JP4444636B2 - Disk subsystem - Google Patents

Description

The present invention relates to an electronic apparatus such as a disk subsystem, a disk array, and a computer having a built-in disk drive, and more particularly to a technique that enables high-speed transfer by connecting array disks to a fabric switch.

In general, when a disk controller and a plurality of disk drives are connected in a disk array, as described in JP-A-10-171746, an SCS is used.
I interfaces or Fiber Channel arbitrated loop topologies are used.

The SCSI interface employs a method of transferring data on the same line in a time-sharing manner, and access to the initiator is a method of performing one-to-one communication per time on one transmission line.

In the Fiber Channel Arbitrated Loop topology, the initiator is looped through the serial interface to the SCSI interface.
Disk drives can be connected, data divided into frames can be transferred in a time-sharing manner, and multiple devices can communicate simultaneously, and the number of connectable disk drives can be expanded to 126.

JP-A-10-171746

In the future, it will be possible to use more disk drives by reducing the size and increasing the density of disk drives.

Since the SCSI interface is a system that performs one-to-one communication per time on one transmission path, communication between a large number of initiators and disk drives cannot be performed simultaneously.
Also, the number of connectable disk drives is as small as 7 to 15. Therefore SC
When one-to-one connection is made with one interface when the number of drives is increased with an interface using SI, a large number of interfaces are required, resulting in difficulty in mounting. Further, since the number of disk drives that can be connected by one control circuit is small, it is necessary to use a large number of control circuits.

On the other hand, when using Fiber Channel, the disk drive protocol cannot be switched because the protocol is different from that of the controller, and loop connection using a Fiber Channel arbitrated loop in which many disk drives share the same loop. I had to. Therefore, when the number of disk drives connected to the same loop is increased, the data transfer speed of the disk drive becomes larger than the maximum data transfer speed of the loop, and as a result, transfer cannot be performed with an efficiency exceeding the maximum data transfer speed of the loop. Connection was possible only at the same data transfer speed as the SCSI interface.

In order to solve the above problems, in the present invention, a protocol control unit is provided between the fiber channel fabric switch and the disk drive in order to enable switch connection between the disk drive and the control device.

The present invention reduces the number of connection lines using a Fiber Channel interface, which is a serial interface, and uses a Fiber Channel fabric topology that enables switch connection, thereby sacrificing the transmission performance of a large number of disk drives in the disk drive interface circuit. It becomes possible to connect without making. Further, by dynamically switching the connection for each control device and each disk drive group, a large number of disk drives can be controlled by a small number of disk drive control circuits. Furthermore, the data migration in the event of a disk drive failure can be performed independently of the disk drive interface control circuit and the disk drive data transfer, thereby improving the system reliability.

Hereinafter, an external storage device (disk subsystem) to which the present invention is applied will be described with reference to the drawings.
Examples will be described. FIG. 1 is an overall view.

In the external storage device shown in the figure, N disk array control circuits (control units) (
1-1) to (1-N) (1-2 in the middle is omitted, the same applies hereinafter), the upper side is connected to a host computer (not shown), and the lower side is M disk drive interfaces (disks). Drive I / F) control circuits (2-1) to (2-M) are provided. Details of the hardware configuration of the disk array control circuit will be described later. M fiber channel fabric switch control circuits (3-1) to (3-M) are disk drive interface (I / F) control circuits (2-1) for controlling the disk drive by the fiber channel interface 5. To (2-M). Then, a total of M × L disk drives (4 (1, 1) to 4 (M,
L)) by means of the Fiber Channel interface 6
The fabric switch control circuits (3-1) to (3-M) are connected.

Also, each disk drive interface control circuit (2-1) to (2-M),
The disk drives 4 (1, 1) to 4 (M, L) for storing data have individual identifiers (ID numbers). The Fiber Channel fabric switch control circuits (3-1) to (3-M) receive the ID numbers of the disk drives to be connected from the disk drive interface control circuits (2-1) to (2-M) and correspond to them. Disk drive interface control circuit (2-1) to (2-M)
And a one-to-one connection between the disk drives 4 (1, 1) to 4 (M, L).

FIG. 2 shows a hardware configuration of the disk array control circuits (1-1) to (1-N). Data transferred from a host computer (not shown) is controlled by the host interface control unit 7 and temporarily stored in the cache memory 8 and is also added with parity data by the parity data generation unit 9, and a data block and a parity data block (M in total). These data and parity blocks are stored in a disk drive group (not shown) by disk drive interface control circuits (2-1) to (2-M), which are corresponding interfaces.

When transferring data to the host computer, if the data to be transferred exists in the cache memory 8, the data is transferred to the host interface control unit 7.
Is transferred to the host computer. If there is no data to be transferred in the cache memory 8, the disk drive interface control circuits (2-1) to (2-1)
2-M) reads the decomposed data from the disk drive group, combines the data decomposed by the parity data generation unit 9, temporarily stores it in the cache memory 8, and transfers the host interface control unit 7 to the host computer .

Note that the above example is a data storage method when RAID is used, and it is naturally possible to store data without using the RAID system. In this case, the data transferred from the host computer (not shown) without the parity data generation unit 9 is temporarily stored in the cache memory 8 by the host interface control unit 7 and any data in the disk drive group. In the case of the mirror method, a plurality of identical data is stored in a plurality of disk drives. When reading data, data is read from the disk drive, temporarily stored in the cache memory 8, and transferred to the host computer by the host interface control unit 7.

The following example also describes a disk subsystem using RAID, but R
Of course, it is not limited to the case of using AID.

FIG. 3 shows fiber channel fabric switch control circuits (3-1) to (3
-M) shows the hardware configuration. Disk drive interface control circuit (
The protocol control unit 16 (first protocol control unit) connected to 2-1) detects the ID number of the disk drives 4 (1, 1) to 4 (1, L) to be accessed and the Fiber Channel protocol. Control. Disk drive 4 (1,1
) To 4 (1, L), the protocol control unit 16 ′ (second protocol control unit) assigns an ID number to the disk drives 4 (1, 1) to 4 (1, L), and switches the control unit. 17 of the disk drive 4 (1, 1) to 4 (1, L) in charge of 17
Report D number. The switch control unit 17 includes each disk drive 4 (1, 1) to
4 (1, L) ID numbers are stored, and the switch 18 is set by the ID numbers received from the disk drive interface control circuits (2-1) to (2-M) to establish a one-to-one connection. To do.

Protocol control may be set to be performed on the protocol control unit 16 'side, data transfer from the host computer, data transfer to the host computer, normal data transfer, and data transfer at the time of disk failure And the protocol control unit 16 and the protocol control unit 16 ′ may be set to be switched.

Further, only one of the protocol control unit 16 and the protocol control unit 16 ′ may be provided, and an ID number detecting unit may be provided instead of the protocol control unit 16, or an ID number assigning unit may be provided instead of the protocol control unit 16 ′. Good.

In addition, the protocol control unit and the switch are provided in the disk drive interface control circuits (2-1) to (2-M) instead of the fiber channel fabric switch as an independent device, and the disk drive 4 (1 , 1) to 4
You may make it connect with (1, L).

FIG. 4 shows fiber channel fabric switch control circuits (3-1) to (3-M
) Operation.

The disk array control circuit (1-1) stores the data divided into M pieces in the disk drive group (10-1). At this time, the disk array control circuit (1
-1) M disk drive interface control circuits (2-1) to (2-M)
) Are the Fiber Channel fabric switch control circuits (3-1) to (3-
The ID number of the disk drive belonging to the disk drive group (10-1) is transmitted to M), and the switch is established. The protocol control unit 16 (see FIG. 3) in the fiber channel fabric switch control circuits (3-1) to (3-M) detects the ID number and requests the switch control unit 17 to switch the switch connection. . Protocol control is performed according to the disk drive. The switch control unit 17 (see FIG. 3) connects the switch 18 (see FIG. 3) to the disk array control circuit (1-1) that is the connection request source and the disk drive 4 that belongs to the disk drive group (10-1) that is the connection request destination. Switch to connect.

At this time, the disk array control circuit (1-1) includes the disk drive group (10-1) and the fiber channel fabric switch control circuit (3-1).
Through (3-M), the other disk array control circuit (1-N) and the other disk drive group (10-2) independently transfer other data. Can be done. That is, even if the disk array control circuit (1-N) establishes a connection to the disk drive group (10-L), the disk array control circuit (1-1), the disk drive group (10-1), and the disk array Since the connection between the control circuit (1-N) and the disk drive group (10-L) can operate independently of each other, the highest data transfer speed possible between each disk array control circuit and the disk drive. Can be used to transfer data.

Although not described in detail, the switch control unit 17 performs the above-described switch switching and switches the connection of the switch 18 in response to a signal indicating that the disk drive is already ready for reading / writing at the time of data reading / writing. Can effectively secure the maximum transfer time.

FIG. 5 shows an expanded embodiment of the present invention. In the embodiment shown above, the protocol controller 16 of the Fiber Channel fabric switch control circuit 3 and the disk drive 4 are connected in a one-to-one correspondence from the protocol controller 16 to the Fiber Channel A plurality of disk drives 4 are loop-connected via an arbitrated loop control circuit 11. By connecting in this way, by connecting a large number of inexpensive disk drives 4, performance equivalent to that provided with a large capacity disk drive can be achieved. Even in this case, not all of the disk drives are connected in a loop, and apparently, the fiber channel arbitrated loop control circuit 11 and a large number of disk drives 4 are one disk drive 4, so that the access performance is There is no decline.

Although not shown, when the maximum data transfer rate of the fiber channel interface is sufficient with respect to the access speed of the disk drive, a plurality of disk drives 4 are connected to the fiber channel arbitrated loop control circuit 1.
1 is connected, a plurality of disk drives are connected in the same loop, and the maximum transfer rate of the Fiber Channel is shared by the plurality of disk drives 4, thereby increasing the number of disk drives 4 without degrading the access performance. Is possible.

FIG. 6 shows a hardware configuration diagram of the arbitrated loop control circuit 11 used in the embodiment shown in FIG.

The arbitrated loop control circuit 11 includes a loop bypass circuit 13, a plurality of disk drive connection ports 12, and a fabric switch connection port 15. A loop bypass circuit switching signal 14 is output from the disk drive 4, and when the disk drive fails, the port is bypassed, and the disk drive can be removed without breaking the loop and without affecting other operating disk drives. It is possible to add.

  FIG. 7 shows another expanded embodiment of the present invention.

In this embodiment, each fiber channel fabric switch control circuit (3-1)
) To (3-M) spare disk control circuit 19 and a plurality of spare disk drives (4-a), (4-
b). In the fiber channel fabric switch control circuit 3, the protocol control unit 16 ′ (FIG. 3) connected to the disk drive group including the failed disk drive 4 (4 (1,2) disk drive group in the figure). Is connected to the protocol control unit 16 ′ connected to the spare disk control circuit 19 through the switch 18. When any disk drive fails, the disk array control circuits (1-1) to (1-N) reconstruct data in the spare disk drive (4-a) or (4-b). .

If errors occur frequently in a specific disk drive 4 and there is a possibility of failure, the data of the disk drive 4 in which errors frequently occur is stored in the spare disk drive (4-a
) Or (4-b) to perform reconstruction. When the disk drive 4 is completely damaged and data transfer is impossible, the cache memory 8 and the parity data generation unit 9 shown in FIG. 2 are used by using the data of the disk drive group of the damaged disk drive 4. To restore the damaged data and spare disk drive (4-
Write to a) or (4-b).

Alternatively, the spare disk control circuit 19 may be performed independently. Therefore, the spare disk control circuit 19 includes a cache memory and a parity data reproducing unit. When the disk drive 4 is completely damaged, the spare disk control circuit 19 reads the remaining disk drive group data,
The damaged data is reproduced and written to the spare disk drive (4-a) or (4-b).

Therefore, access to the spare disk control circuit 19 from each disk drive storing the data that has been divided including the parity data to repair the failed disk drive 4 or the failed part, and the disk array control circuit (1-1) to (1
-N) through the disk drive 4 (4 (1,1) and 4 (1, L) in the figure)
The disk drive group) and the data access from the host computer can operate independently, so that the data can be reconstructed without affecting the data access of the host computer.

Similarly, when the failed disk drive is replaced with a normal disk drive, the spare disk control circuit 15 operates in the same manner as the Fiber Channel fabric.
Spare disk drive (4) for switch control circuits (3-1) to (3-M)
-A), (4-b) and one-to-one connection with the normal disk drive replaced from the failed disk drive, and the disk array control circuits (1-1) to (1-N)
) And the disk drive groups (10-1) to (10-L) (see FIG. 4), the data can be copied independently without disturbing the access from the host computer. Disk drive recovery can be performed.

1 is an overall view of an embodiment of the present invention. It is a detailed diagram of a disk array control circuit. It is a detailed view of a fiber channel fabric switch control circuit. It is a connection diagram of a fiber channel fabric switch. It is a connection diagram of a fiber channel fabric switch and an arbitrated loop. It is a detailed view of a fiber channel arbitrated loop control circuit. It is a connection diagram of a spare disk control circuit.

Explanation of symbols

1 ... disk array control circuit, 2 ... disk drive interface control circuit,
3 ... Fiber channel fabric switch control circuit, 4 ... Disk drive,
5 ... Fiber Channel interface, 6 ... Fiber Channel interface,
DESCRIPTION OF SYMBOLS 7 ... Host interface control part, 8 ... Cache memory, 9 ... Parity data generation part, 10 ... Disk drive group, 11 ... Fiber channel arbitrated loop control circuit, 12 ... Disk drive connection port, 13 ... Loop bypass circuit, 14 DESCRIPTION OF SYMBOLS ... Loop bypass signal switching signal, 15 ... Fabric switch connection port, 16 ... Protocol control part, 17 ... Switch control part, 18 ... Switch, 19 ... Spare disk control circuit.

Claims (6)

Multiple disk drives,
A disk array controller that generates parity data from data received from a host computer, and includes a disk drive interface that transfers data and parity data to the plurality of disk drives;
One switch control circuit connected to the plurality of disk drives and the disk drive interface unit;
A spare disk and a spare disk control circuit connected to the switch control circuit ,
The switch control circuit includes:
A switch, a plurality of protocol control units for detecting the ID numbers of the plurality of disk drives and Fiber Channel protocol control, and a switch for controlling the switches based on the ID numbers received from the disk drive interface unit A control unit,
The spare disk control circuit includes:
The first data stored in the first disk drive in which errors frequently occur among the plurality of disk drives is transferred to the spare disk drive via the switch control circuit,
The transition is
Data transfer from the disk drive interface unit to the second disk drive based on access from the host computer, performed via the switch control circuit, is performed independently.
A disk array device characterized by that. The disk array device according to claim 1,
The disk control apparatus according to claim 1, wherein the switch controller dynamically switches between a path connecting the disk drive interface unit and the switch control circuit and a path connecting the switch control circuit and the plurality of disk drives . The disk array device according to claim 1,
The disk array apparatus characterized in that a path connecting the switch control circuit and the plurality of disk drives is a path for communication through a fiber channel interface . The disk array device according to claim 1,
The switch control circuit receives data or parity data and an ID number of a transfer destination disk drive, and transfers the received data or the parity data to the transfer destination disk drive based on the ID number the disk array apparatus, characterized in that the. The disk array device according to claim 1,
If the first disk drive with frequent errors is replaced with a new disk drive,
The first data stored in the spare disk is transferred to the new disk drive via the spare disk control circuit and the switch control circuit,
The transition is
Data transfer from the disk drive interface unit to the second disk drive based on access from the host computer, performed via the switch control circuit, is performed independently.
A disk array device characterized by that. The disk array device according to claim 1 ,
When the switch control unit receives an access request from the host computer,
The switch control unit switches the connection of the switch after receiving a signal indicating that it is ready to respond to the access request from the disk drive related to the access request. .

Images