconductor memory device 22 is removed from the controller housing 1 (or the drive housing 2).

  The SD memory card slot 222 is an interface for mounting an SD memory card 1000 that provides a replacement area for the semiconductor memory device 22. Note that the data written to the SD memory card 1000 is written after being encrypted by the control unit of the semiconductor memory device 22.

  The status display LEDs 223 include red / blue / green as a set, and four LEDs on the left and right (a total of eight). Each status display LED 223 displays the load status of each of the eight ports. For example, light up red for high loads, light yellow (red: normal + green: normal) for medium loads, light green for light loads, and blue for very light loads. Light up.

  Further, the operation state of the semiconductor memory device 22 may be displayed by the status display LED 223 and used for maintenance purposes. For example, when a failure occurs in the semiconductor memory device 22, it can be flashed in red, and when the remaining capacity of the replacement area is less than or equal to the threshold value, it can be flashed in orange (red: normal + green: dimmed). . In this case, the user installs the SD memory card 1000 in the SD memory card slot 222 of the semiconductor storage device 22 in which the status display LED 223 is blinking orange, so that the semiconductor storage device 22 is not immediately replaced without replacing the semiconductor storage device 22. The storage device 22 can be continuously operated.

  In this way, by providing the status display LED 223 at a position that can be seen from the front side of the semiconductor storage device 22, the user opens the front door of the housing to thereby load the semiconductor storage device 22 mounted in the storage system. The status can be confirmed at a glance, and the volume load can be understood intuitively. For this reason, it is easy to plan a volume relocation for distributing the load.

  In addition, a plurality of air introduction holes 224 are provided on the front surface of the semiconductor memory device 22.

  FIG. 6A is a perspective view showing the structure of the case of the semiconductor memory device 22 according to the first embodiment of the present invention.

  The semiconductor memory device 22 includes an air introduction hole 224 on the front surface and an air discharge port 225 on the back surface. For example, by providing an exhaust fan in the central portion of the controller housing 1 so that the airflow in the controller housing 1 is controlled to flow from the front surface to the central portion and from the back surface to the central portion, from the front surface of the semiconductor memory device 22. Air flows on the back.

  In this way, air introduced from the front surface of the HDD module on which the magnetic disk drive 21 is mounted flows on the outer periphery of the HDD, whereas air can be introduced into the semiconductor memory device 22.

  The semiconductor memory device 22 includes a hole 226 for inserting the SD memory card 1000 at the lower part of the front surface.

  FIG. 6B is a perspective view showing the internal structure of the semiconductor memory device 22 according to the first embodiment of the present invention.

  The semiconductor memory device 22 includes a base substrate 2201, a front substrate 2202, and additional memory substrates 2203, 2204, and 2205).

  The base board 2201 includes an interface connector 228, a nonvolatile memory 2225 (see FIG. 7), control units 2222 and 2223 (see FIG. 7), a connector 2211, and stacking connectors 2212 and 2213. Connector 2211 is connected to front substrate 2202. The stacking connector 2212 is connected to the additional memory board 2203. The stacking connector 2213 is connected to the additional memory board 2204.

  The front substrate 2202 is attached to the base substrate 2201 with a connector 2211. The front substrate 2202 includes four state display LEDs 223 (four in total) on the left and right. Further, the front substrate 2202 includes an SD memory card slot 222 at the bottom.

  The additional memory board 2203 is attached to the base board 2201 by the stacking connector 2212. The additional memory board 2204 is attached to the base board 2201 by the stacking connector 2213. The expansion memory board 2205 is attached to the expansion memory board 2203 by the stacking connector 2214. The additional memory boards 2203, 2204, and 2205 include nonvolatile memory elements.

  The stacking connectors 2212, 2213, and 2214 are arranged along the direction from the front surface to the back surface of the semiconductor memory device 22. Since the semiconductor memory device 22 of the present embodiment has a structure that allows air to flow through the inside thereof, the stacking connector is arranged so that the longitudinal direction of the stacking connector is parallel to the air flow so as not to block the air flow. Has been.

  FIG. 7 is a block diagram showing a configuration of the semiconductor memory device 22 (base substrate 2201) according to the first embodiment of the present invention.

  The semiconductor storage device 22 according to the first embodiment includes an interface connector 228, a physical interface 2221, a protocol control unit 2222, a processor 2223, a memory 2224, a nonvolatile memory 2225, an SD memory card slot 222, a status display LED 223, and a power connector 229. A backup power source 2226 and an additional memory connection connector 2213.

  The interface connector 228 is a high-density connector that accommodates an 8-port SAS interface, as will be described later with reference to FIGS. 8B and 8C.

  The physical interface 2221 is a driver for inputting and outputting signals according to the SAS protocol. The protocol control unit 2222 is a controller for controlling signal input / output in accordance with the SAS protocol.

  In the SAS standard, up to eight physical ports can be accessed with the same SAS address. Therefore, it is possible to control the ports (active system 4 ports, standby system 4 ports) of the semiconductor storage device 22 of the present embodiment by the same logical interface as the storage device having two conventional physical ports. Note that when 8 ports are used in the active system, 8 ports are used in the standby system, so two SAS addresses are required in the active system and the standby system. Note that a SAS address may be set for each link.

  The processor 2223 controls the operation of the semiconductor memory device 22, for example, input / output of data to / from the nonvolatile memory 2225 by executing a program stored in the memory 2224. The memory 2224 stores a program executed by the processor 2223 and data necessary for executing the program. The nonvolatile memory 2225 is a memory that stores data transferred to the semiconductor memory device 22.

  The SD memory card slot 222 is an interface for mounting an SD memory card 1000 that provides a replacement area for the semiconductor memory device 22. The status display LED 223 displays the load status of each port of the semiconductor memory device 22 and the operating status of the semiconductor memory device 22.

  The power connector 229 is an interface for supplying power to the semiconductor memory device 22. The backup power supply 2226 includes a secondary battery or a large-capacity capacitor, and supplies power to each part of the semiconductor memory device 22 for a predetermined time after the supply of power to the semiconductor memory device 22 is stopped. Additional memory connection connectors (stacking connectors) 2212 and 2213 are interfaces for connecting the additional memory board 2204.

  8A to 8C are diagrams showing examples of the arrangement of interface connectors of the storage devices (magnetic disk drive 21 and semiconductor storage device 22) according to the first embodiment of this invention.

  As shown in FIG. 8A, the magnetic disk drive 21 having a SAS interface includes a SAS interface connector 227 including two data input / output ports.

  Further, as shown in FIG. 8B, the semiconductor memory device 22 includes a conventional SAS interface connector 227 and a high-density SAS interface connector 228 including a plurality of data input / output ports. In the case shown in FIG. 8B, 8 ports may be connected via the high-density SAS interface connector 228, 2 ports may be connected via the conventional SAS interface connector 227, and 6 ports may be connected via the high-density SAS interface connector 228. Good. Further, power to the semiconductor memory device 22 is supplied via a conventional SAS interface connector 227.

  Further, as shown in FIG. 8C, the semiconductor memory device 22 may include a high-density SAS interface connector 228 including a plurality of data input / output ports, and a power connector 229. In the case shown in FIG. 8C, 8 ports are provided via the high density SAS interface connector 228. In addition, power to the semiconductor memory device 22 is supplied via the power connector 229.

  In other words, in the storage system of the present embodiment, the magnetic disk drive 21 and the semiconductor storage device 22 have the necessary connectors without interfering with the conventional SAS interface connector 227, high-density SAS interface connector 228, and power connector 229. Can be used. In this case, the magnetic disk drive 21 uses only 2 ports, and the semiconductor memory device 22 uses 8 ports.

  9A to 9H are explanatory diagrams illustrating an example of logical partitioning of a storage area and allocation of paths (bandwidths) of the logically partitioned storage area according to the first embodiment of this invention.

  In the case of the semiconductor memory device 22 shown in FIG. 9A, the storage area is not logically divided and is provided as one logical storage area. In this case, the four paths in the working system form one link of 24 Gbit / s and are used as wide link paths for transferring data at high speed.

  In the case of the semiconductor memory device 22 shown in FIG. 9B, the storage area is logically divided into two areas of equal capacity. In this case, the four paths of the working system form a 12 Gbit / s link and are set to two multilink paths that transfer data at high speed, and each 12 Gbit / s path is a logically divided storage area. Assigned to. Therefore, each logically divided storage area can independently transfer data using a bandwidth of 12 Gbit / s. In the present invention, in order to distinguish the technology that divides a link that can be used as a wide link into a plurality of links and enables the divided links to be transferred independently from the conventional wide link. This will be referred to as a multilink.

  In the case of the semiconductor memory device 22 shown in FIG. 9D, the storage area is logically divided into four areas of equal capacity. In this case, the four 6 Gbit / s paths in the active system are assigned to each logically divided storage area. Therefore, each of the logically divided storage areas can independently transfer data using a 6 Gbit / s band.

  In the case of the semiconductor memory device 22 shown in FIGS. 9A, 9B, and 9D, the division ratio of the storage area and the division ratio of the path are equal. However, in the embodiment of the present invention, the storage area division ratio and the path division ratio may not be equal.

  For example, in the case of the semiconductor memory device 22 shown in FIG. 9C, the storage area is logically divided into three areas of equal capacity. In this case, the four paths in the active system are set to one 12 Gbit / s multilink path and two 6 Gbit / s paths. Then, a 12 Gbit / s path is assigned to one logically divided area, and a 6 Gbit / s path is assigned to each of the other two areas. Thus, each logically divided storage area can transfer data using a bandwidth of 12 Gbit / s or 6 Gbit / s allocated independently.

  By assigning a single link or a multilink path in which a plurality of paths are combined to the logically divided storage area in this way, the bandwidth of the entire I / F can be appropriately distributed according to the performance of the storage device. The performance of the semiconductor memory device 22 that can read and write at a higher speed than the magnetic disk drive can be utilized.

  That is, when 4 ports are used in the active system and the storage area is logically divided into four, it is possible to set a port for each storage device up to four storage area divisions. The processor 400 transmits a mode select command to the semiconductor memory device 22 to instruct the logical division number and the address range of each divided area. In accordance with the received mode select command, the semiconductor memory device 22 associates the logically divided area with the port and assigns an identifier (LUN) to each logically divided area.

  When the storage area is divided beyond the number of usable physical ports, it is desirable in terms of performance to specify the storage area of the I / O target on the controller side (processor 400). However, if the processing capacity of the semiconductor memory device 22 is sufficient, the storage device side may determine the storage area of the I / O target.

  In this manner, by making the storage area division ratio and the path division ratio non-uniform, it is possible to provide a storage area that is logically divided to meet the user's request.

  Further, finer performance can be adjusted by combining the logical division of the storage device with the logical division of the storage system.

  Next, an example of logical division of a storage area and assignment of a path (bandwidth) of the logically divided storage area in the case of the magnetic disk drive 21 will be described.

  In the case of the magnetic disk drive 21 shown in FIG. 9E, since only one active path can be set, the path cannot be logically divided unless the standby path is diverted to the active system. Therefore, even when the storage area is divided into two areas, a 6 Gbit / s path (bandwidth) is shared by the two logically divided areas in each logically divided area.

  If the storage area cannot be divided as in the magnetic disk drive 21 shown in FIG. 9F, one active path is assigned to the storage area. Therefore, a 6 Gbit / s path (bandwidth) is used by one area.

  FIG. 10 is an explanatory diagram of configuration information for realizing the logical partitioning and path allocation shown in FIGS. 9A to 9F.

  The configuration information shown in FIG. 10 is managed by the processor 400 and stored in the shared memory 450. This configuration information includes host group 1701, logical partition group 1702, front end port 1703, system processing capacity 1704, cache capacity, mode 1705, back end zoning 1706, volume group 1707, drive type 1708, and logical partition allocation information ( Area information 1709 and path information 1710).

  The host group 1701 is an identifier of the group of the host computer 10 that accesses this logical partition group (storage area). The logical partition group 1702 is an identifier of a logically partitioned storage area. The front-end port 1703 is an identifier of the port of the channel control unit 100 used for accessing this logical partition group, and includes information on the number of ports used for accessing this logical partition group.

  The system processing capacity 1704 is the performance of access processing to this logical partition group. The cache capacity and mode 1705 are the capacity of the cache memory 400 used for accessing this logical partition group and the access mode (random access or sequential access) to the cache memory 400. The back-end zoning 1706 is a zoning type of a path under the disk control unit 200 used for accessing this logical partition group.

  The volume group 1707 is an identifier of a logical volume that constitutes this logical partition group. The drive type 1708 is the type of storage device that constitutes this logical partition group.

  The logical partition allocation information includes area information 1709 and path information 1710. The area information 1709 is a storage capacity constituting this logical partition group. The path information 1710 is a bandwidth of a path used for accessing this logical partition group.

  The assumed load 1711 is an assumption of an access load to the logical partition group set as an example in this table.

  Since the semiconductor memory device 22 can respond at high speed, the allocation of the capacity of the cache memory 400 can be made smaller than that of the magnetic disk drive 21. In the configuration information shown in FIG. 10, the magnetic disk drive 21 is assigned one port because the access performance to the magnetic disk is a bottleneck based on the data transfer speed. However, in the case of a hybrid type magnetic disk drive in which semiconductor memories are mixed, or in the case where the recording density of the magnetic disk is improved and the performance of the magnetic disk drive is improved, a plurality of ports are provided as in the semiconductor memory device 22. It may be assigned.

  FIG. 11 is a diagram illustrating an example of links set in the storage system according to the first embodiment of this invention. FIG. 11 shows only the active links, and the standby links are preferably provided in the same number as the active links shown.

  The protocol control unit 602A on the controller 600A of the controller housing 1 controls eight physical interfaces 601A. Each physical interface 601A provides one physical port, and each port forms one link.

  Of the links output from the controller housing 1, four links to the semiconductor storage device 22 are gathered together to form two links of 24 Gbit / s. On the other hand, the links to the magnetic disk drive 21 form eight links of 6 Gbit / s.

  The SAS expander 600A of the drive housing 2A connected to the controller housing 1 collects four links from the controller housing 1 by four, forms two of 24 Gbit / s, and links 24 Gbit / s. The two semiconductor memory devices 22 are provided.

  The SAS expander 600B of the drive chassis 2A outputs the 6 Gbit / s eight links to the drive chassis 2B as it is. The SAS expander 600C of the drive housing 2B provides eight links of 6 Gbit / s to the magnetic disk drive 21.

  That is, for data transfer to the semiconductor memory device 22, a multilink in which a plurality of links are combined is used. On the other hand, for data transfer to the magnetic disk drive 21, each path is controlled independently without using a multilink. Note that a part of the path to the storage device may be a multilink, and the path may be formed by combining a multilink and an independent link.

  Next, the virtually configured multilink will be described.

  In the present embodiment, four virtual links of a maximum of 12 Gbit / s are configured by virtual multilinks using four ports of 6 Gbit / s each of the semiconductor memory device 22.

  The administrator logically divides the storage capacity of the semiconductor storage device 22 using a management computer connected to the storage system (for example, via a management network), and configures virtual storage for each divided storage area To do.

  The processor 400 of the storage system allocates the resources of the controller unit 1020 (the processing capacity of each control unit, the processing capacity of the processor, and the capacity of the cache) in accordance with an instruction from the administrator, and sets the zoning of the SAS expander 600 To set a physical port pair for virtual multilink. Furthermore, logical division and virtual multilink are also set in the semiconductor memory device 22 to which virtual multilink is applied.

  After the multi-link setting of the SAS expander 600 and the semiconductor storage device 22 is completed, the processor 400 issues a multi-link start command to the SAS expander 600. After receiving the command, the SAS expander 600 starts link negotiation with the subordinate semiconductor memory device 22. The SAS expander 600 starts the operation in the set mode after the completion of the negotiation, and reports to the processor 400 that the operation is possible.

  After receiving the notification, the processor 400 performs parallel processing of the logical partition group to which two pairs that can operate simultaneously (for example, port 0/1 and port 2/3, port 1/2 and port 0/3) belong. The disk control unit 200 and the like are set so that priority is given.

  The disk control unit 200 monitors the processing queue written from the controller, and controls the execution order of commands so as to preferentially execute I / O that can be processed in parallel. As a result, the bandwidth can be used efficiently.

  Further, the SAS expander 600 can form a storage domain for each logical partition while sharing the physical port among the plurality of logical partitions by virtual zoning in which one physical port is assigned to the plurality of logical partitions.

  A plurality of ports between the semiconductor memory device 22 and the SAS expander 600 may be suspended for each logically divided link. Further, the corresponding storage area of the semiconductor memory device 22 may be stopped in response to the suspension of the port (link). Thus, by stopping the operation of the storage system for each logical division, the power consumption can be reduced.

  Further, some of the plurality of ports between the semiconductor memory device 22 and the SAS expander 600 may be suspended according to the load of each logical partition.

  12A to 12D are diagrams for explaining examples of links set in the storage system according to the first embodiment of this invention. In the link examples shown in FIGS. 12A to 12D, one or a plurality of links are time-divided and shared by a plurality of storage areas.

  FIG. 12A shows an example of a virtual multilink that transfers data using time-division links allocated to logically divided storage areas.

  The storage area is divided into four storage areas 1101 to 1104. A port (link) is allocated so as to be shared by a plurality of storage areas. That is, links 1 and 2 are allocated to the storage area 1101, and the data 1111 is transferred to the storage area 1101 by the links 1 and 2. Links 2 and 3 are allocated to the storage area 1102, and the data 1112 is transferred to the storage area 1102 by the links 2 and 3. Links 1 and 4 are allocated to the storage area 1103, and the data 1113 is transferred to the storage area 1103 by the links 1 and 4. Links 3 and 4 are allocated to the storage area 1104, and the data 1114 is transferred to the storage area 1104 through the links 3 and 4.

  In this case, when transferring data to the semiconductor memory device 22, the processor 400 or the disk control unit 200 checks the assigned port to determine whether or not the port is free. Then, the data is transferred so as to form one link using the vacant port.

  Thus, by using each link in a time-sharing manner, data can be transferred at a maximum of 12 Gbit / s in each storage area.

  FIG. 12B shows a state in which data is transferred using the standby link 5 when a failure occurs in the link 1 shown in FIG. 12A.

  Since a failure has occurred in link 1, standby link 5 is assigned to the active system instead of link 1.

  Therefore, links 2 and 5 are allocated to the storage area 1101, and data 1111 transferred to the storage area 1101 is transferred by the links 2 and 5. Links 4 and 5 are allocated to the storage area 1103, and data 1113 transferred to the storage area 1103 is transferred by the links 4 and 5. The allocation of links to the storage areas 1102 and 1104 is the same as that shown in FIG. 12A.

  As described above, in this embodiment, the link can be formed by freely combining the working and standby paths, so that even when a failure occurs in some links, the conventional wide link of the SAS standard is used. As described above, it is not necessary to switch all of the collected links. That is, when a failure occurs in a part of the working path, data can be transferred using a part of the standby system link by switching only the part of the path in which the failure has occurred. For this reason, even when a failure occurs, data can be transferred using the same bandwidth as in normal times.

  FIG. 12C illustrates an example in which data is transferred using the active links 1 to 4 and the standby links 5 to 8 in a normal state.

  Links 1, 2, 7, and 8 are allocated to the storage area 1101, and the data 1111 is transferred to the storage area 1101 by the links 1, 2, 7, and 8. Links 2, 3, 5, and 8 are allocated to the storage area 1102, and the data 1112 is transferred to the storage area 1102 by the links 2, 3, 5, and 8. Links 1, 4, 6 and 7 are allocated to the storage area 1103, and data 1113 is transferred to the storage area 1103 by the links 1, 4, 6 and 7. Links 3, 4, 5 and 6 are allocated to the storage area 1104, and the data 1114 is transferred to the storage area 1104 via the links 3, 4, 5 and 6.

  Thus, when the active and standby links are used in a time-sharing manner, data can be transferred at a maximum of 24 Gbit / s in each storage area, and the data transfer band provided in the storage system can be used effectively. .

  FIG. 12D shows a state in which data transfer is performed by degenerating data transfer to the storage area using link 1 when a failure occurs in link 1 shown in FIG. 12C.

  Since a failure has occurred in link 1, links 7 and 8 are allocated to storage area 1101, and data 1111 is transferred to storage area 1101 by links 7 and 8. Links 6 and 7 are allocated to the storage area 1103, and the data 1113 is transferred to the storage area 1103 by the links 6 and 7. The allocation of links to the storage areas 1102 and 1104 is the same as that shown in FIG. 12C.

  As described above, in the present embodiment, since the link can be formed by freely combining the working and standby paths, all of the collected links are switched as in the case of the conventional SAS standard wide link. There is no need. In other words, it is not necessary to prepare a standby path having the same bandwidth as the active path. Therefore, when a failure occurs in a part of the paths, the paths can be degenerated and data transfer can be continued.

  FIG. 13A is an explanatory diagram illustrating a configuration of a management table 1200 managed by the processor 400 according to the first embodiment of this invention.

  The management table 1200 shows the relationship between the logically partitioned storage areas and the ports of the disk controller 200, and includes a logical partition group 1201, a virtual multilink 1202, a target port 1203, and a disk group 1204.

  The logical partition group 1201 is an identifier of the logical partition of the logically partitioned storage system. The virtual multilink 1202 is the number of links used as a virtual multilink. The target port 1203 is a port identifier of the disk control unit 200. The disk group 1204 is an identifier of a port of the SAS expander 600 (a storage device connected to this port).

  Note that an entry 1205 of the management table 1200 indicates port assignment at the normal time, and an entry 1206 indicates port assignment when a failure occurs in the port A0.

  FIG. 13B is an explanatory diagram illustrating a configuration of the management table 1210 managed by the SAS expander 600 according to the first embodiment of this invention.

  The management table 1210 shows the relationship between the logically divided storage areas and the ports of the SAS expander 600, and includes mapping information of the physical port 1211, the virtual zone group 1212, the virtual multilink 1213, and the physical link 1214.

  The physical port 1211 is a port identifier of the SAS expander 600. The virtual zone group 1212 is an identifier of a logical partition of the logically divided storage system. The virtual multilink 1213 is the number of links used as a virtual multilink. The physical link 1214 is a physical link identifier.

  As can be seen from the management table 1210, the virtual link corresponds to one zone, and the physical link is shared by a plurality of zones.

  FIG. 14 is a flowchart of processing executed in the processor 400 and the disk control unit 200 of the controller housing 1 according to the first embodiment of this invention.

  First, upon receiving an I / O request (write request, read request) from the host computer 10, the processor 400 stores a command for executing I / O to the storage devices 21 and 22 in the processing queue, The disk controller 200 is instructed that the command is stored in the processing queue (S101).

  After receiving the instruction that the command is stored in the processing queue, the disk control unit 200 reads the command stored in the processing queue (S102). Then, the disk control unit 200 determines whether there is a command that can be executed in parallel with the command read from the processing queue (S103).

  If there is a command that can be executed in parallel with the command read from the processing queue as a result of the determination, the disk control unit 200 executes the command read from the processing queue and the command that can be executed in parallel ( That is, an I / O command corresponding to the command to be executed is output to the SAS interface (SAS expander 600). Then, after the execution of both commands is completed, the command read from the processing queue is deleted from the queue (S104), and the command executed in parallel is deleted from the queue (S105). Thereafter, the process proceeds to step S107.

  On the other hand, when there is no command that can be executed in parallel with the command read from the processing queue, the disk control unit 200 executes the command read from the processing queue. Then, after the execution of the command is completed, the command read from the processing queue is deleted from the queue (S106). Thereafter, the process proceeds to step S107.

  In step S107, the disk control unit 200 determines whether the command is stored in all the processing queues. If no command is stored in any processing queue as a result of the determination, the process shifts to a standby state. On the other hand, if the command is stored in any of the processing queues, the process returns to step S102, and the command stored in the processing queue is processed.

  FIG. 15 is a flowchart of processing executed in the SAS expander 600 of the controller housing 1 according to the first embodiment of this invention.

  When the SAS expander 600 receives an I / O command from the disk controller 200, the SAS expander 600 refers to the state of the port to which data is to be transferred by the I / O command, and determines whether this port is transferring data. Determination is made (S111).

  As a result of the determination, if a port to which data is to be transferred by an I / O command is being transferred, the management table 1210 (FIG. 13B) is further referred to and whether or not this port is a multi-linked port. Is determined (S112).

  If it is determined that the port is a multi-linked port, it is further determined whether or not the I / O can be transferred by the multi-link (S113).

  As a result of the determination, if this I / O can be transferred by the multilink, the I / O is processed (data is transferred) by the multilink according to the received I / O command (S114).

  On the other hand, if the port is not a multi-linked port (NO in S112), or if this I / O cannot be transferred through the multi-link (NO in S113), the process returns to step S111 to try to transfer data. Wait until the port is free.

  Further, if the port to which data is to be transferred by the I / O command is not being transferred (NO in S111), it is further determined whether or not the port to which data is transferred is a multi-linked port ( S115).

  As a result of the determination, if this port is a multi-linked port, I / O is processed (data transfer) using the multi-link in accordance with the received I / O command (S114).

  On the other hand, if the port is not a multi-linked port, I / O is processed (data is transferred) by normal processing according to the received I / O command (S116).

  FIG. 16 is a flowchart of path change processing when a failure occurs according to the first embodiment of this invention. By the path changing process shown in FIG. 16, the path is changed as shown in FIG. 12B.

  After the failure of the port for which the virtual multilink is set is detected, the SAS expander 600 notifies the disk controller 200 of the failure notification of the port where the failure has occurred (S121).

  The disk control unit 200 refers to the management table 1200 (FIG. 13A) and determines whether or not all the wide links including the port where the failure is detected have become unusable (S122).

  As a result of the determination, when all the wide links including the port where the failure is detected become unusable, the path is changed to a standby port corresponding to this port (S124).

  On the other hand, when a part of the wide link including the port in which the failure is detected is usable, the link pair included in the multilink is changed, and the change of the path is notified to the SAS expander 600. (S123).

  After receiving the path change notification, the SAS expander 600 redirects the I / O of the link to the changed path (S125).

  In the case of the path change shown in FIG. 12D, a setting is made to reduce the number of links used by the virtual multilink including the path where the failure has occurred.

  As described above, in the first embodiment, by assigning one path or multilink path to the logically divided storage area, the bandwidth of the entire I / F is appropriately allocated according to the performance of the storage device. And the performance of the storage system can be improved.

<Embodiment 2>
Next, a second embodiment of the present invention will be described.

  In the conventional storage system, a verification code for ensuring that the data is correct is added for each predetermined unit of writing to the storage device. For example, verification codes such as LA and LRC are added every 512 bytes in a semiconductor storage device (so-called SSD) and a SATA disk drive, and every 520 bytes in an FC disk drive.

  Furthermore, even if the controller issues a write command to the storage device, there may be a problem that data is not written to the storage device.

  Therefore, in the prior art, after writing data, the written data is read and the write data and the read data are compared for each byte. Alternatively, all the verification codes of the written data and the verification codes of the read data are compared.

  However, the conventional method described above consumes a transfer bandwidth to read all the written data and consumes the processing capacity of the processor to compare the data, thus hindering improvement in performance. This deterioration in performance becomes a serious problem in a storage device capable of high-speed processing such as a semiconductor storage device.

In order to deal with this problem, it has been proposed to compare only the data of the first and last blocks for sequential data (see, for example, JP-A-2005-165443).
However, the above method has a problem that the overhead required for command generation / issuance and status confirmation for two times is large with respect to one writing, and there is a concern about a decrease in throughput.

  In order to solve this problem, in the second embodiment, among the verification codes added to the data of a plurality of blocks of the storage device, only the last block is added to the data of several blocks from the last block. Use a verification code to verify the data.

  The verification code of the second embodiment is, for example, 8 bytes, an address verification code (sequential number such as an address in a RAID group), a continuity verification code (an identifier generated every time data is written, written) Serial number of data, flags indicating the beginning and end of data, etc., data verification code (code such as CRC or ECC for data guarantee), and / or data attribute code (data type (read only or parity) ) Etc. can be used.

  FIG. 17A and FIG. 17B are diagrams for explaining the outline of the second exemplary embodiment of the present invention.

  As shown in FIG. 17A, the processor 400 (controller 1020) writes a final block to the storage device (the magnetic disk drive 21 or the semiconductor storage device 22) (1601), and then issues a read command for reading the written data. The written data is read (1602). Then, the verification code of the written data is compared with the verification code of the read data (1603). As a result, when the verification is successful, the processor 400 returns a write response to the host computer 10. On the other hand, if the verification fails (the verification code of the written data does not match the verification code of the read data), the processor 400 reports an error to the host computer 10. In the second embodiment, the processor 400 issues a read instruction for reading data of the last block (or several blocks from the end), and compares only the verification codes of the last block (or several blocks from the last).

  As shown in FIG. 17B, in the verification of the write data of the second embodiment, the port applicator 610 connected to the SAS expander 600 or the HDD replaces the processor 400 described above with the verification code of the write data. The verification code of the read data may be compared (1603).

  In this case, the SAS expander 600 (or the port applicator 610) writes the final block in the storage devices 21 and 22 (1601), and then holds the WRITE status of the storage devices 21 and 22 and holds the final block (or Data from the last several blocks) is read from the storage device (1602). Thereafter, the SAS expander 600 compares the verification code of the write data with the verification code of the read data (1603). As a result, if the verification is successful, the SAS expander 600 returns the pending WRITE response to the controller 1020 (1604). On the other hand, if the verification fails, an error is reported to the controller 1020 (1604).

  As described above, the SAS expander 600 or the port applicator 610 verifies the write data with respect to the write data, so that the load of the write data verification processing is distributed and the controller load can be reduced.

  Further, the storage devices 21 and 22 may verify the write data. As described above, if the write data verification is distributed in the storage devices 21 and 22, the load on the controller can be reduced.

  In this embodiment, since the amount of data processed in the data verification process executed by the processor 400 (SAS expander 600, port applicator 610) is small, the wait time of the WRITE response can be shortened, and the controller 1020 times out. Can be suppressed. In addition, the write data can be verified with high reliability while reducing the processing capacity and data transfer bandwidth of the CPU consumed for the verification process.

  In the verification of write data according to the second embodiment, reliability can be further improved by comparing all verification codes of written blocks when a predetermined condition is satisfied. This predetermined condition includes a case where the controller 1020 detects an error in the storage devices 21 and 22 within a predetermined time before writing data, and an error occurs when writing data or reading verification data. is there. In such a case, the operations of the storage devices 21 and 22 may be unstable, or the storage devices 21 and 22 may be out of order. Therefore, the reliability of data writing can be improved by verifying the entire written data.

  In addition, it is better to verify the entire written data periodically (for example, every predetermined time). By doing so, the failure of the storage devices 21 and 22 can be detected early, The reliability of data writing can be further improved.

  In addition, it is possible to lengthen the range for reading data for comparison of verification codes, and improve the possibility of detecting data mismatch due to verification codes. That is, the number of verification codes read out for comparison increases, so the probability that an abnormal verification code is included increases.

  It is also preferable to embed a management code for each writing in the verification code and compare the verification codes. The management code may use information that can verify the continuity of data, such as an identifier generated every time data is written, a serial number of the written data, and a flag indicating the beginning and end of the data.

  FIG. 18 is an explanatory diagram illustrating an error detection model according to the second embodiment of this invention.

  In FIG. 18, for example, a case where 10 blocks of data 1801 are written is considered. When the data is normally written in the storage devices 21 and 22, the tenth data is written in the tenth sector (1802).

  On the other hand, when the writing is interrupted due to an interrupt caused by internal processing having a high priority in the storage devices 21 and 22, the storage devices 21 and 22 may resume writing from the sector 1804 that has shifted after the end of the interrupt processing. is there. Although it is originally a bug, it may be overlooked in cases where reproduction and extraction are difficult, such as interrupt conflicts under specific conditions. As a result, an unexpected situation occurs. 22 may not be able to detect an error.

  However, even in this case, it is possible to verify whether the data has been normally written by reading the last one block of data (1801). Furthermore, by reading the data of one block before and after the last block, it is possible to confirm the continuity of data and the influence on the outside writing target area 1805.

  Whether to read only the data of the last block or the data of several blocks before and after the last block may be determined by the setting of the administrator or the user. In addition, when the data to be written is larger than a predetermined threshold, a plurality of blocks of data may be read for verification. Specifically, when data having a RAID stripe size or larger is written, a plurality of blocks of data may be read for verification. Further, when the ratio between the write data length and the read data length is larger than a predetermined threshold value, data of a plurality of blocks may be read for verification. In addition, when the bandwidth consumed for data writing is larger than a predetermined threshold, a plurality of blocks of data may be read for verification. Further, when the load on the storage device is larger than a predetermined threshold, data of a plurality of blocks may be read for verification.

  Thus, the reliability can be further improved by controlling the data length read for verification by the length of the data to be written.

  FIG. 19 is a flowchart of a data verification process executed by the processor 400 according to the second embodiment of this invention.

  First, when write data is received from the host computer 10, the channel control unit 100 stores the received data in the cache memory 300 (S131). Then, the processor 400 reads the data stored in the cache memory 300 and generates a verification code for the data received from the host computer 10 (S132). The processor 400 generates a write command and writes the data to the target storage device (S133).

  Thereafter, the processor 400 determines whether or not the load on the processor 400 is lower than a predetermined threshold, and whether or not the length of data to be written exceeds a predetermined threshold (S134).

  As a result of the determination, when the load on the processor 400 is low or the written data is long, the data of a predetermined number of blocks including the last written block is read (S137), and the verification code and the write data that are read Are compared with each other to determine whether or not they match (S138). As a result, if they do not match, it is determined that the data has not been written correctly, and error processing is executed (S139). On the other hand, if the two match, it is determined that the data has been correctly written, and the data verification process ends.

  As a result of the determination in step S134, when the load on the processor 400 is high or the written data is short, the data of the last written block is read (S135), and the read verification code and the verification code of the write data are Are compared to determine whether or not they match (S136). As a result, if they do not match, it is determined that the data has not been written correctly, and error processing is executed (S139). On the other hand, if the two match, it is determined that the data has been correctly written, and the data verification process ends.

  As described above, in the second embodiment, since the bandwidth consumed for verifying the write data is small, the performance is excellent. For this reason, it can respond to a user's high performance request.

  The second embodiment can be applied regardless of whether the write data is sequential data, unlike the technique described in Japanese Patent Laid-Open No. 2005-165443 described above.

  Also, in the second embodiment, unlike the technique described in Japanese Patent Laid-Open No. 2005-165443 described above, the data of the first block is not used. For this reason, continuous block data can be read with one read command, and there is no need to issue two read commands to read the first block and the last block as in the prior art.

  Also, in the second embodiment, unlike the technique described in Japanese Patent Laid-Open No. 2005-165443 described above, only the verification codes are compared without comparing all the data of the blocks.

  Furthermore, in the second embodiment, data after the last block is also read and verified. Therefore, it is possible to improve data reliability in a range not included in the conventional technology.

  Furthermore, in the second embodiment, by using information with regularity in the verification code, the cause of the error is specified by using the regularity of the verification code in addition to the OK or NG determination such as data comparison. Can do. Therefore, the reliability of the data writing process can be improved.

  Furthermore, in the second embodiment, since there are few data to be compared and the comparison can be performed by simple arithmetic processing, it is easy to incorporate into the software.

21 Magnetic disk drive 22 Semiconductor memory device 100 Channel controller 200 Disk controller 300 Cache memory 400 Processor (CPU)
500 Data controller 600 SAS expander 610 Port applicator (port expander)

Claims (11)

A storage system comprising a storage device that stores data, a controller that controls input and output of data to and from the storage device, and an interface that connects the storage device and the controller,
The storage device includes a plurality of physical ports connected to the interface,
The controller is
Providing the storage area of the storage device logically divided into a plurality of storage areas;
The storage system according to claim 1, wherein one or a plurality of the physical ports are allocated to the logically divided storage area. The controller is
Two physical ports are exclusively assigned to one logically divided storage area,
2. The storage system according to claim 1, wherein data is transferred so as to form one link by using two ports allocated to the logically divided storage area. The controller is
The two physical ports are exclusively assigned to the logically divided one storage area as working ports,
Two other physical ports are assigned to the logically divided storage areas as standby ports,
When a failure occurs in one active port assigned to the logically divided storage area, one standby port assigned to the storage area and an active port in which the failure has not occurred The storage system according to claim 2, wherein the data is transferred so as to form one link. The controller is
Assigning the plurality of physical ports to be shared by the logically divided storage areas;
The data is transferred so as to form one link by using a port that can be used at the time of data transfer among a plurality of ports allocated to the logically divided storage area. Storage system.   The storage system according to claim 1, wherein the controller allocates the plurality of ports equally or unevenly to the logically divided storage areas. A part of the storage area of the storage device is set as a replacement area used when the storage area becomes unrewritable,
The storage system according to claim 1, further comprising an interface for mounting an auxiliary storage device used when it is determined that the replacement area is insufficient. A storage device that stores data; a cache memory that temporarily stores data to be written to the storage device; a controller that controls input / output of data to / from the storage device; the storage device; the cache memory; and the controller A storage system comprising an interface for connecting
The controller is
When transferring data from the cache memory to the storage device, for each data writing block, generate a verification code to ensure the validity of the written data,
Transferring the data to which the generated verification code is added to the storage device;
Of the verification codes added to the data written in the plurality of blocks of the storage device, the verification code added to the data of one or a plurality of blocks from the first block is used without being used in the storage device. A storage system characterized by verifying data.   The controller uses only the verification code added to the data of the last block among the verification codes added to the data written to the plurality of blocks of the storage device, and the data written to the storage device The storage system according to claim 7, wherein the storage system is verified.   When the data transferred to the storage device is larger than a predetermined threshold, the controller determines the length of the verification code added to the data written in the plurality of blocks of the storage device from the last to the head. 8. The storage system according to claim 7, wherein the data written in the storage device is verified by using a verification code added to a small predetermined number of blocks of data.   The controller uses the verification code added to the data of a predetermined number of blocks before and after the last block among the verification codes added to the data written in the plurality of blocks of the storage device, 8. The storage system according to claim 7, wherein data written in the device is verified. A storage system comprising a storage device that stores data, a controller that controls input and output of data to and from the storage device, and an interface that connects the storage device and the controller,
The storage device includes a plurality of physical ports connected to the interface,
The controller is
Providing the storage area of the storage device logically divided into a plurality of storage areas;
Two physical ports are exclusively assigned as active ports to one logically divided storage area,
Two other physical ports are assigned to the logically divided storage areas as standby ports,
During normal operation, data is transferred so as to form one link using the two active ports assigned to the logically divided storage areas,
When a failure occurs in one active port assigned to the logically divided storage area, one standby port assigned to the storage area and an active port in which the failure has not occurred A storage system characterized by transferring data so as to form a single link.