JP2001159958A - Parity data writing system for display array device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve the performance of read/write processing of a disk array device by reducing the application ratios of control devices and resources to be used for the generation and writing of new parity data and allocating the control devices and the resources to the read/write processing, based on requests from a host computer. SOLUTION: Out of host data write processing, parity data generation processing and parity data write processing to be performed by the control devices in a conventional system, the parity data generation processing and the parity data write processing are performed not by the control devices but by parity generation devices. Since data transfer to be performed at the parity generation is performed directly between the parity generation devices without passing the control devices, the resources of the control devices are not used at all.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to the control of a disk array device, and more particularly, to a method for receiving data from a host computer and generating redundant data on a fiber channel loop without a control device, thereby forming a disk array device. To a method of writing data to a disk device that is operating.

[0002]

2. Description of the Related Art When writing data received from a host computer (hereinafter, referred to as host data) to a disk device, redundant data belonging to the same redundant data group as the disk device (hereinafter, referred to as an ECC group) separately from the host data. A process of newly generating data (hereinafter referred to as parity data) and writing the data to the corresponding disk device is required. At this time, conventionally, the control device has a parity generation function, and in order to newly generate parity data belonging to the same ECC group as that of the disk, old data before writing to the disk device and writing of new parity data before writing new parity data to the disk device are used. The old parity data is read by the controller,
Using these and host data, new parity data must be generated by the parity generation function and written to the corresponding disk device. This significantly increases the load on the controller and the utilization of resources in the controller.

[0003]

As described above, when the control device writes host data, the old data read process and the old parity data read process are performed for one write process, and the new parity data is written. After the generation, it is necessary to perform four processes of writing new parity data to the disk. This increases the load on the control device. Also, when reading the old data / old parity data, resources in the control device are used, so that the read / write processing that is normally performed is pressed down.

The present invention reduces the usage rate of a control device and resources used for the generation of the new parity data and the writing of the new parity data, and allots the reading / writing based on a request from the host computer. Improve the performance of read / write processing of a disk array device.

[0005]

In order to solve the above-mentioned problems, according to the method of the present invention, parity data generation processing among host data write processing, parity data generation processing, and parity data write processing performed by the controller. And the write processing is performed not by the control device but by the parity generation device. Further, since data transfer performed at the time of parity generation is directly performed between the parity generation devices without passing through the control device, resources of the control device are not used at all. Therefore, the control device need only perform the process of writing host data to the disk device and the normal read process. By these things,
The controller is released from the parity data generation processing and the parity data writing processing for the host data.
In addition, the usage rate of the control device resources is minimized.

[0006]

FIG. 1 shows an example of the configuration of a disk array device. 1-1 is a disk array device, and 1-2 is a host computer to which the disk array device is connected. Host computer 1 of disk array device 1-1
-2 is performed via a control device 1-3 on the host computer side provided in the disk array device. The control device 1-3 is connected via a cache memory 1-4 to a disk device-side control device 1-6 that controls a plurality of disk devices 1-5 constituting the disk array device. In addition, there is a parity generation device 1-8 between the control device 1-6 and the plurality of disk devices 1-5, and the control device 1-6, the parity generation device 1-8, and the plurality of disk devices 1-5 respectively They are connected via fiber channel loops. FIG. 2 shows the details of the portion 1-7, and shows an example of the connection configuration of the disk devices constituting the disk array.

In FIG. 2, a plurality of disk devices such as 2-1 are connected to a loop controller 2-4 in a controller 2-3 by one loop 2-2. The data transfer speed of the loop 2-2 is higher than the data transfer speed of one disk device, and enables simultaneous communication between a plurality of disk device-loop control unit pairs by time division multiplexing. The control device 2-3 includes a plurality of loop control units 2-4, and a plurality of loops are connected.
A parity generation device group 2-6 is connected between the loop control unit 2-4 and the disk device 2-1 via a fiber channel loop 2-2.
The same number of parity generation devices 2-7 as the number of loops are provided in −6, and the parity generation devices 2-7 can communicate with each other.

The parity generator 2-7 includes a cache 2-10 for holding data received via a fiber channel loop, an arithmetic unit 2-11 for generating parity data from data on the cache, and a parity generator. There is an accessible memory 2-12. Memory 2
-12 has a write information management table as shown in Table 1. The table has an entry for each write request, and each entry includes fields for a disk device identifier for which a write request has been made, a command at the time of the write request, a start address, and the number of transfer blocks.

[0009]

[Table 1]

When communication is performed between the parity generation devices, a parity generation information management table as shown in Table 2 is held in the shared memory 2-9, and the parity generation information mainly corresponds to the disk device that has requested writing. Used to search for a disk device.

[0011]

[Table 2]

The table has an entry for each parity generation request. The data disk device identifier, disk device type (data / parity), the parity disk device identifier corresponding to the data disk device, and the parity disk There is a parity generation device identifier connected to the loop to which the device belongs. This table can be accessed from all the parity generation devices 2-7 in the parity generation device group 2-6.

Next, referring to FIGS. 3 and 4, a data disk (for example, 2-D data) for the data 3-1 on the cache 2-5 to which a write request has been issued from the host computer.
The flow of the process of writing in 1) will be described.

The host data writing process is performed by the control unit 2
-3 through the loop control unit 2-4 and starts when the parity generation device 2-7 in the same loop 2-2 as the data disk to be written (for example, 2-1) receives the host data 3-1. 4-1. When the parity generation device 2-7 receives the data, it refers to the command of the received data to determine whether it is a write request 4-2. If the received data is a reading process, the parity generation device 2-
7 transmits the received data to the next disk device and ends the processing 4-10. If the received data is a write process, the information shown in Table 1 is obtained from the received data,
4-3 to be registered in -12. Next, the received data is stored in the cache 2-10 4-4. Then, a request for reading old data is made to the data disk device 2-1 based on the contents of Table 1 registered in the memory 2-12 4-5, and the received old data is stored in the cache 2-10 4- 6.
Next, the host data in the cache 2-10 stored in 4-4 is written to the data disk device 2-1 based on the contents of Table 1 registered in the memory 2-12 4-7. Here, the host data write processing is completed, but a parity generation instruction is issued next.

The parity generation instruction is performed by searching the parity generation information management table of Table 2 in the shared memory 2-9 using the data disk identifier as a key, and searching for a corresponding parity generation device (for example, 3-2). 4-9 is performed by transmitting a parity generation request, host data, and old data. The parity generation request is sent to the parity disk device 3-3 connected to the parity generation device 3-2 obtained above.
3 is generated based on the head address of the data in Table 1 and the number of transfer blocks in Table 1, and transmitted to the parity generation device 3-2 that performs parity generation. At that time, the cache 2-10
Also send the host data and old data stored in
9. End the process 4-10.

Next, the parity generation processing and parity data writing processing will be described with reference to FIGS. 6 in FIG.
-1, the parity generation device 2 (3-2) receiving the parity generation request receives the parity generation command via the communication cable 5-1 and stores the information shown in Table 1 in the memory 5-5. 6-2. Next, host data and old data are received via the communication cable 5-1 and stored in the cache 5-2 6-3. Next, Table 1 in the memory 5-5
6-4 instructs the parity disk device 3-3 to read the old parity data based on the contents shown in (4). 6-5 storing the old parity data received from the parity disk device 3-3 in the cache 5-2. As a result, all the data for generating the new parity data is prepared, and these data are input to the arithmetic unit 5-4 to generate the new parity data 6-6. Thereafter, the new parity data is written to the parity disk device 3-3 along the path 5-3 6-7. Thus, the parity generation processing is completed, and a series of processing associated with the 6-8 host data write request is completed.

[0017]

According to the conventional write processing,
In addition to the process of writing to the target data disk device, the control device performs all of the parity generation process and the process of writing new parity data to the parity disk device.
In the present invention, the new parity generation process and the write process to the parity disk device are performed by the parity generation device, and the resources such as the microprocessor and the memory on the control device are allocated only to the normal write process. The performance of the array device can be improved.

[Brief description of the drawings]

FIG. 1 is a diagram showing a configuration example of a disk array device.

FIG. 2 is a diagram showing an example of an internal configuration of a parity generation device and a connection configuration example of a disk device.

FIG. 3 is a view showing a process of writing host data to a data disk device.

FIG. 4 is a diagram showing a flow of host data write processing to a data disk.

FIG. 5 is a diagram showing a state of parity generation and parity data write processing.

FIG. 6 is a diagram showing a flow of parity generation and write processing to a parity disk.

[Explanation of symbols]

1-1: Disk array device, 1-2: Host computer, 1-3: Control device, 1-4: Cache memory,
1-5: Plurality of disk devices, 1-6: Control device on the disk device side, 1-8: Parity generation device, 2-1: Disk device, 2-2: Loop, 2-3: Control device, 2- 4
... Loop control unit, 2-6 ... Parity generation device group, 2-7
... Parity generation device, 2-10 ... Cache, 2-11
... Arithmetic unit, 2-12... Accessible memory.

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Takao Sato 2880 Kozu, Odawara-shi, Kanagawa Prefecture Within the Hitachi, Ltd.Storage Systems Division (72) Inventor Masatsugu Nozaki 6-81 Ouecho, Naka-ku, Yokohama-shi, Kanagawa Hitachi Software Engineering Co., Ltd. In-house F-term (reference) 5B065 BA01 CA15 CE11 EA01 EA12 ZA08 5D066 BA02 BA08

Claims (1)

[Claims] There are a plurality of disk devices having redundancy for improving reliability against a failure and a control device for controlling these disk devices, and a plurality of disk devices are simultaneously controlled by time division multiplexing. A control device and the plurality of disk devices storing data are connected by a fiber channel loop enabling communication between the device pairs, and m (m> 1) data storage systems are connected. An RA which has a redundant data group in which redundant data exists in a data unit and divides and stores data in a plurality of storage media
In the disk array device using the ID method, data on a cache received from a host computer is written to the disk device of the disk device group, and the redundant data in the redundant data group to which the disk device belongs is written to the control device. A method in which the data is generated on the loop of the Fiber Channel and written to the disk device in order to reduce processing.