JPH04340619A - Disk array system - Google Patents

Abstract

PURPOSE:To obtain a disk array system capable of activating plural disk cells in a linkage. CONSTITUTION:A disk array 1 where access ports capable of mutually writing/ reading the stored contents of adjacent disk devices and disk devices are connected like a torus by these access ports 1b, data multiplexers 2 which select the data to be transferred, and an input/output controller 3 are provided. Since data is normally written and read even at the time of fault and the processing for fault recovery is performed on the disk device side, the reliability and the throughput of the system are improved.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a disk device for a computer system, and more particularly to a disk array system with improved operation continuation processing in the event of a failure.

0002

2. Description of the Related Art Generally, when a failure occurs in an external storage device of a computer system, the stored data is destroyed in most cases, and this has a large impact on the operation of the computer system. Therefore, conventionally, an operation method has been adopted in which two external storage devices are used to avoid the failure.

A conventional external storage device that uses this type of failure avoidance method will be described below with reference to the drawings. As shown in FIG. 29, during normal operation, the execution program accesses only the disk device 6a via the selector 5, and exchanges data with the main memory 4 via the input/output control device 3. Backup data is stored in the disk device 6b, and when a failure occurs in the disk device 6a, the selector 5 replaces the disk device 6a with the disk device 6b.
to continue operating the computer system based on the backup data. When recovering from a failure, data is transferred from the disk device 6b in which the failure has not occurred to the disk device 6a.

FIG. 30 shows another example of conventional obstacle avoidance.
During normal operation, data from the main memory 4 is simultaneously transferred to the disk devices 6a and 6b via the input/output control device 3 and selector 5. When transferring data from the disk device 6a to the main memory 4, the same data is read from the disk device 6a and the disk device 6b at the same time, the two data are compared by the data comparator 7, and if the two data do not match, The disk device 6a or 6b in which a failure has occurred is identified and disconnected from the computer system to continue operation. When recovering from a failure, data is transferred from a disk device in which no failure has occurred to a disk device in which a failure has occurred.

[0005]

However, in the above-described conventional disk device, it is necessary to execute processing for recovery from a failure through the input/output control device. In particular, when restoring backup data, in addition to normal data transfer, data transfer between disk devices increases the load on the input/output control unit and reduces the throughput of the entire computer system. Ta.

SUMMARY OF THE INVENTION The present invention is intended to solve these conventional problems, and it is an object of the present invention to provide a disk array system that can recover from a failure without increasing the load on the input/output control device.

[0007]

[Means for Solving the Problems] In order to achieve the above object, in the present invention, in a disk array system consisting of a plurality of disk devices, the storage contents of each adjacent disk device can be written and read. have mutually possible access ports, and the disk devices are connected in a torus shape by the access ports.

[0008]

[Operation] Since the present invention is configured as described above, in preparation for the occurrence of a disk device failure, the storage contents of the disk device are backed up to a plurality of disk devices via the access port, and when the failure is recovered, the The backed up storage contents can be read from multiple disk devices and re-stored in the disk device from which the failure has been recovered. Further, when a failure occurs in a disk device, it is possible to disconnect the failed disk device and continue operation based on the storage contents backed up in a plurality of other disk devices.

[0009]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Examples of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram of an embodiment of a disk device of the present invention. In the figure, reference numeral 1 denotes a disk array composed of a plurality of disk cells 1a arranged in a matrix, and the disk cells 1a are connected to adjacent disks in the horizontal and vertical directions shown in the figure through access ports 1b. It is connected to the cell 1a in a torus shape. In this embodiment, the disk array 1 is composed of four disk cells 1a in both the horizontal and vertical directions, but there is no particular restriction on the number. A data multiplexer 2 selects data to be transferred, and is connected to all of the left and right disk cells 1a via their respective access ports 1b. Reference numeral 3 denotes an input/output control device, which is composed of a buffer memory 3a and a controller 3b. Data in the main memory 4 is transferred via the buffer memory 3a of the input/output control device 3 to one disk cell 1a designated by the controller 3b. Data in the disk cell 1a designated by the controller 3b of the input/output control device 3 is transferred to the main memory 4 via the buffer memory 3a of the input/output control device 3. Further, the disk cell 1a itself has a function of mutually transferring data between adjacent disk cells 1a, and backup data is exchanged between the two disk cells 1a.

FIG. 2 is a block diagram showing the detailed configuration of the disk cell 1a. In the same figure, 1b is the left and right direction,
These are four access ports that transfer data between vertically adjacent disk cells 1a and one access port that transfers data between data multiplexer 2. 1c is a disk controller that controls data transfer between adjacent disk cells 1a in the horizontal and vertical directions and between the data multiplexer 2; 1d;
is a disk device, and 1e is a storage medium. Now, when a read request is made to the access port 1b from one of the four adjacent disk cells 1a, the disk controller 1c issues a read request to the disk device 1d to read data from the storage medium 1e. Therefore, the disk controller 1c uses the access port 1b that received the read request.
The data is output to the adjacent disk cell 1a and transferred to the adjacent disk cell 1a. Furthermore, when a read request is issued from the data multiplexer 2 to the access port 1b, the disk controller 1c similarly issues a read request to the disk device 1d to read the data from the storage medium 1e, and sends the read request to the access port 1b connected to the data multiplexer 2. Output data. Further, when a write request is made to the access port 1b from one of the four adjacent disk cells 1a or from the data multiplexer 2, the disk controller 1c issues a write request to the disk device 1d to write data to the storage medium 1e. .

As shown in FIG. 3, the data storage area on the storage medium 1e is divided into eight areas: N, S, W, E, Nb, Sb, Wb, and Eb. Of these, N, S
, W and E are used for normal data recording, and the remaining four areas Nb, Sb, Wb and Eb are used to store backup data of adjacent disk cells 1a.

Next, referring to FIG. 4, the classification of areas used on the storage medium 1e when storing backup data will be explained. Hereinafter, the disk cell 1a that transfers data with the data multiplexer 2 will be referred to as a primary disk cell, and the disk cell 1a to which backup data is transferred from the primary disk cell will be referred to as a secondary disk cell. In the example in the figure, disk cell 1f is the primary disk cell, disk cells 1g and 1h
, 1i and 1j are secondary disk cells. Also, the position of the secondary disk cell 1h viewed from the primary disk cell 1f is called the N position, the position of the secondary disk cell 1g is called the S position, the position of the secondary disk cell 1i is called the E position, and the position of the secondary disk cell 1j is called the N position. This is called the W position. Since the disk cells 1a are connected in a torus shape, any disk cell 1
a can be designated as the primary disk cell 1f, and along with this, the adjacent secondary disk cell 1
The positions of g, 1h, 1i and 1j are also automatically determined. The backup data of the primary disk cell 1f is automatically stored in the four secondary disk cells as follows. That is, the data in the S area of the primary disk cell 1f is stored in the secondary disk cell 1g at the S position.
The data in the Nb area of the primary disk cell 1f is stored in the Nb area of the primary disk cell 1f in the secondary disk cell 1h at the N position.
The data in the E area of the primary disk cell 1f is stored in the Sb area of the primary disk cell 1f.
The data in the W area of the primary disk cell 1f is stored in the Wb area of the primary disk cell 1f in the secondary disk cell 1j at the W position.
It is stored in the Eb area of .

As shown in FIG. 5, data transfer between the data multiplexer 2 and the disk cells 1a on the matrix is performed using the access port 1b of each disk cell.

Next, the operation when writing data to the primary disk cell 1f will be explained. The relationship between each disk cell when writing data to a disk cell is as shown in FIG.
The data in a is transferred to the primary disk cell 1f. Backup data of the primary disk cell 1f is stored in the secondary disk cells 1g, 1h, 1i, and 1j. As shown in FIGS. 7 and 8, the input/output control device 3 transfers the memory area 4b in the memory space 4a of the main memory 4 to the memory area 3c in the buffer memory 3a. The controller 3b controls data transfer and at the same time controls which area of the buffer memory 3a the data is transferred to. Next, data is transferred from the memory area 3c to the primary disk cell 1f through the data multiplexer 2. The primary disk cell 1f stores the transferred data in any one of areas N, S, W, and E shown in FIG. 3, and transfers the backup data to the secondary disk cells 1g, 1h, 1i, and 1j. The disk controller 1c determines data transfer and in which area the transferred data is stored.

Next, the operation when reading data from the disk array 1 will be described with reference to FIGS. 9 and 10. When reading data from the primary disk cell 1f, data is also read from the secondary disk cells 1g, 1h, 1i, and 1j that store backup data, and is input to the data multiplexer 2 and stored in the memory area of the input/output control device 3. 3c, and subsequently transferred to the memory area 4b of the main memory 4. At this time, the controller 3b controls which disk cell to transfer data from.

Next, the operation during backup processing will be explained with reference to FIGS. 11 to 14. The position of the disk cell that stores backup data varies depending on the storage area of the data transferred to the primary disk cell 1f. As shown in FIG. 11, the data transferred to the N area of the primary disk cell 1f is stored in the secondary disk cell 1h. Backup data is stored in the Sb area of . Similarly, as shown in FIG. 12, the data transferred to the S area of the primary disk cell 1f is stored as backup data in the Nb area of the secondary disk cell 1g, and as shown in FIG.
The data transferred to the W area of the secondary disk 1j
As shown in FIG. 14, the data transferred to the E area of the primary disk cell 1f is stored in the Wb area of the secondary disk cell 1i.

Next, a data write operation when the primary disk cell 1f cannot be used due to a failure will be explained with reference to FIG. Since the primary disk cell 1f cannot be used by the input/output control device 3, the process is changed so that the data in the memory area 4b of the main memory 4 is transferred to the secondary disk cells 1g, 1h, 1i, and 1j. The transfer destination areas at this time are the Nb area for the secondary disk cell 1g, the Sb area for the secondary disk cell 1h, the Wb area for the secondary disk cell 1i, and the Eb area for the secondary disk cell 1j.

Next, the operation when reading data from the failed primary disk cell 1f will be described with reference to FIGS. 16 and 17. Since the primary disk cell 1f cannot be used, the input/output control device 3
changes the process so that data is transferred from the adjacent secondary disk cells 1g, 1h, 1i, and 1j to the memory area 4b of the main memory 4. At this time, the transfer source area is the Nb area for the secondary disk cell 1g, the Sb area for the secondary disk cell 1h, the Wb area for the secondary disk cell 1i, and the Eb area for the secondary disk cell 1j.

Next, the operation when the primary disk cell 1f, which has become unusable due to a failure, is recovered from the failure will be explained with reference to FIGS. 18 to 25. Primary disk cell 1 before failure
In order to recover data of f, data is sequentially transferred from secondary disk cells 1g, 1h, 1i and 1j. However, the data transfer order may be arbitrary regardless of the following explanation procedure. The N area of the primary disk cell 1f is recovered from the Sb area of the secondary disk cell 1h (Fig. 1
8), and the S area of the primary disk cell 1f is recovered from the Nb area of the secondary disk cell 1g (see Figure 19).
), the W area of the primary disk cell 1f is recovered from the Eb area of the secondary disk cell 1j (see FIG. 20), and the E area of the primary disk cell 1f is recovered from the Wb area of the secondary disk cell 1i (see FIG. 21). , the Nb area of the primary disk cell 1f is recovered from the S area of the secondary disk cell 1h (see FIG. 22).
, the Sb area of the primary disk cell 1f is recovered from the N area of the secondary disk cell 1g (see FIG. 23),
The Wb area of the primary disk cell 1f is recovered from the E area of the secondary disk cell 1j (see FIG. 24), and the Eb area of the primary disk cell 1f is recovered from the W area of the secondary disk cell 1i (see FIG. 25). Furthermore, since data transfer can be performed between disk cells without going through the input/output control device 3, the load on the input/output control device 3 can be reduced during failure recovery. FIG. 26 is a flowchart showing the procedure at the time of failure recovery described above, and since steps 125 to 132 correspond to FIGS. 18 to 25, respectively, their explanations will be omitted.

Next, the procedure for writing to the disk array 1 will be explained with reference to the flowchart of FIG. When a data write request occurs, the data is transferred from the main memory 4 to the buffer memory 3a of the input/output control device 3 (step 100), and the input/output control device 3 selects which disk cell 1a is to be the primary disk cell 1f. (Step 101). At this time, if the primary disk cell 1f to which the data is written has a fault, the primary disk cell 1f is disconnected from the system (steps 102, 111), and if no fault has occurred, the input/output control device 3 is disconnected. Data is transferred from the buffer memory 3a to the primary disk cell 1f (step 103). If a failure has occurred in the primary disk cell 1f after data transfer, it is disconnected from the system (steps 104, 111). Next, backup data is created, but at this time, the secondary disk cells 1g, 1h, 1i and 1 are used to store the backup data.
If a fault has occurred in any of j, disconnect the faulty disk cell from the system (steps 105, 110), and if no fault has occurred, transfer from primary disk cell 1f to secondary disk cell 1.
The data are stored in predetermined areas g, 1h, 1i, and 1j (step 106). If a failure occurs in any of the secondary disk cells 1g, 1h, 1i, and 1j after the backup data is created, the disk cell in which the failure has occurred is disconnected from the system (step 108,
110). If no failure has occurred in the primary disk cell 1f or any of the secondary disk cells 1g, 1h, 1i, and 1j in the processing so far, the buffer memory 3a of the input/output control device 3 is cleared (step 109). , it is assumed that writing has completed normally. If a failure occurs in the primary disk cell 1f during data transfer from the input/output control device 3 and the failed disk cell is disconnected from the system (step 111), any failure occurs in the secondary disk cells 1g, 1h, 1i, and 1j. (Step 11)
2) If a fault has occurred, disconnect the faulty disk cell from the system (step 115).
), the write process ends abnormally. If no failure has occurred, the buffer memory 3a of the input/output control device 3 is cleared (step 109), and the write process ends normally (step 109).

As described above, by operating a plurality of disk cells in conjunction with each other, the primary disk cell 1
Even if a failure occurs in any one of the primary disk cells f, the secondary disk cells 1g, 1h, 1i, and 1j adjacent to the failed primary disk cell 1f are used to restore the adjacent two disk cells in the disk array disk array 1. As long as all the disk cells 1a do not fail at the same time, the system as a whole will not fail, so reliability can be improved.

Next, the procedure for reading data from the disk array 1 will be explained with reference to the flowchart shown in FIG. When a data read request occurs, the controller 3a of the input/output control device 3 selects which disk cell 1a is to be the primary disk cell 1f (step 116). At this time, primary disk cell 1 from which data is read is
If no fault has occurred in the primary disk cell 1f, data is transferred from the primary disk cell 1f to the buffer memory 3a of the input/output control device 3 (step 118). Next, the data is transferred from the buffer memory 3a to the main memory 4 (step 123), the buffer memory 3a of the input/output control device 3 is cleared (step 124), and the data reading ends normally. If a failure has occurred in the primary disk cell 1f (step 117), this primary disk cell 1f is disconnected from the system (step 119).
), and if a failure has occurred in any of the secondary disk cells 1g, 1h, 1i, and 1j (step 120), the disk cell in which the failure has occurred is disconnected from the system (step 122), and a data read error is detected. It ends. Secondary disk cell 1g, 1h,
If no failure has occurred in 1i and 1j, the data is transferred to the buffer memory 3a of the input/output control device 3 (step 121), and then the data is transferred from the buffer memory 3a to the main memory 4 (step 123). The buffer memory 3a of the input/output control device 3 is cleared (step 124), and data reading ends normally.

As described above, by operating a plurality of disk cells in conjunction with each other, the primary disk cell 1
Even if a failure occurs in any one of the primary disk cells f, reading data from the disk array 1 using the secondary disk cells 1g, 1h, 1i, and 1j adjacent to the failed primary disk cell 1f. Can be done normally. Further, as long as two adjacent disk cells 1a in the disk array 1 do not fail at the same time, the entire system will not fail, so reliability can be improved.

[0024]

[Effects of the Invention] As explained above, the present invention allows data reading and writing to be performed normally even in the event of a failure by operating a plurality of disk cells in conjunction with each other, thereby improving system reliability. Therefore, failure recovery can be performed without increasing the load on the input/output control device and the host device, which has the effect of increasing the throughput of the entire system.

[Brief explanation of drawings]

FIG. 1 is a block diagram of an embodiment of the present invention;

[
Figure 2: Block configuration diagram of disk cell,

FIG. 3 is an explanatory diagram showing data storage area divisions of a disk device medium;

FIG. 4 is an explanatory diagram showing data storage area divisions of a disk device medium;

FIG. 5 is a block diagram showing the relationship between disk cells and data multiplexers;

FIG. 6 is an explanatory diagram showing data write processing operation,

[Figure 7
]Explanatory diagram showing data write processing operation,

FIG. 8 is an explanatory diagram showing data write processing operation;

FIG. 9 is an explanatory diagram showing data read processing operation;

FIG. 10 is an explanatory diagram showing data read processing operation;

FIG. 11 is an explanatory diagram showing data backup processing operation;

FIG. 12 is an explanatory diagram showing data backup processing operation;

FIG. 13 is an explanatory diagram showing data backup processing operation;

FIG. 14 is an explanatory diagram showing data backup processing operation;

FIG. 15 is an explanatory diagram showing write processing operation when a failure occurs;

FIG. 16 is an explanatory diagram showing read processing operation when a failure occurs;

FIG. 17 is an explanatory diagram showing read processing operation when a failure occurs;

FIG. 18 is an explanatory diagram showing processing operations during failure recovery;

[Figure 1
9] Explanatory diagram showing processing operations during failure recovery,

FIG. 20 is an explanatory diagram showing processing operations during failure recovery;

FIG. 21 is an explanatory diagram showing processing operations during failure recovery;

FIG. 22 is an explanatory diagram showing processing operations during failure recovery;

FIG. 23 is an explanatory diagram showing processing operations during failure recovery;

FIG. 24 is an explanatory diagram showing processing operations during failure recovery;

FIG. 25 is an explanatory diagram showing processing operations during failure recovery;

[Fig. 26] Flowchart showing the procedure at the time of failure recovery,

FIG. 27 is a flowchart showing the procedure when writing data;

FIG. 28 is a flowchart showing the procedure when reading data;

FIG. 29 is a block diagram of a conventional disk device;

FIG. 30 is a block diagram of a conventional disk device.

[Explanation of symbols]

1 Disk array, 1a Disc cell, 1b Access port, 1c Disk controller 1d Disk device, 1e Storage medium, 2 Data multiplexer, 3 Input/output control device, 3a Buffer memory, 3b Controller, 4 Main memory, 5 Selector, 6 Disk device, 7. Data comparator.

Claims (3)

[Claims] Claim 1: In a disk array system consisting of a plurality of disk devices, each adjacent disk device has an access port that can mutually write and read the storage contents of each disk device, and this access port allows the disk device to A disk array system characterized by being connected in a torus shape. 2. In preparation for the occurrence of a disk device failure, the storage contents of the disk device are backed up to a plurality of disk devices via the access port, and the storage contents backed up from the plurality of disk devices when recovery from the failure occurs. 2. The disk array system according to claim 1, further comprising a disk controller that reads out the data and stores it again in the disk device from which the failure has been recovered. 3. Claim 1, further comprising a disk controller which, when a failure occurs in a disk device, disconnects the failed disk device and continues operation based on the storage contents backed up in a plurality of other disk devices. Disk array system described.