JPH0713700A - Disk array device - Google Patents

Abstract

PURPOSE:To remarkably shorten processing time by improving access efficiency to plural disks with the number of trams smaller than that of a conventional system. CONSTITUTION:A manager tram MT is connected to only data trams DT0-DT3 due to the limitation of the number of serial ports, and data from a host is transferred to the data trams DT0-DT3. The data trams DT0-DT3 write transferred data in disk devices D0-D3. The data tram completing processing earliest among the data trams DT0-DT3 transfers the data to a data tram DT4. The data tram DT4 receiving the data writes the data in a disk device D4. In this way, it is possible to improve processing efficiency even with the small number of trams.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a disk array device in which a plurality of disk devices are arranged and data is distributed and stored in the plurality of disk devices.

[0002]

2. Description of the Related Art A large amount of data is distributed and stored in a plurality of small disk devices, and a RAID (Redundant Arra) is used as a disk array device for accessing these disk devices.
ys of Inexpensive Disks) are known. RAID
There are 5 levels from 1 to 5. At levels 2 to 4, data is distributed and stored on other disks except some disks, and parity information for error checking is stored on the remaining disks. Remember. In RAID level 5, parity information is distributed and stored in all disks. In this type of system, a plurality of communication processors (hereinafter,
A system that speeds up disk access is also proposed by arranging trams) in a tree shape and enabling parallel processing for a plurality of disks.

[0003]

However, in the above-mentioned conventional system, since the number of serial ports possessed by one tram is limited, all trams provided respectively corresponding to a plurality of disk devices can be used. It is impossible to transfer data in parallel at once. For this reason, the trams are arranged in a tree shape so that data is transferred via trams of multiple stages, so that the number of trams increases, the configuration becomes complicated, and the data flow and processing are fixed. As a result, there is a problem that a wasteful time is generated in the processing, and the time until completion of access to a plurality of disks cannot be further speeded up.

The present invention has been made in order to solve such a problem, and it is possible to improve the access efficiency to a plurality of disks with a smaller number of trams than before, and to significantly reduce the processing time. The purpose is to provide a device.

[0005]

DISCLOSURE OF THE INVENTION A disk array device according to the present invention stores data in a plurality of disk devices and one disk device corresponding to each of these disk devices. The disk access means and the data transfer include: a plurality of disk access means for accessing; and a data transfer means for exchanging data with a part of the plurality of disk access means. The means comprises a communication processor (tram) each having a plurality of serial ports and capable of parallel data transfer. Among the plurality of disk access means, the disk access means that does not exchange data with the data transfer means is The process is completed first in the disk access means for exchanging data with the data transfer means. And characterized in that for exchanging the data with the disk access unit.

[0006]

According to the present invention, data is transferred from the data transfer means to a part of the tram as the disk access means provided corresponding to each disk device, and the remaining trams are transferred to other trams. Data is transferred to and from the tram that has completed the processing first, so even if a long time is spent processing a particular tram, the data is transferred via another tram. Will be possible. As described above, according to the present invention in which the data transfer path can be flexibly changed as compared with the conventional device in which the data flow is fixed, the waiting time of the processing can be significantly reduced with a small number of trams to improve the access speed. You can speed it up.

[0007]

Embodiments of the present invention will be described below with reference to the accompanying drawings. FIG. 1 is a block diagram showing the configuration of a disk array device according to an embodiment of the present invention. This device stores data in a distributed manner in five disk devices D0 to D4. Each disk device D0 to D4 has a data tram DT0 to DT as a disk access means.
4 are connected to each other. These data tram D
Each of T0 to DT4 is composed of, for example, a communication processor having four serial ports capable of parallel data transfer, and a transputer (trade name) manufactured by Inmos Corp. or the like can be preferably used. A manager tram MT is provided as a data transfer means for accessing a host computer (not shown), and the manager tram MT is also composed of a communication processor having four serial ports similar to the data trams DT0 to DT4. .

The parallel port of the manager tram MT is connected to a host computer (not shown), and the four serial ports are each one of the data trams DT0 to DT3.
Connected with one serial port. Also, the four serial ports of the data tram DT4 are
It is connected to each one serial port of T0 to DT3. The data trams DT0 to DT3 are the remaining 2
They may be connected to each other via one serial port. The data trams DT0 to DT4 are also provided with SCSI ports (parallel), and the data trams DT0 to DT4 are connected to the disk devices D0 to D4 via these SCSI ports, respectively.

Next, the operation of the disk array device thus constructed will be described. Figure 2 shows RAID level 3
It is a figure for demonstrating ~ 5. Disk device D
If there are four data a, b, c, d to be stored in 0 to D4, as shown in FIG. 2A, in RAID3, each data a, b, c, d has four disk devices. It is divided into D0 to D3 and written in bit units or byte units. The parity information ECC of the divided and written data is written in the remaining disk device D4. In RAID3, a plurality of disk devices D0
Since data is distributed and stored in D3 to D3, long data can be accessed at high speed.

RAID 4 is basically the same as RAID 3, but as shown in FIG. 2B, each data a, b,
c and d are divided in sector units. Therefore, RAID
In 3, it is necessary to simultaneously access all the disk devices, but in RAID 4, individual disk devices can be accessed independently and in parallel.

RAID 5 is, as shown in FIG.
The disk device that stores the parity information is not fixed, and the parity information ECC is embedded in the disk devices D0 to D4 in order.

FIG. 3 is a flow chart showing a write access method when the disk array device of this embodiment is applied to RAID3 to RAID5, and parity information EC
An example of calculating C by the manager tram MT is shown.
When the manager tram MT receives a write command from the host computer (S1, S2), the manager tram MT notifies the data tram D0 to D3 of this (S3), and then caches the data in the internal memory (S
4), while dividing the cached data into four,
The parity information ECC is calculated (S5). Then, each divided data and the parity information ECC are transferred to the data trams DT0 to DT3. At this time, the data transferred to each of the data trams DT0 to DT3 is as follows when corresponding to the data of FIG.

(1) In the case of RAID3 a0, ECC → DT0 a1, ECC → DT1 a2, ECC → DT2 a3, ECC → DT3 (2) In the case of RAID4 a0, ECC → DT0 b0, ECC → DT1 c0, ECC → DT2 d0, ECC → DT3 (3) In the case of RAID5 (at time P in FIG. 2) c0, c3 → DT0 c1, c3 → DT1 ECc, c3 → DT2 c2, c3 → DT3

In the case of RAID5, since the data stored in the disk device D4 is not always the parity information, the data stored in the disk device D4 is checked every access. The manager tram MT is the data tram D
When the data is transferred to T0 to DT3, the write completion notification from these data trams DT0 to DT3 is waited (S
6).

On the other hand, the data tram DT which has received commands and data from the manager tram MT (S7, S8)
Of the received data, 0 to DT3 write the data to be written in the disk device Di connected thereto to the disk devices D0 to D3 (S9), and transfer the remaining data to the data tram DT4 together with the write command (S).
10, S11). Of the data trams DT0 to DT3, the data tram DT4 receives the write command and data from the data tram DTi that has transferred the command and data fastest (S14, S15), and writes the data to the disk device D4 (S16). ). When the data writing is completed (S17), the data trams DT0 to DT
A notification of completion is sent to 3 (S18). The data trams DT0 to DT0 which have received the write completion notification from the data tram DT4
The DT3 sends a write completion notification to the manager tram MT (S13). This completes the writing process.

As described above, according to this embodiment, the data tram DT4 is written by using the data transferred to the data tram DT4 fastest among the data trams DT0 to DT3. The time can be shortened. In particular, not all the data trams DT0 to DT3 access the disk devices D0 to D3 or perform arithmetic processing depending on the situation.
Can be used efficiently, and the processing efficiency can be improved even with a small number of trams.

FIG. 4 shows the parity information for the data tram DT.
0 to DT3 is a flowchart showing an example of calculation on the side, which is particularly effective for RAID 3 and 4. In this embodiment, the manager tram MT transfers all the cached data to the data trams DT0 to DT3 (S2).
1). The data trams DT0 to DT3 calculate parity information ECC based on the transferred data (S2).
2). The data tram DT4 is the data tram D.
Of the T0 to DT3, the parity information ECC is received from the data tram DTi that has completed the parity calculation fastest.

In the above embodiment, the disk device D
Although the write access to 0 to D4 is taken as an example, it goes without saying that the read access can be performed by the same procedure.

[0019]

As described above, according to the present invention, the data access means that does not exchange data with the data transfer means is the data access means that finishes the processing first among the other data access means. Since data is transmitted and received, there is an effect that the waiting time of processing can be significantly shortened and the access speed can be increased.

[Brief description of drawings]

FIG. 1 is a block diagram of a disk array device according to an embodiment of the present invention.

FIG. 2 is a diagram for explaining a data storage form of RAID3 to RAID5.

FIG. 3 is a flowchart showing an example of a data write operation in the same device.

FIG. 4 is a flowchart showing another example of a data write operation in the same device.

[Explanation of symbols]

D0 to D4 ... Disk device, DT0 to DT3 ... Data tram, MT ... Manager tram.

Claims (1)

[Claims] 1. A plurality of disk devices, and a plurality of disk access means which are provided in one-to-one correspondence with these disk devices and which access data to the corresponding disk devices, and a plurality of these disk access devices. Data transfer means for exchanging data with a part of the disk access means of the disk access means, and the disk access means and the data transfer means,
Each of the plurality of disk access means is a communication processor that has a plurality of serial ports and is capable of parallel data transfer. The disk access means that does not exchange data with the data transfer means is A disk array device, characterized in that the data is transferred to and from the disk access means which has completed the processing first among the disk access means for transfer.