JPH01140326A - Magnetic disk device - Google Patents

Abstract

PURPOSE:To shorten the transfer time of data by making access to plural disk devices in parallel, and performing the division and the parallel transfer of the data. CONSTITUTION:A small magnetic disk integration device 2 is constituted of plural small magnetic disk devices 3, a master controller 21 which controls those plural disk devices 3 comprehensively, and busses 221 and 222 which connect the master controller 21 to the disk devices 3 or the disk devices 3 themselves. And the data, even when it is the data with long length, can be read out and written in parallel from the plural disk devices 3 by a function to divide continuous data to the one with appropriate length and write them on the plural disk devices by distributing. In such a way, it is possible to shorten the transfer time of the data.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a disk device, and more particularly, to a five-capacity disk device that integrates a plurality of magnetic disks and performs data transfer at high speed.

[Conventional technology]

As a device that integrates small magnetic disk devices and divides data between each disk device to increase the transfer speed,
An example of this can be found in JP-A-60-57428. However, in this device, the order in which data is read/written from each disk is determined by a higher-level collective disk device control unit. Further, although a direct memory access control unit (hereinafter abbreviated as DMAC) was provided for each disk, a buffer memory was not provided for each disk.

[Problem that the invention seeks to solve]

In the above conventional technology, the order of disk reading or writing is determined by the higher-level controller, so even if disk access is completed, the user must wait until the previous disk device completes data transfer. There wasn't. This method is suitable for systems that often send long continuous data, but is disadvantageous for systems that often read relatively short data sets at the same time.

The first object of the present invention is to perform data transfer at high speed, both in systems where there are many input/output requests for relatively short random data sets (and also in systems where there are many requests to transfer long continuous data). There is a particular thing.

Further, in the above-mentioned conventional technology, since a backup memory is not provided in each disk device, the seek time of the disk appears to be unknown (although the rotational waiting time does not change).

A second object of the present invention is to apparently shorten this rotation waiting time.

[Means for solving problems]

In order to achieve the above object, the present invention includes a plurality of small magnetic disk devices and a host controller that controls the small magnetic disk devices, and when the data length is longer than a predetermined transfer unit length, the transfer unit In a magnetic disk device that is divided into length units and written in parallel to each separate small magnetic disk device, the upper controller is equipped with a buffer, and when reading data, the read data is longer than the transfer unit length. If the data set is completed, it is stored in the above buffer and sent to the host combiator. If the read data is less than the transfer unit length, the data read from the multiple small magnetic disk devices is accessed. It has a function of sending data to the host combinator in the order of completion.

[Effect]

The magnetic disk device according to the present invention has a function of issuing commands to multiple small disk devices in parallel, and when writing data, divides continuous data into required lengths and writes each piece of data distributed to each disk. Function: When reading data from a disk, consecutive data is arranged in a buffer in the upper controller and then transferred to the host system.
A function that sends non-consecutive data to the host computer once it has been read into the buffer, and a function that sends non-consecutive data to the host computer when the order of data to be transferred is specified by the host computer.
It has a function of transferring data from a buffer in a host controller of a disk device to a host combination 1-ta in a specified order according to the aK, and each small magnetic disk device is provided with a buffer memory.

The function of issuing commands to multiple disk devices in parallel makes it possible to control multiple disk devices in parallel.

With the function of dividing continuous data into appropriate lengths and distributing each piece of data to multiple disks for writing, it is possible to shorten transfer time by reading and writing long data from multiple disk devices in parallel. Happened.

When reading data from a disk or a fly, the host does not need to be aware of the fact that the data has been divided, and the data can be read as the original continuous data by rearranging the divided data in order in the upper controller and then transmitting it to the host. You can receive the data.

A function that allows arbitrary data to be sent faster than other data based on commands from the host computer, regardless of the order in which transfer requests arrive or the order in which disk accesses are completed, allows high-priority data to be transferred to the host combinator as quickly as possible. becomes possible.

By providing a buffer memory in each small magnetic disk device, if access to another disk is completed while the bus between the controller and the disk is being used by one disk, the disk that has completed the access will wait for rotation. Data can be stored in a buffer provided in the disk device without the need to do so. This data is transferred to the upper controller's buffer as soon as the bus is opened. Therefore, the apparent transfer waiting time is significantly shortened, and since data is read from the buffer memory, the read time is shortened compared to reading data from a disk.

〔Example〕

An embodiment of the present invention will be described below with reference to FIGS. 1 to 5.

<Configuration of Embodiment> FIG. 1 is a block diagram showing a complete overview of an example of a system of the present invention. As shown in the figure, the small magnetic disk storage device 2 includes a plurality of small magnetic disk devices W (CHDD) 5.
, a mask controller (upper controller) 21 that centrally controls the plurality of small magnetic disk devices 3, a bus 221 that connects the mask controller 21 and the small magnetic disk devices 5 or the small magnetic disk devices 3,
222.

As shown in FIG. 2, the master controller 21 in FIG. 1 includes a port register 211, a directory 212.
3, port registers 214 and 215, and a buffer 216.

In addition, each small disk device shown in FIG. 1 has a port register 67.58, a memory 731. It consists of a heart rate controller (HDC) 62, a central processing unit (CPU) s3, a direct memory access controller (DMAC) 54, and a portion of a disk driver 35.

<Operation of the Embodiment> First, reading data from the disk will be explained.

When a request to read data on the disk comes from the host computer, the signal is transmitted to the directory 212 or 215 through the port register 211. Directory 212 or 213 determines the data length from the transmitted command. When the master controller 21 transfers data to the small magnetic disk device 3, the master controller 21 performs the transfer using a certain predetermined data length as one unit. This length is determined according to the average data length handled by the host system.

The directory determines whether the data length of the requested data is longer than the transfer unit length, and if it is shorter, starts accessing the small magnetic disk device 3 in which the requested data is recorded. When the access to the small disk is completed and an access completion signal is transmitted to the directories 212 and 216 through the port 214 or 215, the directory checks whether the port register 211 is free or not, and if the port register 211 is free, the data is sent directly to the host computer 1. If the buffer 216 is not empty, the data from the small magnetic disk device 3 is transferred to the buffer 216. Port register 21
1 is empty (until the data is stored in the buffer 216,
When the port register 216 becomes empty, data is transferred from the buffer 216 to the host computer 1 using the directory. According to this embodiment, for a data set with a short p1 data length, it is possible to start reading data from a disk device that has completed access earlier, regardless of the order of access requests.

In the above embodiment, data is transferred to the host computer 1 in the order in which accesses are completed, but data may be transferred to the host computer 1 in the order of access requests. The data read out from the disk device fff5 first is stored in the buffer 216 until it is transferred to the host computer 1, and then transferred to the host computer 1 through the directories 212 and 213 when its turn comes. Furthermore, the order of data transfer to the host computer 1 can be arbitrarily changed by the host computer 1. While data is being transferred in the order read from the disk device 5 or in the order of data transfer requests from the host computer 1, if the data transfer request is sent from the host computer 1 with a high priority,
Directories 212 and 215 immediately start accessing the disk device 3 in which the requested data is recorded.

If the port register 211 is in use when the access is completed, the data transferred from the disk device 3 is temporarily stored in the pad 7a 216, and the data transferred from the disk device 3 is stored in the port register 216.
11 is transferred to the host computer 1 as soon as it becomes available.

If the requested data length is longer than the transfer unit length, the directory issues an access command all at once to a plurality of small magnetic disk drives 3 in which the requested data is stored. Upon receiving the access command, each disk device simultaneously starts an operation for reading signals, and writes the data requested by the host to the memory within the small magnetic disk device. This memory has a capacity sufficient to store data of a transfer unit length. Upon completion of writing data to the memory, the disk device sends a signal to the master controller 21 to report that preparations for data transfer have been completed. The directory 212 or 213 of the master controller 21 that received the signal is
The data is stored in the host, and the requested data is transferred to the host in order from the beginning. According to this embodiment, even when the data length is long, the data is read out in parallel from a plurality of disk devices 5, so that the effective transfer speed is improved.

Data is written to the disk device 3 in the same manner.

Next, writing data to the disk will be explained.

If the write data length is less than or equal to the unit transfer length when transferring between the master controller 21 and small disk device #5, the mask controller 2 that received the transfer command
Directory 1 212 or 213 directly transfers the write request data to the designated disk device @5. When the specified disk is being read or written, or when two buses are both in use, the buffer 216
Request data is written to the designated disk device 6, and when the bus and designated disk device 6 become free, writing of data to the disk device is started.

If the write data length is longer than the unit transfer length, the directory 212 or 213 issues an access request all at once to each of the small magnetic disk devices 5 designated as disks into which the data will be written. Then directory 2
12 or 213 transfers data to each small magnetic disk device 3. Each small magnetic disk device 6 is
The data is temporarily stored in the internal memory 51 as shown in FIG.
Store K.

When the access is completed, data is written from the memory 31 to the disk medium using the head 36. If either of the designated disk devices 3 is in the process of reading or writing, the directory 212 or 213 is -
Once the data to be imported to the disk KIIIF is stored in buffer 2.
16, and when the disk device finishes the read/write process, the data is transferred from the buffer 216 to the small disk device t5.

In the small magnetic disk device 3, the master taco/troller 21
The commands sent from the CPU 35 are sent to the CPU 35 through the port register 37 or 38, and the CPU CPtJs3 analyzes the sent commands, and according to the results, the memory 51, the hard disk controller 52, and the DMAC!
54, controls the disk driver 35, etc. The port register S7 is connected to the bus 221, and the port register 38 is connected to the bus 222, and each Novus is used to connect the master controller 21, 6 or other small magnetic disk device r1t.
Data transfer with 3 is performed.

FIG. 4 is a time chart showing the effect of this method.

FIG. 4a shows a conventional method without a memory, and FIG. 4a shows the present method with a memory.

With reference to FIG. 4, a case where four disk devices are accessed at once will be described as an example.

In the conventional method, as shown in Figure a, the order of data transfer from disks is determined by the host in advance, so even if disk b-a completes access earlier than disk a, disk a's data transfer order is determined in advance by the host. Disk b must wait until the data transfer is complete.

■Also, since there is no menu, data transfer on disk A does not end even when the head reaches the location on the recording medium where the requested data is written on disk B. 5) Data cannot be read from disk B. If you can't, you'll just pass by, and then you'll have to wait until that spot comes under the head. As a result, the rotational waiting time becomes long.Furthermore, the data transfer time from the disk device 5 to the mask controller 21 is determined by the data reading speed from the disk device 3.

On the other hand, in this method shown in 19b, (1) data starts to be read from the disk that has been accessed, so the notator transfer to the mask controller starts earlier.

■Memo IJ (Because it has it, data can be stored in memory immediately after disk access is completed.
[The waiting time for one turn is significantly reduced.

(2) The data read into the memory is transferred from the memory to the mask controller 21. Therefore, disk device 3
The transfer speed is improved compared to the data read speed from
Data transfer time becomes shorter.

As a result of the above, the effective transfer time is significantly shortened, as shown in FIG. 4 &b.

〔Effect of the invention〕

According to the present invention, in a small magnetic disk integrated mass storage device, multiple disk devices can be accessed in parallel, and data can be divided and transferred in parallel, thereby reducing data transfer time. (can do.

Since the disk devices can be accessed in parallel, the response time can be shortened for systems where data transfer requests are made randomly.

Furthermore, since each small magnetic disk device has a buffer, the apparent rotational waiting time and data transfer time are shortened.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an overview of an example of a system according to the present invention, FIG. 2 is an internal configuration diagram showing an embodiment of the mask controller shown in FIG. 1, and FIG. FIG. 4, which is an internal configuration diagram showing one embodiment, is a time chart for comparison between the method according to the present invention and the conventional method. DESCRIPTION OF SYMBOLS 1...Host computer, 2...Small magnetic disk integrated device, 21...Mask controller, 224.222...Internal bus, 3...Small magnetic disk device. Representative Patent Attorney Katsoo Ogawa Figure 1 Page 2 Page 3

Claims (1)

[Claims] 1. A plurality of small magnetic disk devices and a host controller that controls the small magnetic disk devices are provided, and if the data length is longer than a certain transfer unit length, the transfer unit length is set to one unit. In the magnetic disk device that is divided into two types of small magnetic disk devices and writes in parallel to each separate small magnetic disk device, the upper controller is equipped with a buffer, and when reading data, if the read data is longer than the transfer unit length, the upper controller transfers data to the buffer. Once the data set is complete, it is sent to the host computer, and if the read data is less than the transfer unit length, the data read from the plurality of small magnetic disk devices is sent to the host computer in the order of access completion. A magnetic disk device characterized by having a function of sending data to a computer. 2. The host controller further has a function of transmitting data read from the plurality of small magnetic disk devices to the host computer in an order according to instructions from the host computer. A magnetic disk device according to scope 1. 3. The method according to claim 1, wherein each of the plurality of small magnetic disk devices includes a buffer memory having a capacity of at least a predetermined transfer unit length for data transfer with the host controller. Magnetic disk device.