JPH07121308A - Disk device write-back control method - Google Patents

Abstract

(57) [Summary] [Object] The present invention relates to a write-back control method for a disk device, which eliminates the rotation waiting time of the disk medium that occurs during write-back processing, and enables high-speed and efficient write-back control. The purpose is to [Structure] When an instruction is received, it is determined whether or not it is a sequential write, and if it is a sequential write, the data is prefetched to the data buffer 7, and the write processing specified by the new instruction is performed by the previous instruction. The writing process is controlled to be continued. In addition, when multiple write commands that are issued consecutively are received, the commands are pre-ordered in the order in which they were received, and then these commands are reordered to determine whether the commands are sequential write commands. For example, as described above, the writing process to the disk medium is controlled to be continued from the writing process performed by the previous command.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a write back control method for a disk device. In recent years, in order to increase the access speed of magnetic disk devices, etc., the speed of disk rotation has increased,
Various physical methods such as increasing track capacity are used, but the most basic ones are shortening of seek time and rotation waiting time.

These two time reduction methods can be realized to some extent by improving the control method of existing hardware. By the way, in recent years, when a write cache mechanism has become quite natural in a magnetic disk device, it is clear that the write-back control method has a great influence on the performance of the write cache mechanism, and by extension, the performance of the magnetic disk device itself.

Therefore, how to efficiently perform the write-back operation becomes a problem, and in order to realize them by hardware, speeding up the write-back operation by shortening the seek time and the rotation waiting time is one of the means. ing.

[0004]

FIG. 22 is a block diagram of a conventional magnetic disk device. In FIG. 22, 1 is a magnetic disk device, 2 is a host (host computer), 3 is an interface control unit, 5 is an input / output control unit, and 6 Is the data buffer controller,
Reference numeral 7 is a data buffer, 8 is a read / write control unit, and 9 is a disk medium.

§1: Description of magnetic disk device--FIG. 2
See FIG. 2. The illustrated magnetic disk device is a device equipped with a write cache mechanism, and the magnetic disk device 1 includes an interface control unit 3, a read / write control unit 8, a disk medium 9, etc. (only those necessary for explanation). (Shown).

Further, the interface control section 3 is provided with an input / output control section 5, a data buffer control section 6, a data buffer 7 and the like. When operating the magnetic disk device, the magnetic disk device 1 is connected to the host 2 for operation. Functions and the like of the above-mentioned respective parts are as follows.

The input / output control unit 5 controls various interfaces with the host 2. The data buffer control unit 6 controls data read / write with respect to the data buffer 7.

The data buffer 7 temporarily stores the write data transferred from the host 2.
Data read / write is performed by the data buffer control unit 6.

The read / write controller 8 controls data read / rat to the disk medium 9. §2: Outline of write-back control in magnetic disk device In a magnetic disk device equipped with a write cache mechanism, control is performed as follows.

When the host 2 issues a data write command (write command), the input / output control unit 5 receives the command. Then, when the input / output control unit 5 analyzes the received command and recognizes that it is a write command, it issues an instruction to the data buffer control unit 6 to temporarily write data transferred from the host 2 to the data buffer 7 Stored in (data queuing).

After that, when the data buffer 7 is full or a command is issued from the host 2, the data in the data buffer 7 is transferred to the disk medium 9
Write to (write back).

In this case, the write back of the data queued in the data buffer 7 is performed in the block unit designated by the write command. The write back control method is as follows.

Even if the data queuing for one command is completed, the data queuing for the next write command is not performed until the write back for the data is completed, and the write back operation for the previous command is completed. ,
A method of starting data queuing of the next write command (write back control method 1).

Even if the device is capable of queuing the data of a plurality of write commands, once the write back for one command is completed, the write processing is temporarily stopped, and at that point, the data buffer 7 still remains. If the data transferred by another write command remains, write back for the next one command is newly started (write back control method 2).

[0015]

SUMMARY OF THE INVENTION The above-mentioned conventional devices have the following problems. In a device in which the capacity of the data buffer equipped in the magnetic disk device is not sufficient for the amount of data transferred from the host in one write command, each time write back is performed for each write command, Waiting for the disk medium to rotate has occurred.

Due to the delay of the write-back operation, the host cannot send the write data to the magnetic disk device and waits until the end of the write-back operation.

Even in the case of sequential write, the writing process (write back) is interrupted for each command, so that rotation waiting occurs. Therefore, the time required for write back becomes long.

In the write back control method 1, the host waits for the transfer of write data for a long time. In the write back control method 2, when the write data is queued one after another, the data buffer is immediately filled with the write data, and as a result, the data queuing cannot be performed. Therefore, the host has to wait for the transfer of the write data for a long time.

The present invention solves such a conventional problem and eliminates the rotation waiting time of the disk medium which occurs during the write-back operation when the host performs the sequential write, and the write-back control is fast and efficient. The purpose is to be able to.

[0020]

FIG. 1 is a diagram for explaining the principle of the present invention. In FIG. 1, the same parts as those in FIG. 22 are designated by the same reference numerals. Further, 12 is a command queue, 14 is a reordering execution unit, and 20 and 21 are hosts (host computers). The present invention has the following configuration to solve the above problems.

In a disk device equipped with a write cache mechanism for storing the data transferred from the host 2 in the temporary data buffer 7 and then writing the data in the data buffer to the disk medium 9, the host issues it. When an instruction is received, it is determined whether the instruction is a sequential write (a continuous data block group on the disk medium is sequentially written over a plurality of instructions), and if the instruction is a sequential write instruction, the disc is The data to be written to the medium 9 is prefetched (data queuing) to the data buffer 7, and the write processing (write back) to the disk medium specified by the new instruction is continued from the write processing performed by the previous instruction. A write-back control method for a disk device that controls to perform.

In the disk device having a write cache mechanism for storing the data transferred from the host 2 in the temporary data buffer 7 and then writing the data in the data buffer to the disk medium 9, the host 2 continuously operates. When receiving multiple write commands issued,
After prefetching (command queuing) the commands in the order in which they were received, these commands are optimized (reordering) in the order in which the writing process is completed in the shortest time, and the optimized commands are sequentially fetched and sequentially. It is determined whether or not it is a write (a continuous data block group on the disk medium is sequentially written over a plurality of instructions) command, and if it is a sequential write command, the data to be written to the disk medium 9 is stored in the data buffer 7. A write-back control method for a disk device that performs pre-fetching (data queuing) and controls write processing (write-back) to a disk medium specified by a new instruction so as to follow the write processing performed by the previous instruction .

A plurality of hosts 20 in a disk device having a write cache mechanism for storing data transferred from the host in the temporary data buffer 7 and then writing the data in the data buffer to the disk medium 9. When the command issued by 21 is received, it is judged whether it is a sequential write command (a continuous data block group on the disk medium is sequentially written over a plurality of commands), and it is judged as a sequential write command. In this case, the data to be written to the disk medium 9 is prefetched (data queuing) in the data buffer 7, and the writing process (write back) to the disk medium designated by the new command is performed by the previous command. The write-back of the disk unit that controls the writing process to continue. Control method.

A plurality of hosts 20 in a disk device having a write cache mechanism for storing the data transferred from the host in the temporary data buffer 7 and then writing the data in the data buffer to the disk medium 9. When 21 receives a plurality of write commands issued continuously, after prefetching (command queuing) in the order in which the commands are received, these commands are optimized in the order in which the write process is completed in the shortest time (reordering). ), The optimized instructions are sequentially fetched, and it is judged whether or not the sequential write (a continuous data block group on the disk medium is sequentially written over a plurality of instructions) is instructed. , The data to be written to the disk medium 9 is prefetched to the data buffer 7 (data Together with queuing) to,
A write-back control method for a disk device, which controls a write process (write-back) to a disk medium designated by a new command so as to continue the write process performed by a previous command.

[0025]

The operation of the present invention based on the above configuration will be described with reference to FIG. : In the magnetic disk device equipped with the write cache mechanism, when the host (one or more devices) performs sequential write, the input / output control unit 5 detects this,
The write data (write data) is queued in the data buffer 7, and the read / write control unit 8 performs the write processing of the next instruction subsequent to the end of the write (write back) processing of the current instruction. Give instructions.

In the case of a magnetic disk device equipped with the command queue 12, the host (one or more machines)
When the I / O controller 5 receives a plurality of commands issued continuously, the I / O controller 5 stacks these commands in the command queue 12 (command queuing), and then the reordering execution unit 14 performs reordering.

When the reordered instruction is a sequential write, the write back process is continuously performed in the same manner as the case of the sequential write.

In this way, when the host performs sequential write, the magnetic disk device 1 detects this and continuously executes the write processing of a plurality of instructions, thereby waiting for the rotation of the disk medium during the write back processing. You can save time.

As a result, the real time of the write back process can be shortened, and at the same time, by activating the output of the data buffer 7, more data can be queued.

[0030]

Embodiments of the present invention will be described below with reference to the drawings. 2 to 21 are views showing an embodiment of the present invention. In FIGS. 2 to 21, the same parts as those in FIGS. 1 and 22 are designated by the same reference numerals. Reference numeral 11 denotes a command information saving memory.

(Description of First Embodiment) §1: Outline Description of First Embodiment In the first embodiment, in a magnetic disk device having a write cache mechanism, a host extends a continuous data block group on a disk medium a plurality of times. When writing sequentially in sequence (hereinafter referred to as “sequential write”), it is detected that the magnetic disk device is sequential write, and the data (write data) to be written to the disk medium is prefetched in the data buffer (hereinafter “data”). Queuing ").

By performing the writing process designated by the new command subsequently to the writing process performed by the previous command, the rotation waiting time of the disk medium is eliminated and the access speed is improved.

§2: Description of magnetic disk device--FIG. 2
Reference FIG. 2 is a block diagram of the magnetic disk device of the first embodiment.
In the first embodiment, SCSI is used as an interface between the host (host computer) 2 and the magnetic disk device 1.
(Small Computer System Interface) is adopted, and the magnetic disk device is a device equipped with command queuing and write cache mechanism.

As shown in the figure, the magnetic disk device 1 includes
Interface controller 3, read / write controller 8,
A disk medium 9 and the like (only those necessary for explanation are shown) are provided.

Further, the interface control section 3 is provided with an input / output control section 5, a data buffer control section 6, a data buffer 7, a command information saving memory 11, a command queue 12 and the like.

When operating the magnetic disk device, the magnetic disk device 1 (1 machine) is connected to the host 2 (1 machine) for operation. Functions and the like of the above-mentioned respective parts are as follows. The input / output controller 5 controls various interfaces with the host 2. When the input / output control unit 5 receives a write command from the host 2,
The command is analyzed and the analysis result is stored in the command information saving memory 11.

When another write command is newly received during execution of the write command, the command is stored in the command queue 12, and when the write command being executed is completed, the command queue 12 is restarted. Get the command to be processed.

The write data received from the host 2 is
The data buffer controller 6 stores the data in the data buffer 7. The read / write control unit 8 is instructed to start the write back operation.

The data buffer control section 6 controls data read / write with respect to the data buffer 7. The data buffer control unit 6 has two pointers for writing and reading to the data buffer 7, and using these pointers, input / output control is performed so that underrun and overrun do not occur in the data buffer 7. The data transfer between the unit 5 and the read / write control unit 8 and the data buffer 7 is adjusted.

The data buffer 7 has two ports, a read port and a write port, and is a buffer for saving the queued write data. The data buffer control unit 6 writes / reads the data. Is done.

In the first embodiment, the data buffer 7 is 32 [K
The data buffer has a ring buffer structure composed of RAM having a capacity of [Byte]. : The command information saving memory 11 includes the input / output control unit 5
Is a memory for storing the analysis result of the write command (write command issued by the host) analyzed by (host information is temporarily saved).

The command queue 12 is a memory for saving a newly received command during execution of a write command. In the case of the first embodiment, the command queue 12 is a command queue having a ring buffer structure composed of RAM having a capacity of 4 [KByte].

The read / write controller 8 receives an instruction from the input / output controller 5 and writes back the data in the data buffer 7. The read / write controller 8 is composed of hardware for accessing the disk medium and firmware for controlling the hardware.

The disk medium 9 is a data recording medium. In the first embodiment, the data byte length is LBA (Logical Block Addres) formatted to 4096 bytes.
s) (logical block address) from “000000” h to ““ 00F00
A medium having 0 "h.

§3: Outline of operation of magnetic disk device The outline of operation of the magnetic disk device is as follows. When the host 2 issues a command, the input / output control unit 5 receives and analyzes this command. When the input / output control unit 5 recognizes that the command is a write command, it instructs the data buffer control unit 6 to store data. In response to this instruction, the data buffer control unit 6 stores the data transferred from the host 2 in the data buffer 7 (data queuing).

At this time, at the same time, the input / output control section 5 reads / reads.
The write control unit 8 is notified of the LBA (= “000000” h) and the number of transfer blocks, and is instructed to start the writing process.
In response to the instruction, the read / write control unit 8 searches for a target data block, and when the block is detected, starts writing (write-back) the data stored in the data buffer 7.

On the other hand, the input / output control unit 5 receives from the host 2
When all the write data are received, the host 2 is informed of the “GOOD” status (report that the write command data is normally received), and the next command waiting state is entered.

Further, the read / write control unit 8 stops when the writing process instructed by the input / output control unit 5 is completed. §4: Description of write command--see FIG. 3 FIG. 3 is an explanatory diagram of the first embodiment (an example of a write command). An example of the write command issued by the host will be described below with reference to FIG.

In FIG. 3, examples of the write command described in the first embodiment are indicated by the reference signs (a) to (h). For each write command indicated by the symbols (a) to (h), a CDB (Command Description Block:
Command description block) [hex], write start LBA
(Logical block address), the number of transfer blocks, and the number of transfer bytes are shown.

For example, in the write command shown by the code (a), the CDB is "0A 00 00 00 01 00", the write start address is "000000" h, the transfer block number is "01" h, and the transfer byte number is 4 It is [KByte].

The write command indicated by the symbol (b) is C
The DB is “0A 00 80 00 05 00”, the write start address is “008000” h, the transfer block number is “05” h, and the transfer byte number is 20 [KByte].

The write command indicated by the code (c) is C
The DB is “0A 00 80 05 05 00”, the write start address is “008005” h, the transfer block number is “05” h, and the transfer byte number is 20 [KByte].

The write command indicated by the symbol (d) is C
The DB is “0A 00 80 0A 05 00”, the write start address is “00800A” h, the transfer block number is “05” h, and the transfer byte number is 20 [KByte]. The other commands are also shown below.

§5: Explanation of processing of the first embodiment by a flow chart--See FIGS. 4 and 5. FIG. 4 is a processing flowchart 1 of the first embodiment, and FIG. 5 is a processing flowchart 2 of the first embodiment. The process of the first embodiment will be described below with reference to FIGS. 4 and 5. Note that S1 to S
Reference numeral 17 indicates a processing step.

This processing is an example of processing when the host 2 sequentially issues the write commands (b), (c) and (d) shown in FIG. First, it is assumed that the host 2 issues the write command (b). At this time, the input / output control unit 5
The command is received (S1) and its contents are analyzed (S2). As a result of the analysis, if it is confirmed that the write command is for blocks of LBA "008000" h to "05" h (S
3) The information (CDB analysis result) is stored in the command information saving memory 11 and saved (S4). When a command other than the write command is received, other processing is performed.

At this time, the input / output control unit 5 notifies the read / write control unit 8 of the information of the analysis result, instructs the start of the write operation, and at the same time starts receiving the write data from the host 2 (write operation). Start data transfer).

In this way, while the I / O controller 5 is executing data transfer, the LBA "008" on the disk medium 9 is
Seek to a track with 000 ”h blocks,
A target data block is searched (S5).

Then, the read / write controller 8 starts writing the data queued in the data buffer 7 to the disk medium 9 when the target data block is detected. That is, the read / write controller 8
Starts writing back to the disk medium 9 when the target data block is detected.

On the other hand, the input / output control unit 5 gives an instruction to the data buffer control unit 6 to start the process of storing the write data in the data buffer 7 (data queuing) (S6). That is, data queuing to the data buffer 7 and write back from the data buffer 7 to the disk medium 9 are independently and separately performed.

After that, when the input / output control unit 5 confirms that the data queuing of the command (b) is normally completed (S7), the input / output control unit 5 instructs the host 2 to complete the processing. The “GOOD” status to the effect is reported (S8).

By this report, the host 2 is informed that the next command can be executed (usually, at this point, the read / write controller 8 still seeks, detects the target data block, or executes the write operation). Inside).

Then, the host 2 sends a write command (c).
Is issued. At this time, the input / output control unit 5 receives the command (S9) and analyzes the content thereof (S1).
0). As a result, if it is confirmed that the write command is for LBA “008005” h to “05” h blocks (S
11), the information (CDB analysis result) is stored in the command information saving memory 11 and saved (S12). In addition,
When a command other than the write command is received, when the write-back processing of the write command (b) is completed,
Perform other processing.

The input / output control unit 5 compares the last write LBA of the write command (b) executed last time with the last write LBA of the write command (c) received this time, and confirms whether these LBAs are continuous. It is determined whether or not it is a sequential write (S1).
3).

As a result, if the LBAs are not continuous (not sequential write), the write command (b)
When the write-back process is completed, another process is performed. However, in this example, the last LBA of the write command (b)
Is “008004” h and the write start LBA of the write command (c) is “008005”, the input / output control unit 5
Judge as a sequential write.

When it is determined that the write command is sequential write (S13), the input / output control unit 5
Queues the write data of the write command (c) in the empty area of the data buffer 7 (S14).

Here, when the data buffer 7 runs out of free space during data queuing, the data which has been written to the disk medium 9 is overwritten. This control is performed by the data buffer control unit 6 according to an instruction from the input / output control unit 5.

Further, if the read / write control unit 8 is in operation at this time, the input / output control unit 5 continues the write process of the write command (b) currently being executed, and subsequently receives the write command ( The read / write control unit 8 is instructed to perform the write back of (c) (S1).
5).

Even if the read / write control unit 8 receives the instruction of the input / output control unit 5 and the write process (write back process) of the write command (b) ends, the read / write control unit 8 does not stop and continues to write command (c). ) The writing process (write back process) is performed.

On the other hand, when the data queuing of the write command (c) is completed (S16), the input / output control unit 5
The "GOOD" status is reported to the host 2 (S17).

The host 2, which has received the "GOOD" status, subsequently issues a write command (d).
In this example, since the write command (d) is also the target of sequential writing, the processing is performed in the same manner as the processing S9 to S17. After that, even when another command is issued, the same processing is performed.

After the host 2 issues the above command,
When the magnetic disk device is not accessed, in the magnetic disk device, the I / O controller 5 reports the “GOOD” status for this command to the host 2,
Waiting for the next command, the read / write controller 8
The process is stopped when all the write-back processes are completed.

§6: Description of processing of the first embodiment based on a time chart--see FIG. 6 FIG. 6 is a time chart of the first embodiment. Hereinafter, the processing of the first embodiment will be described based on the time chart of FIG. The processing content is the same as that described above.

In FIG. 6, (A) is host 2, (B)
Shows a time chart of each processing of the input / output control unit 5 and (C) of the read / write control unit 8. Also, [1] to [1
7] is the timing of each process, and the contents are as follows.

In [1], the host 2 issues the write command (b). In [2], the input / output control unit 5 receives the command. In [3], the input / output control unit 5 instructs the read / write control unit 8 to write back the write command (b). In [4], seek is performed to the target block on the disk medium. In [5], the input / output control unit 5 starts data queuing of the write command (b).

At [6], data transfer of the write command (b) is performed from the host 2 to the input / output control unit 5. In [7], the input / output control unit 5 reports the “GOOD” status of the write command (b) to the host 2. In [8], the host 2 receives the "GOOD" status. In [9], the host 2 issues the write command (c). [1
0], the input / output control unit 5 receives the command.

In [11], the input / output control unit 5 causes the read / write
The write controller 8 is instructed to write back the write command (c) subsequent to the write command (b). In [12], the read / write control unit 8 writes the data of the write command (b) (write back to the disk medium).

At [13], the input / output control unit 5 starts the data queuing of the write command (c). [14]
Then, the data transfer of the write command (c) from the host 2 to the input / output control unit 5 is started. In [15], read /
The write controller 8 writes the data of the write command (c) (writes back to the disk medium).

In [16], the input / output control unit 5 reports the “GOOD” status to the host 2. In [17], the host 2 receives the "GOOD" status. (Explanation of Second Embodiment) §1: Outline of Second Embodiment In the second embodiment, a mechanism for prefetching commands from the host to the magnetic disk device (hereinafter referred to as “command queuing function”)
It has a command queue necessary for that purpose.

When the magnetic disk device prefetches a plurality of write commands, these commands are optimized in such an order that the write processing is completed in the shortest time (hereinafter referred to as "reordering"), and the result is sequential. If it is the same as the write, those commands are continuously written. As a result, the access speed can be improved.

Since the write command example of the first embodiment shown in FIG. 3 is the same as that of the second embodiment, it will be described with reference to FIG. §2: Description of configuration of magnetic disk device--see FIG. 7 FIG. 7 is a magnetic disk device block diagram of the second embodiment.
The magnetic disk device of the second embodiment is a device in which a reordering execution unit 14 is further added to the magnetic disk device of the first embodiment (see FIG. 2), and other configurations are the same as those of the first embodiment.

The reordering executing section 14 reorders the commands in the command queue 12 (which are queued write commands issued by the host) (process for rearranging the order). The rest of the configuration is the same as that of the first embodiment, so the description is omitted.

§3: Description of reordering process ... See FIG. 8. FIG. 8 is an explanatory diagram of reordering according to the second embodiment.
Shows Example 1 and FIG. 8B shows Example 2. In each figure, the left side shows before reordering and the right side shows after reordering.

When the input / output control unit 5 receives the write command issued by the host 2, it stores the command in the temporary command queue 12 in the order of reception. afterwards,
The reordering execution unit 14 performs reordering.

In this reordering, not the order in which the commands are received, but the continuity of the LBA is checked, and the order of the commands is exchanged so that the commands can be processed in the shortest time. For example, the host 2 sends a write command (d) →
Assuming that the commands are sequentially issued in the order of (c) → (b), the input / output control unit 5 outputs the received command as in Example 1.
Once stored in the command queue 12 in this order. After that, reordering is performed and the commands in the command queue 12 are exchanged in the order of (b) → (c) → (d).

Also, the host 2 sends a write command (h)
Assuming that the commands are sequentially issued in the order of → (g) → (f), the input / output control unit 5 temporarily stores the received commands in the command queue 12 in this order as in Example 2.
After that, reordering is performed, and (f) → (g) →
Replace in the order of (h).

§4: Description of processing of the second embodiment by flowchart ... See FIGS. 9 and 10. FIG. 9 is a processing flowchart 1 of the second embodiment, and FIG. 10 is a processing flowchart 2 of the second embodiment. Hereinafter, FIG. 9 and FIG.
The processing of the second embodiment will be described based on 0. Note that S2
1 to S41 indicate processing steps.

This processing is performed by the host 2 without waiting for the "GOOD" status report from the magnetic disk device.
Write commands (d), (c), and (b) shown in
It is an example of processing in the case of continuously issuing in the order of (d) → (c) → (b).

The input / output control unit 5 receives from the host 2 (d) →
The CDBs of the write commands continuously issued in the order of (c) → (b) are received (S21), and these write commands are sent to the temporary command queue 12 as (d) → (c) → (b).
Are stored in the order of (reception order) (see before reordering in FIG. 8A) (S22).

After that, the reordering execution section 14 carries out reordering (S23). By this reordering, the reordering execution unit 14 switches the processing of these commands to the order in which it will be completed in the shortest time (see after reordering in FIG. 8A) (S24).

Next, the input / output control unit 5 takes out the CDB of the command to be processed first from the command queue 12 (S25) and analyzes it (S26). In this example, the write command (b) is first taken out and analyzed.

As a result of the analysis, LBA "008000" h to "05" h
When it recognizes that it is a write command for blocks,
The information (CDB analysis result) is stored in the command information saving memory 11 and saved (S28). When a command other than the write command is received, other processing is performed.

At this time, the input / output control unit 5 notifies the read / write control unit 8 of the information on the analysis result, instructs the start of the write operation, and at the same time starts receiving the write data from the host 2 (write Start data transfer).

In this way, while the I / O controller 5 is executing data transfer, the LBA "008" on the disk medium 9 is
Seek to a track with 000 ”h blocks,
A target data block is searched (S29).

Then, the read / write controller 8 starts writing the data queued in the data buffer 7 to the disk medium 9 when the target data block is detected. That is, the read / write controller 8
Starts writing back to the disk medium 9 when the target data block is detected.

On the other hand, the input / output control unit 5 gives an instruction to the data buffer control unit 6 to start the process of storing the write data in the data buffer 7 (data queuing) (S30).

After that, when the input / output control unit 5 confirms that the data queuing of the command (b) has been normally completed (S31), the input / output control unit 5 instructs the host 2 to complete the processing. The "GOOD" status to the effect is reported (S32).

With this report, the host 2 is informed that the next command can be executed (usually, at this point, the read / write controller 8 still seeks, detects the target data block, or executes the write operation). Inside).

Next, the input / output control unit 5 takes out the command to be processed next, that is, the CDB of the write command (c) in this example from the command queue 12 (S33),
This command is analyzed (S34).

As a result of the analysis, LBAs "008005" h to "05" h
When the write command is recognized for the block (S35), the information (CDB analysis result) is stored in the command information save memory 11 and saved (S36). When a command other than the write command is received, other processing is performed.

In the input / output control unit 5, whether the last write LBA of the write command (b) executed last time is continuous with the write start LBA of the write command (c) currently being processed, that is, the sequential It is determined whether or not it is a light (S37).

As a result, if the LBAs are not continuous (not sequential write), the write command (b)
When the write-back process is completed, another process is performed. However, in this example, the last LBA of the write command (b)
Is “008004” h and the write start LBA of the write command (c) is “008005”, the input / output control unit 5
Judge as a sequential write.

When it is determined that the write command is sequential write (S37), the input / output control unit 5
Queues the write data of the write command (c) in the empty area of the data buffer 7 (S38).

Here, when the data buffer 7 runs out of free space during data queuing, the data overwritten on the disk medium 9 is overwritten. This control is performed by the data buffer control unit 6.

Further, if the read / write control unit 8 is in operation at this time, the input / output control unit 5 continues the writing process of the write command (b) currently being executed,
The read / write controller 8 is instructed to write back the write command (c) (S39).

When the read / write control unit 8 receives an instruction from the input / output control unit 5 and the write process (write back process) of the write command (b) ends, the read / write control unit 8 does not stop and continues to write ( Control is performed so as to perform the writing process (write back process) of c).

On the other hand, when the data queuing of the write command (c) is completed (S40), the input / output control unit 5
The "GOOD" status is reported to the host 2 (S41).

In the input / output control unit 5 which has reported the "GOOD" status, the command to be processed next, that is, the CDB of the write command (d) in this example, is taken out from the command queue 12.

In this example, since the write command (d) is also the target of sequential writing, the same processing as the above processing is performed. Therefore, when the host 2 does not access the magnetic disk device after continuously issuing the write commands (d), (c), and (b), in the magnetic disk device, the input / output control unit 5 is After the "GOOD" status for the command is reported to the host, it waits for the next command. Therefore, the read / write control unit 8 stops at the time when all the write back operations are completed.

§5: Description of processing of the second embodiment based on a time chart--see FIG. 11 FIG. 11 is a time chart of the second embodiment. Below, Figure 1
The process of the second embodiment will be described based on the time chart of No. 1. The processing content is the same as that described above.

In FIG. 11, (A) is the host 2,
7B is a time chart of each processing of the input / output control unit 5, FIG. 9C is the reordering execution unit 14, and FIG. Further, [1] to [21] are timings of each processing, and the contents thereof are as follows.

In [1], the host 2 issues the write command (d). In [2], the input / output control unit 5 receives the command and stores it in the command queue 12.
In [3], the host 2 issues a write command (c). In [4], the input / output control unit 5 receives the command and stores it in the command queue 12. In [5], the host 2 issues the write command (b). In [6], the input / output control unit 5 receives the command, and the command queue 1
Store in 2.

In [7], the reordering execution unit 14 executes reordering, and the write command is (b) →
It is determined in the order of (c) → (d). In [8], the input / output control unit 5 sends a write command (b) to the read / write control unit 8.
Write back. In [9], seek is performed to the target block on the disk medium. In [10], the input / output control unit 5 starts data queuing of the write command (b).

In [11], the input / output control unit 5 from the host 2
Data transfer of write command (b) is performed. [12]
Then, the input / output control unit 5 reports the “GOOD” status of the write command (b) to the host 2. [13]
Then, the host 2 receives the "GOOD" status. In [14], the input / output control unit 5 causes the command queue 12
The write command (c) is taken out from.

In [15], the input / output control unit 5 instructs the read / write control unit 8 to write back the write command (c) subsequent to the write back of the write command (b). In [16], the read / write controller 8
Write data of command (b) (disk medium 9
Write back to).

At [17], the input / output control unit 5 starts the data queuing of the write command (c). [18]
Then, data transfer of the write command (c) is performed from the host 2 to the input / output control unit 5. In [19], the input / output control unit 5 sends the write command (c) “GOOD” to the host 2.
Report status. In [20], the host 2 receives the "GOOD" status.

At [21], the read / write controller 8 writes the data of the write command (c) (writes back to the disk medium 9). The processing of the write command (d) is the same as the processing of the write command (c).

(Explanation of Third Embodiment) §1: General Description of Third Embodiment In the third embodiment, one magnetic disk device having the same configuration as that of the second embodiment is replaced by a plurality of hosts (this (2 in the example)
In the environment where is connected,
When write commands are received from a plurality of hosts and the commands become sequential writes, those commands are continuously written.

As a result, the rotation waiting time of the disk medium is eliminated and the access speed is improved. Since the write command example shown in FIG. 3 is the same in the third embodiment,
The explanation will be made with the aid of the above.

§2: Description of configuration of magnetic disk device
-Refer to FIG. 12 FIG. 12 is a magnetic disk device block diagram of the third embodiment. As shown in the figure, the magnetic disk device 1 is provided with an interface controller 3, a read / write controller 8, a disk medium 9, and the like.

Further, the interface control section 3 is provided with an input / output control section 5, a data buffer control section 6, a data buffer 7, a command information saving memory 11, a command queue 12 and the like.

At the time of operation, two hosts 20, 21 are connected to the magnetic disk device 1. Since the configuration of the magnetic disk device 1 is the same as that of the first embodiment, detailed description will be omitted.

§3: Processing explanation of Embodiment 3 by flow chart ... See FIGS. 13 to 15. FIG. 13 is a processing flowchart 1 of Embodiment 3, FIG. 14 is a processing flowchart 2 of Embodiment 3, and FIG. 15 is an embodiment. It is a processing flowchart 3 of 3. Hereinafter, the process of the third embodiment will be described with reference to FIGS. Note that S51 to S
74 indicates a processing step.

In this embodiment, one host 20
In this example, the other host 21 issues the write command (f) before the data queuing of the write command (e) is issued.

For example, assume that the host 20 issues a write command (e). At this time, the input / output control unit 5
The command is received (S51), and its contents are analyzed (S52). As a result of the analysis, LBA "008000" h to "10" h
When it is confirmed that the write command is for a block (S53), the information (CDB analysis result) is stored in the command information saving memory 11 and saved (S54). When a command other than the write command is received, other processing is performed.

At this time, the input / output control unit 5 notifies the read / write control unit 8 of the information of the analysis result, instructs the start of the write operation, and at the same time starts receiving the write data from the host 21 (write operation). (Start data transfer)
To do.

In this way, while the I / O controller 5 is executing data transfer, the LBA "008" on the disk medium 9 is
Seek to a track with 000 ”h blocks,
A target data block is searched (S55).

Then, the read / write control unit 8 starts writing the data queued in the data buffer 7 to the disk medium 9 when the target data block is detected. That is, the read / write controller 8
Starts writing back to the disk medium 9 when the target data block is detected.

On the other hand, the input / output control unit 5 gives an instruction to the data buffer control unit 6 to start the process of storing the write data in the data buffer 7 (data queuing) (S56). That is, data queuing to the data buffer 7 and write back from the data buffer 7 to the disk medium 9 are independently and separately performed.

When performing the data queuing, the input / output control unit 5 determines whether or not there is a free area in the data buffer 7 (S57). Then, if there is a free area, data queuing is continued.

After that, when the input / output control unit 5 confirms that the data queuing of the command (e) has been normally completed (S58), the input / output control unit 5 instructs the host 20 to end the processing. The “GOOD” status to the effect is reported (S59).

However, when there is no free area in the data buffer 7 due to the data queuing of the write command (e) (S57), the input / output control unit 5 suspends the data queuing (S70).

Then, the input / output control unit 5 receives the write command (f) from the host 21 (S71) and stores it in the command queue 12 (S72). Data queuing of the write command (e) is performed as soon as an empty area is created in the data buffer 7 by write back (S7).
3) The process is restarted (S74). In this case, the data of the data block for which writing has been completed is overwritten.

After reporting the "GOOD" status in the processing (S59), the input / output control unit 5 takes out the CDB of the write command (f) to be processed next from the command queue 12 (S60) and analyzes it. (S61).

As a result, when it is confirmed that the write command is for LBA "008010" h to "01" h blocks (S62), the information (CDB analysis result) is stored in the command information saving memory 11. And save (S63).
When a command other than the write command is received, another process is performed after the write back process of the write command (b) is completed.

Further, the input / output control unit 5 compares the last write LBA of the write command (e) executed last time with the write LBA of the write command (f) currently being processed, and these LBAs are consecutive. It is determined whether or not there is, that is, whether or not it is a sequential write (S
64).

As a result, if the LBAs are not continuous (not sequential write), the write command (e)
When the write-back process is completed, another process is performed. However, in this example, it is determined that the write commands (e) and (f) are sequential writes.

Then, the input / output control unit 5 queues the write data of the write command (f) in the empty area of the data buffer 7 (the area in which the data of the data block for which the writing is completed is stored) (S65). ).

Further, if the read / write control unit 8 is in operation at this time, the input / output control unit 5 continues the write process of the write command (e) currently being executed,
The read / write control unit 8 is instructed to write back the write command (f) received this time (S6).
6).

When the read / write control unit 8 receives an instruction from the input / output control unit 5, even if the write process (write back process) of the write command (e) is completed, the read / write control unit 8 does not stop and continues to write ( The writing process (write back process) of f) is controlled.

On the other hand, when the data queuing of the write command (e) is completed (S67), the input / output control section 5
The "GOOD" status is reported to the host 2 (S68).

After that, when neither host accesses the magnetic disk device, in the magnetic disk device, the input / output control unit 5 waits for the next command, and the read / write control unit 8 completely The process is stopped when the write back process is completed.

§4: Description of Processing of Third Embodiment by Time Chart ... See FIG. 16 FIG. 16 is a time chart of the third embodiment. Below, Figure 1
The processing of the third embodiment will be described based on the time chart of No. 6. The processing content is the same as that described above.

In FIG. 16, (A) is the host 20,
(B) is the host 21, (C) is the input / output control unit 5, (D)
Is a data buffer control unit 6, and (E) is a time chart of each process of the read / write control unit 8. Also,
[1] to [21] are timings of each processing, and the contents thereof are as follows.

In [1], the host 20 issues the write command (e). In [2], the input / output control unit 5 receives the command. In [3], the input / output control unit 5 instructs the read / write control unit 8 to write back the write command (e). In [4], seek is performed to the target block on the disk medium. In [5], the input / output control unit 5 starts data queuing of the write command (e).

In [6], the input / output control unit 5 is transferred from the host 20.
Data transfer of the write command (e) is performed. In [7], there is no free area in the data buffer 7, and the data transfer of the write command (e) to the data buffer 7 is temporarily stopped.

At [8], the host 21 issues a write command (f). In [9], the input / output control unit 5 receives the command and stores it in the command queue 12. [1
0], the data buffer control unit 6 causes the data buffer 7
The data transfer of the write command (e) to is resumed.
In [11], data transfer (continuation) of the write command (e) is performed from the host 20 to the input / output control unit 5.

In [12], the input / output control unit 5 sets the host 20
Report the "GOOD" status of the write command (e). In [13], the host 20 receives the "GOOD" status. In [14], the input / output control unit 5 takes out the write command (f) from the command queue 12.

In [15], the input / output control unit 5 instructs the read / write control unit 8 to perform write back of the write command (f) subsequent to write back of the write command (e). In [16], the read / write controller 8 writes the data of the write command (e) (writes back to the disk medium 9).

At [17], the input / output control unit 5 starts the data queuing of the write command (f). [18]
Then, the write command (f) is transferred from the host 21 to the input / output control unit 5. In [19], the input / output control unit 5 sends the write command (f) “GOOD” to the host 21.
Report status. In [20], the host 21 receives the “GOOD” status. In [21], the read / write controller 8 writes the data of the write command (f) (writes back to the disk medium 9).

(Explanation of Fourth Embodiment) §1: General Description of Fourth Embodiment In the fourth embodiment, in the environment similar to that of the third embodiment, the magnetic disk device executes a plurality of write commands from a plurality of hosts. When prefetched, these instructions are reordered, and if the result is similar to sequential write, writing of these instructions is performed continuously. As a result, improvement in access speed can be achieved.

§2: Description of configuration of magnetic disk device
-Refer to FIG. 17 FIG. 17 is a block diagram of a magnetic disk device according to the fourth embodiment. Since the magnetic disk device of the fourth embodiment is the same as the magnetic disk device of the second embodiment (see FIG. 7), the description thereof will be omitted. However, during operation, two hosts 20, 21 are connected to this magnetic disk device.

§3: Description of the process of the fourth embodiment by a flow chart ... See FIGS. 18 to 20. FIG. 18 is a process flow chart 1 of the fourth embodiment, FIG. 19 is a process flow chart 2 of the fourth embodiment, and FIG. It is the processing flowchart 3 of 4. Hereinafter, the processing of the fourth embodiment will be described with reference to FIGS. Note that S81 to S
Reference numeral 105 denotes a processing step.

In this example, before one host 20 issues a write command (e) and the data queuing of the command is completed (“GOO from the magnetic disk device”).
This is an example in which the other host 21 issues write commands (h) → (g) → (f) in this order (without waiting for the “D” status).

For example, assume that the host 20 issues a write command (e). At this time, the input / output control unit 5
The command is received (S81), and its contents are analyzed (S82). As a result of the analysis, LBA "008000" h to "10" h
When it is confirmed that the write command is for a block (S83), the information (CDB analysis result) is stored in the command information saving memory 11 and saved (S84). When a command other than the write command is received, other processing is performed.

At this time, the input / output control unit 5 notifies the read / write control unit 8 of the information of the analysis result, instructs the start of the write operation, and at the same time starts receiving the write data from the host 20 (write operation). (Start data transfer)
To do.

In this way, while the input / output control unit 5 is executing data transfer, the LBA "008" on the disk medium 9 is
Seek to a track with 000 ”h blocks,
A target data block is searched (S85).

Then, when the target data block is detected, the read / write controller 8 starts writing (write back) the data queued in the data buffer 7 to the disk medium 9.

On the other hand, the input / output control unit 5 gives an instruction to the data buffer control unit 6 to start the process of storing the write data in the data buffer 7 (data queuing) (S86).

When performing the data queuing, the input / output control unit 5 determines whether or not there is a free area in the data buffer 7 (S87). Then, if there is a free area, data queuing is continued.

After that, when the input / output control unit 5 confirms that the data queuing of the command (e) has been normally completed (S88), the input / output control unit 5 instructs the host 20 to complete the processing. The “GOOD” status indicating that it has been reported is reported (S89).

However, when there is no free area in the data buffer 7 due to the data queuing of the write command (e) (S87), the input / output control unit 5 suspends the data queuing (S99).

Then, the input / output control unit 5 receives the write commands (h), (g), and (f) continuously issued by the host 21 in the order of (h) → (g) → (f) (S100). ), And the command queue 12 in the order of (h) → (g) → (f).
(S101).

Thereafter, the reordering executing section 14 carries out reordering (S102). As a result, the write commands are exchanged in the order of (f) → (g) → (h),
It is stored in the command queue 12 (S103).

In the data queuing of the write command (e), the processing is restarted as soon as a free area is created in the data buffer 7 by the write back (S104) (S1).
05). In this case, the data of the data block for which writing has been completed is overwritten.

After reporting the "GOOD" status in the processing (S89), the input / output control unit 5 takes out the CDB of the write command (f) to be processed next from the command queue 12 (S90) and analyzes it. (S91).

As a result, when it is confirmed that the write command is for LBA "008010" h to "01" h blocks (S92), the information (CDB analysis result) is stored in the command information saving memory 11. And save (S93).
When a command other than the write command is received, another process is performed after the write back process of the write command (e) is completed.

Further, the input / output control section 5 compares the last write LBA of the write command (e) executed last time with the write LBA of the write command (f) currently being processed, and these LBAs are consecutive. It is determined whether or not there is, that is, whether or not it is a sequential write (S
94).

As a result, if the LBAs are not continuous (not sequential write), the write command (e)
When the write-back process is completed, another process is performed. However, in this example, it is determined that the write commands (e) and (f) are sequential writes.

Then, the input / output control unit 5 queues the write data of the write command (f) in the empty area of the data buffer 7 (the area in which the data of the data block for which writing has been completed was stored) (S95). ).

Further, if the read / write control unit 8 is in operation at this time, the input / output control unit 5 continues the write process of the write command (e) currently being executed,
The read / write controller 8 is instructed to write back the write command (f) received this time (S9).
6).

When the read / write control unit 8 receives an instruction from the input / output control unit 5, even if the write process (write back process) of the write command (e) is completed, the read / write control unit 8 does not stop and continues to write ( The writing process (write back process) of f) is controlled.

On the other hand, when the data queuing of the write command (f) is completed (S97), the input / output control section 5
The "GOOD" status is reported to the host 21 (S98). Thereafter, the write commands (g) and (h) are similarly processed.

After that, when neither host accesses the magnetic disk device, in the magnetic disk device, the input / output control unit 5 waits for the next command and all the read / write control units 8 The process is stopped when the write back process is completed.

§4: Description of processing of the fourth embodiment by a time chart--see FIG. 21 FIG. 21 is a time chart of the fourth embodiment. Below, FIG.
The process of the fourth embodiment will be described based on the time chart of No. 1. The processing content is the same as that described above.

In FIG. 21, (A) is the host 20,
(B) is the host 21, (C) is the input / output control unit 5, (D)
Shows a time chart during each processing of the data buffer control unit 6, (E) the reordering execution unit 14, and (F) a read / write control unit 8. Further, [1] to [26] are timings of each processing, and the contents thereof are as follows.

At [1], the host 20 issues the write command (e). In [2], the input / output control unit 5 receives the command. In [3], the input / output control unit 5 instructs the read / write control unit 8 to write back the write command (e). In [4], seek is performed to the target block on the disk medium. In [5], the input / output control unit 5 starts data queuing of the write command (e).

In [6], the input / output control unit 5 is transferred from the host 20.
Data transfer of the write command (e) is performed. In [7], there is no free area in the data buffer 7, and the data transfer of the write command (e) to the data buffer 7 is temporarily stopped.

At [8], the host 21 issues a write command (h). In [9], the input / output control unit 5 receives the command and stores it in the command queue 12. [1
0], the host 21 issues a write command (g). In [11], the input / output control unit 5 receives the command and stores it in the command queue 12.

At [12], the host 21 issues the write command (f). In [13], the input / output control unit 5 receives the command and stores it in the command queue 12.
In [14], the reordering execution unit 14 performs reordering, and the command execution order is (f) → (g) →
It is decided in the order of (h).

In [15], the read / write controller 8 writes the data of the write command (e) (writes back to the disk medium 9). In [16], the data buffer control unit 6 restarts the data transfer of the write command (e) to the data buffer 7. In [17], host 2
Data transfer of the write command (e) from 0 to the input / output control unit 5 (continuation) is performed.

In [18], the input / output control unit 5 causes the host 20
Report the "GOOD" status of the write command (e). In [19], the host 20 receives the “GOOD” status. In [20], the input / output control unit 5 takes out the write command (f) from the command queue 12.

At [21], the input / output control unit 5 instructs the read / write control unit 8 to write back the write command (f) subsequent to the write back of the write command (e). In [22], the input / output control unit 5
Starts the data queuing of the write command (f).

At [23], the write command (f) is transferred from the host 21 to the input / output control section 5. [2
4], the input / output control unit 5 reports the “GOOD” status of the write command (f) to the host 21. [25]
Then, the host 21 receives the "GOOD" status.

At [26], the read / write controller 8 writes the data of the write command (f) (writes back to the disk medium 9). The write commands (g) and (h) are processed in the same manner as the write command (f).

(Other Embodiments) The embodiments have been described above, but the present invention can also be implemented as follows. In the third and fourth embodiments, the number of hosts can be any number of three or more.

Not limited to the magnetic disk device, for example,
It is also applicable to a magneto-optical disk device and the like.

[0187]

As described above, the present invention has the following effects. : For sequential write, the magnetic disk device recognizes this by itself, so without waiting for disk rotation,
You can write back. Therefore, the write back process can be speeded up.

When the host performs the sequential write, the rotation waiting time of the disk medium which has been generated during the write back processing is eliminated, and the write back control can be performed at high speed and with high efficiency. Also, the access speed is improved.

[0189] Since the host receives the write commands continuously issued and queues them for command reordering, the processing time can be shortened.

[Brief description of drawings]

FIG. 1 is a diagram illustrating the principle of the present invention.

FIG. 2 is a block diagram of a magnetic disk device according to a first embodiment.

FIG. 3 is an explanatory diagram of Example 1 (an example of a write command).

FIG. 4 is a processing flowchart 1 of the first embodiment.

FIG. 5 is a processing flowchart 2 of the first embodiment.

FIG. 6 is a time chart of the first embodiment.

FIG. 7 is a block diagram of a magnetic disk device according to a second embodiment.

FIG. 8 is an explanatory diagram of reordering according to the second embodiment.

FIG. 9 is a processing flowchart 1 of the second embodiment.

FIG. 10 is a processing flowchart 2 of the second embodiment.

FIG. 11 is a time chart of the second embodiment.

FIG. 12 is a block diagram of a magnetic disk device according to a third embodiment.

FIG. 13 is a processing flowchart 1 of the third embodiment.

FIG. 14 is a processing flowchart 2 of the third embodiment.

FIG. 15 is a processing flowchart 3 of the third embodiment.

16 is a time chart of Example 3. FIG.

FIG. 17 is a block diagram of a magnetic disk device according to a fourth embodiment.

FIG. 18 is a processing flowchart 1 of the fourth embodiment.

FIG. 19 is a processing flowchart 2 of the fourth embodiment.

FIG. 20 is a processing flowchart 3 of the fourth embodiment.

FIG. 21 is a time chart of Example 4.

FIG. 22 is a block diagram of a conventional magnetic disk device.

[Explanation of symbols]

 1 magnetic disk device 2, 20, 21 host (host computer) 3 interface control unit 5 input / output control unit 6 data buffer control unit 7 data buffer 8 read / write control unit 9 disk medium 12 command queue 14 reordering execution unit

Claims (4)

[Claims] 1. A disk device provided with a write cache mechanism for storing data transferred from a host in a temporary data buffer (7) and then writing the data in the data buffer to a disk medium (9), When the command issued by the host (2) is received, it is determined whether the command is a sequential write (a continuous data block group on the disk medium is sequentially written over a plurality of commands), and the sequential write command is used. If it is determined that there is data, the data to be written to the disk medium (9) is prefetched (data queuing) to the data buffer (7), and the writing process (write back) to the disk medium specified by the new command is performed. , A feature that controls to continue the writing process performed by the previous instruction. Write-back control method for disk device. 2. A disk device comprising a write cache mechanism for storing data transferred from a host in a temporary data buffer (7) and thereafter writing the data in the data buffer to a disk medium (9), When the host (2) receives multiple write commands that are issued consecutively, it prefetches (command queuing) the commands in the order in which they were received, and then optimizes these commands in the order in which the write process ends in the shortest time. (Reordering)
Then, the optimized instructions are sequentially taken out, and it is determined whether or not the sequential write (a continuous data block group on the disk medium is sequentially written over a plurality of times of instructions) is instructed. The data buffer (7) stores the data to be written in the disk medium (9).
A disk characterized by pre-emption (data queuing) and control so that the write processing (write back) to the disk medium specified by the new instruction is performed after the write processing performed by the previous instruction. Device write back control method. 3. A disk device comprising a write cache mechanism for storing data transferred from a host in a temporary data buffer (7) and thereafter writing the data in the data buffer to a disk medium (9), When receiving an instruction issued by a plurality of hosts (20, 21), it is judged whether or not the instruction is a sequential write (a continuous data block group on the disk medium is sequentially written over a plurality of instructions). When it is determined that the command is a sequential write command, the data to be written to the disk medium (9) is prefetched (data queuing) in the data buffer (7) and the writing process to the disk medium specified by the new command ( Write back) so that it continues from the write processing performed by the previous instruction. And a write-back control method for a disk device. 4. A disk device provided with a write cache mechanism for storing data transferred from a host in a temporary data buffer (7) and then writing the data in the data buffer to a disk medium (9), When a plurality of write commands issued continuously by a plurality of hosts (20, 21) are received, the commands are prefetched (command queuing) in the order in which the commands are received, and then these commands are written in the shortest time. Optimization (reordering) is performed in order, the optimized instructions are sequentially taken out, and it is determined whether or not they are sequential write (sequential data block groups on the disk medium are sequentially written over a plurality of instructions) instructions. , If it is a sequential write command, the data to be written to the disk medium (9) (7)
A disk characterized by pre-emption (data queuing) and control so that the write processing (write back) to the disk medium specified by the new instruction is performed after the write processing performed by the previous instruction. Device write back control method.