JPS63316151A - Data transfer accelerating device - Google Patents

Abstract

PURPOSE:To accelerate data transfer as a whole, by performing the data transfer between a disk memory device and a data buffer device in parallel in each sub system. CONSTITUTION:A data processing system 1 is provided with an I/O processor 2 including a division means 22 which performs a division processing on a logic data block number and performs the division of a disk memory device 5 and that of a physical data lock number, and makes each sub system holding a physical data block included in the range of a logic data block perform the data transfer between the memory device 5 and the data buffer 3 in parallel. A controller 4 provided with the buffer 3 controls the memory device 5 under direct control. Those devices constitute one sub system, and four sub systems are connected to one processor 2, and by responding to an input/output request from the system 1, the operations of all of the sub systems under control can be controlled.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to data transfer control of a disk storage control device, and in particular to a method for increasing the substantial data transfer speed by processing data transfer to multiple disk storage devices in parallel. Regarding.

[Conventional technology]

Conventionally, the data transfer rate could only take a value specific to the disk storage device. However, through the data buffer, the data transfer rate between the data processing system and the data buffer and between the data buffer and the disk storage device can be different. However, when handling large amounts of data, it is not possible to maintain such a data transfer rate continuously.

[Problem that the invention seeks to solve]

Since the data buffer method described above stores data as the difference between the data transfer speeds of two devices, for example, if the data transfer speed on the data processing system side is high and that on the disk storage device side is relatively low. It has the following drawbacks:

■ Prior data is accumulated in a buffer prior to a read request from the data processing system, and high-speed data is transferred from the data buffer at the time of a read request. If you try to read data, you will have to wait until the data is accumulated, so you cannot expect substantial high-speed data transfer.

■ Is the data processing system's write request interval short?
When attempting to write a large amount of data, the data buffer becomes full with the written data, and thereafter the data transfer speed on the disk storage device side drops.

An object of the present invention is to provide a data transfer speed-up device ζt that solves the above-mentioned problems.

[Means for solving problems]

The present invention includes N subsystems including a disk storage device, a controller 4η for controlling data transfer to the disk storage device, and a data buffer device provided in the controller 1, a data processing system, and these N subsystems. directs input/output data transfer to and from the N subsystems, and sequentially records logical data block contents recognized by the data processing system across storage media of N disk storage devices corresponding to N subsystems. When the data processing system specifies the logical data block range and orders sequential reading or sequential writing of the data block contents, the logical data block number should be divided and accessed first. It has a disk storage device, a division means for determining a physical cylinder number, a physical track number, and a physical record number, and operates in parallel for each of the subsystems holding physical data blocks included in the range of the logical data block. The present invention is a data transfer speeding up device characterized in that data transfer between a disk storage device and a data buffer device is directed.

〔Example〕

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

In FIG. 1, the apparatus of the present invention directs input and output data transfers between a data processing system 1 and the subsystems, including four subsystems including a controller 4 with a data buffer 3 and a disk storage device 5. The logical data blocks are allocated so as to be recorded sequentially across the storage media of N disk storage devices corresponding to the N subsystems, and the data processing system defines the range of the logical data blocks and stores the data blocks. When ordering sequential reading or writing of contents, perform division processing on the logical data block number to determine the disk storage device and physical data block number (consisting of physical cylinder number, physical track number, and physical record number). data between a disk storage device 5 and a data buffer 3 in parallel for each of the subsystems holding physical data blocks included in the range of logical data blocks. It is designed to give commands to perform the transfer.

A controller 4 which is superior to the data buffer 3 controls the disk storage device 5 directly under its control.

These constitute one subsystem, and four of the subsystems are connected to one I10 processor 2 to direct the operations of all subordinate subsystems in response to input/output requests from the data processing system 1.

The 110 processors 2 can communicate with the individual controllers 4 and can access data in the individual data buffers 3.

FIG. 2 is a block diagram showing an example of the connection between the I10 processor 2 and one of the subsystems. Commands and parameters issued by the data processing system are transmitted through the interface control line 11.
and is received by the I10 control unit 25 via the data bus 10. The microprocessor section 21 receives and decodes the commands and parameters received by the control section 25 of the machine 10. The microprocessor section 21 instructs the controller 4 via the communication control section 23 to read or write a designated physical data block range. The microprocessor section 41 of the controller 4 controls the disk storage device 5 based on the instructions received through the communication control section 42 . The data transfer control section 43 controls data transfer between the data buffer 3 and the disk storage device 5 through the byte width conversion section 31. The byte width converter 31 converts the byte width on the data buffer 3 side to four times the byte width transferred between the disk storage device 5 and the data buffer 3 . Both the microprocessor section 21 and the microprocessor section 41 can access data in the data buffer 3, and in particular operate block pointers provided in the data buffer 3. 24.32 is a bus control unit.

Figure 3 shows disk storage devices A, B, C, D (DKA, DKB, DKC,
This figure shows the logical data block numbers assigned to the 0 cylinder and 0 track of the DKD. The number of data blocks per track is eight. In the example shown in Figure 3, the logical data block number assignment is 8 to the DKA track.
(block numbers 0 to 7), the next 8 blocks (block numbers 8 to 15) are assigned to the 1-rack of DKB, and in the same way, blocks with consecutive logical data block numbers are sequentially assigned to tracks of DKC and DKD. , DKA to D
Once the same cylinder/track has been identified up to KD, the next logical data block number for the next track on the same cylinder is assigned in the same manner as above. Once the last track on the cylinder has been allocated, the allocation continues from the 0 track of the next cylinder.

In this way, logical data block numbers are assigned to all blocks from DKA to DKD across all cylinders/all tracks. However, these numbers indicate how they are viewed from the data processing system 1, and the I10 processor 2 converts the logical data block number into a physical data block number, that is, physical cylinder number/physical track number, so that the controller 4 can process it. /Convert to physical record number.

Next, an example of the operation of reading data block contents will be explained. First, the data processing system 1 provides the reading start logical data block number and the reading end logical data block number as parameters to the I10 processor 2, and then issues a command to read the data block to the I10 processor 2. 17. The processor 2 uses the starting logical data block number as the dividend and uses the division unit 22 to calculate the divisor "32".
'Divide by (8 blocks x 4 disks). Generally, when the number of blocks per track is n, it is divided by 4n.

The remainder at this time is the R9 quotient as Q. Furthermore, the remainder when Q is divided by the number of tracks per cylinder (m) is r, ioi
Assuming 9, the location of the starting logical data block is as shown in the table below.

(Left below) Since the physical data block number can be obtained for the end logical data block number in the same way, the physical data block range for each disk storage device can be determined. As a result, the I10 processor 2 gives the data block range to each controller 4 and instructs the read operation.

The example in FIG. 3 shows the initial part of the read operation for logical data block numbers 0 to 255. The upper hatched area represents the data portion read into the data buffer 3 under the control of the corresponding controller 4 up to a certain initial point in each disk storage device. Since each disk storage device is independent, the relative rotational position of the storage medium surface is undefined. The upper row in Figure 3 is
These relative rotational positions and the arbitrary data block position at which reading is started are shown randomly.

The controller 4 stores the physical data block number that started within its subsystem in the block pointer of the data buffer 3 and updates it with the physical data block number each time a completely read data block occurs.

The I10 processor 2 can identify the read data block by selecting a subordinate subsystem and referring to the block pointer contents of its data buffer 3.

In FIG. 3, since the starting logical data blocks exist in D and A, the I10 processor 2 first sequentially reads the contents of the stored data blocks from the data buffer 3 of the subsystem including the DKA, and the data processing system Transfer to 1. In FIG. 3, while data blocks for at least one track of the DKA are being transferred, the average data transfer rate is approximately the same as that of the disk storage device 5, but after the data block transfer of this track is completed.

Since the subsequent data block has already been read by the DKB-DKD subsystem, the data transfer rate can be maintained at approximately four times the rate. The lower part of FIG. 3 illustrates this, and the mesh pattern represents data that has already been read.

Furthermore, during this time, each controller 4 reads the remaining data blocks in parallel, so that high-speed data transfer can be performed for the entire remaining data blocks.

Next, an example of the operation of writing data block contents will be explained. The data processing system 1 gives the write start logical data block number and the write end logical data block number as parameters to the 10 processor 2, and then issues a data block write command to the I10 processor 2.

The I10 processor 2 specifies the disk storage device in which the data block to be started is located by performing the same division as described above on the logical data block number, and determines the physical data block range for each disk storage device. As a result, the 10 processor 2 gives the data block range to each controller 4 and instructs the controller 4 to perform a write operation.

The I10 processor 2 initializes and places the block pointer of each data buffer 3, and sequentially receives necessary data block contents from the data processing system 1 from the subsystem having the physical data block assigned to the logical data block to be started. and transfer it to each data buffer 3. The data transfer rate at this time is high. Each controller 4 refers to the block pointer of the data buffer 3 and controls the data blocks already stored in the data buffer to be sequentially written to the disk storage device 5. In regular writing, since the four controllers 4 perform parallel processing, writing is performed at high speed while the difference in the amount of data initially accumulated in the data buffers is almost maintained.

〔Effect of the invention〕

As explained above, the present invention allocates logical data blocks so that they can be processed in parallel across multiple disk storage devices, provides a data buffer in the controller of each disk storage device, and stores the logical data requested by the data processing system. I that includes access to the block
The I10 processor provides each controller with the physical data block number to be accessed by converting the logical data block number to a physical data block number, and selects high-speed data transfer processing with the data processing system via the I10 processor. By transferring data between the data buffer and the disk storage device in parallel, each controller can transfer data at high speed overall when reading or writing large amounts of data sequentially. There is an effect that can be achieved.

[Brief explanation of drawings]

Fig. 1 is a system configuration diagram showing an embodiment of the present invention, Fig. 2 is a block diagram showing an example of connection between the I10 processor and subsystems, and Fig. 3 is an allocation of logical data block numbers to physical data blocks. is a diagram showing an example. 1...Data processing system 2...I10 processor 3
...Data buffer 4...Controller 5...
・Disk storage device

Claims (1)

[Claims] (1) N subsystems including a disk storage device, a controller that controls data transfer to the disk storage device, and a data buffer device provided in the controller, and a communication system between the data processing system and these N subsystems. Directs input/output data transfer and allocates storage media of N disk storage devices corresponding to N subsystems to sequentially record logical data block contents recognized by a data processing system. When the data processing system specifies the range of a logical data block and orders sequential reading or sequential writing of the data block contents, it divides the logical data block number and determines the disk storage device to be accessed first. a disk storage device having a division means for determining a physical cylinder number, a physical track number, and a physical record number, and in parallel for each of the subsystems holding physical data blocks included in the range of the logical data block; 1. A data transfer speed-up device, characterized in that it directs data transfer between a data buffer device and a data buffer device.