JP2001014111A - Control method for rotary type storage device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve processing performance at the execution of a tagged command. SOLUTION: In a magnetic disk device 20 provided with a magnetic disk 40 and a magnetic disk controller 30 provided with a cache memory 33, a command queue 34 and a control processor 31, a lock-ahead interruption information table 35 is provided, a look-ahead processing is performed in rotation standby time still remaining even when the execution orders of commands in the common queue 34 are rearranged and reduce so as to be efficient execution orders at the time of executing the tagged commands of random access received from a host computer 10 and the interruption/restart of look-ahead are managed by utilizing the look-ahead interruption information table 35. Thus, the hit rate of read commands issued by the host computer 10 in the cache memory 33 is improved and the processing performance at the execution of the tagged command is improved.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a control technique for a rotary storage device, and more particularly, to the control of a rotary storage device such as a magnetic disk device having a function of rearranging commands in an efficient execution order. Regarding effective technology to apply.

[0002]

2. Description of the Related Art A SCSI (Small C) widely used as a connection interface for a small external storage device is used.
output System Interface)
, The target (magnetic disk device) is Sim
The command execution order can be changed so that a command with a tag such as ple Queue Tag has an efficient command execution order with other queued commands.

In general, in a magnetic disk device, in response to a command with a tag of random access, a moving time (seek time) of a head for reading or writing,
The command processing performance of the magnetic disk device is rearranged by rearranging the command execution order so that the sum of the rotation waiting times until reading or writing to a target sector (address at which reading or writing is performed) is minimized. Has been improved.

On the other hand, for a sequential access command, data corresponding to a host request length is read from the magnetic disk from an address at which reading of a read command issued by the host computer should be started, and the data is transferred to the host computer. At the same time, the subsequent data is read from the magnetic disk into the cache. Next, when the host computer issues a read command for the subsequent data, the data is transferred from the cache without performing disk access, thereby improving the command processing performance of the magnetic disk device.

[0005]

As described above, a read-ahead process is performed for a sequential access command, and the data is transferred from the cache when the host computer issues a read command for the subsequent data. To improve performance. That is, when executing a sequential access command, the magnetic disk device reads prefetched data from the magnetic disk to the cache until the next command is issued by the host computer.

On the other hand, for a random access command, the performance is improved by changing the command execution order so that the command execution order becomes efficient in consideration of the seek time and the rotation waiting time. I have. But,
Regardless of which command registered in the command queue is selected as the next command to be executed, the rotation waiting time becomes longer, and there is a combination of commands that cannot be rearranged in an efficient execution order. . Even if the rotation waiting time is reduced by rearranging the commands so that the execution order is efficient, the rotation waiting time does not always become zero. If the host computer does not issue a new command and there is no command to be executed in the command queue, the host computer issues no command until the next host computer issues the command. There is a problem that it becomes a state just waiting to be done.

[0007] It is an object of the present invention to improve the processing performance of a rotary storage device by performing a pre-reading process using a rotation waiting time which remains even after rearranging the execution order of a plurality of commands. .

Another object of the present invention is to improve the throughput of data read processing via a cache memory by effectively utilizing the rotation waiting time between a plurality of command execution processing for prefetch processing to a cache memory. An object of the present invention is to provide a control technology of a rotary storage device that can be used.

[0009]

According to the present invention, there is provided a rotary storage medium for storing data exchanged with a higher-level device;
Including a cache memory that temporarily holds data via a data transfer path between the rotary storage medium and the host device, the execution order of a plurality of commands queued in the command queue may be changed. A method of controlling a rotary storage device having a function of rearranging a plurality of possible commands into an execution order that minimizes the sum of a seek time and a rotation wait time for reading or writing data in the rotary storage medium. In this method, the data is prefetched from the rotary storage medium to the cache memory by utilizing the remaining rotation waiting time even after the command is rearranged.

More specifically, as an example, when a command with a random access tag is executed, even if the command execution order is reduced by rearranging the command execution order so as to obtain an efficient execution order, the rotation waiting time still remains. When a higher-level device issues a read command for an address for which a pre-read process has been performed by executing pre-read processing, a command with a tag is executed by transferring data from the cache memory without accessing the rotary storage medium. Processing performance can be improved.

[0011]

Embodiments of the present invention will be described below in detail with reference to the drawings.

FIG. 1 is a conceptual diagram showing an example of the configuration of a magnetic disk device including a magnetic disk control device for implementing a method of controlling a rotary storage device according to an embodiment of the present invention.

In FIG. 1, reference numeral 10 denotes a host computer, and 30 denotes a magnetic disk controller. The host computer 10 and the magnetic disk controller 30 are connected by a host interface 50. The host interface 50 uses a SCSI interface. 40
Denotes a magnetic disk, and is connected to the magnetic disk controller 30 via a disk interface 60. The magnetic disk control device 30 includes a control processor 31, an internal memory 32, and a cache memory 33.

The internal memory 32 has a command queue 34,
A prefetch interruption information table 35 used for controlling prefetch processing described later is provided.

As exemplified in FIG. 4, the prefetch interruption information table 35 is configured to store information of a request number 35a, an address 35b for starting prefetch, and a remaining length 35c for prefetch in association with each other. .

When the host computer 10 issues a read command A 70 to the magnetic disk device 20, the magnetic disk device 20 reads the read command A 7
0 is registered in the command queue 34. Command queue 3
4 among the commands registered in the magnetic disk 40
A command that minimizes the sum of the seek time and the rotation waiting time is selected based on the current head position, and reading or writing to the magnetic disk 40 is performed.

It is assumed that the read command A70 is selected as the next command to be executed from the commands registered in the command queue 34. In the present embodiment,
Even if the commands are rearranged so as to have an efficient execution order and the rotational waiting time is reduced, the read-ahead processing is performed on the magnetic disk 40 in order to perform the read-ahead processing for the rotational wait time remaining even after reducing the rotational wait time. Do it.
In order to prevent the cache memory 33 from being overwritten by performing the prefetching process beyond the size of the cache memory 33, the size of the cache memory 33 is set to the length for performing the prefetching process. When the cache memory 33 is divided into a plurality of segments and used, the length of the prefetch process is set to the size of one segment. The maximum value of the number of commands that can be prefetched is the number of divisions of the cache memory 33. That is, the number of request numbers 35 a (request number N) in the prefetch interruption information table 35 of FIG. 4 is determined by the number of divisions of the cache memory 33.

Since there is a queued command in the command queue 34 in addition to the currently executed read command, the prefetch processing must be interrupted. The pre-reading process is interrupted at the timing when the rotation waiting time of the next reading or writing process is minimized,
The address at which prefetching is resumed and the remaining length for prefetching are registered in the prefetch interruption information table 35 shown in FIG.

When the host computer 10 does not issue a new command and there is no command to be executed in the command queue 34, in the conventional reference technology, it becomes a waiting time until the host computer issues the next command.
In this embodiment, the prefetch processing registered in the prefetch suspension information table 35 is restarted during this waiting time. The magnetic disk 40 starts reading from the address in the prefetch interruption information table 35 where the prefetching process is restarted to the remaining length for prefetching. The pre-read data 37 read from the magnetic disk 40 is stored in the host computer 1
It is not transferred to 0 but is stored in the cache memory 33.

When the host computer 10 issues a read command B 71 to the pre-read address, the pre-read data 37 has already been read into the cache memory 33, so that the magnetic disk drive 20
Can transfer data without performing disk access.

Next, an example of a method for performing the pre-reading process will be described in detail with reference to the flow charts of the reading command process shown in FIGS. In each of FIGS. 2 and 3, indicates the connection relationship of the processing flow between the flowcharts of FIGS. 2 and 3, and S1 to S30 indicate the numbers of the respective processing steps.

If there is a command to be executed in the command queue 34 (step S01), it is determined whether or not the command is the highest priority command (read command) (step S02). The command execution order is rearranged so that an efficient execution order is obtained in consideration of the rotation waiting time and the seek time (step S03). When the commands to be executed next are read commands (step S09) as a result of rearranging them so as to have an efficient execution order, read processing for the length obtained by adding the host request length and the prefetch length is performed (step S10). The host request data 36 for the host request length is transferred to the host computer 10, and the prefetch length is stored in the cache memory 33. If there is a command to be executed other than the currently executed command in the command queue 34, the host computer 10
Command should take precedence over prefetching,
The prefetch process must be interrupted.

Here, the trigger for stopping the prefetching process will be described with reference to FIG. Even if the pre-reading process is immediately stopped to execute the command queued in the command queue 34 as shown in the upper part of FIG. 6, the rotation waiting time is long before the data of the next command is actually read. May occur.

Therefore, in the present embodiment, as exemplified in the lower part of FIG. 6, the rotation time is taken into consideration in advance by seeking the seek time to the address requested from the host computer 10 and the rotation waiting time before reading. The prefetching process is stopped so that the waiting time is minimized (step S11). As a result, the time during which the prefetch processing can be executed is extended, and the prefetch data can be read into the cache memory 33 more efficiently. This method is not very effective in cases where there are many queued commands and there is little rotation waiting before the next reading, but there are cases where the number of queued commands can not be optimally sorted and there is no tag This is effective in the case of processing the command. It is also effective when the number of sectors per track increases as the recording density increases. After stopping the prefetching process by the above method, in order to restart the prefetching process,
The address at which the prefetching process is restarted and the remaining length for performing the prefetching process are registered in the prefetch interruption information table 35 (step S12), and the command that has just been executed is deleted from the command queue 34 (step S13).

Although the command queue 34 is not empty (step S01), it may be a write command such that a status is specified when all data transferred by a read command or a write command is written to the disk. A command that must be processed with priority over other commands is not registered in the command queue 34, and the status may be reported when all data transferred by the write command is read into the cache memory 33. If only the specified write command is registered in the command queue 34 (step S02), the commands are arranged in an efficient execution order, including the prefetching process having a higher priority among the prefetching processes. (Step S04). Here, among the pre-reading processes, the pre-reading process having a higher priority order includes the pre-reading process designated by the host computer 10 and the
In the future, when the host computer 10 issues a read command, it means a pre-read process of data determined by the magnetic disk device 20.

A case in which the host computer 10 instructs prefetch will be described. When the host computer 10 issues a read command, if the host computer 10 determines that the subsequent data is needed later,
B (Command Descriptor Bloc)
k) Alternatively, a command is issued to the magnetic disk device 20 with the prefetch request flag provided in the message made valid.

A case where the magnetic disk device 20 determines that the pre-read process is necessary will be described. For example, in the case where commands to be read into a plurality of continuous areas registered in the command queue 34 are processed as one command by rearranging the execution order, it is necessary to determine that the prefetch processing is necessary. As shown in Example 1 of FIG. 5, the read command A from the address L to the address (M-1) and the read command A from the address M to the address (N-
When the read commands B up to 1) are present in the command queue 34, the two commands are changed in order of command execution in order to shorten the processing time, and are read as the read commands C from address L to address (N-1). It is executed collectively. When the read commands for two consecutive areas are collectively executed as one command as described above, it is necessary to recognize that the prefetch process is necessary after the address N, and to perform the prefetch process. Even if the two read commands A and B are not continuous as in Example 2 in FIG. 5, even if the α sector which is the difference between the addresses of the read commands A and B is within a certain threshold number of sectors. As in the case of the first example, it recognizes that the prefetching process after the address N is necessary, and performs the prefetching process.

Even if the write commands are rearranged so as to have an efficient execution order and the rotation waiting time is reduced, the above-described high-priority prefetching process is performed during the remaining rotation waiting time (step S05). In order to perform the next writing process, the prefetching process is stopped in consideration of the seek time and the rotation wait (step S06). After the prefetching process is stopped, the address at which the prefetching process was stopped and the remaining prefetching length are stored in the prefetching suspension information table 35 in order to restart the prefetching process.
Then, it is registered as the address 35b for starting prefetching and the remaining length 35c for prefetching (step S07), and writes the write data read in the cache memory 33 to the disk (step S08).

Also, a write-and-verify command or an XOR write command (for reading data at a specific address on the magnetic disk 40,
(A command to write the result of performing an exclusive OR operation with the write data to the address given to the original address and returning the result to the original address) in a single command (eg, a command including a plurality of read and write processes). S1
The pre-read process 4) will be described below using a write-and-verify command as an example. The first write process is completed (step S17), and the prefetch process registered in the prefetch suspension information table 35 is restarted during the rotation waiting time until the next verify process is performed (step S1).
8). In order to execute the verify processing, the prefetch processing must be stopped in consideration of the seek time and the rotation waiting time (step S19). After the prefetching process is stopped, the address at which the prefetching process was stopped and the remaining prefetching length are registered in the prefetching suspension information table 35 for resuming the prefetching (step S20), a verifying process is performed (step S21), and the prefetching is suspended. The prefetching process registered in the information table 35 is restarted (step S22).

When executing a write command in which the status is to be reported when all data transferred by the write command is written to the disk, write processing for the host request length is performed (step S15). The prefetch processing registered in the prefetch suspension information table 35 is restarted (step S16).

If the host computer 10 does not issue a new command, executes all the commands registered in the command queue 34, and there is no command to be executed in the command queue 34, the prefetch interruption information table 35
Restarts the pre-reading process registered in (step S2)
3). From the address at which the prefetching process is restarted, data of the remaining length for prefetching is read from the magnetic disk 40 to the cache memory 33. When a command is issued from the host computer 10, the command from the host computer 10 must be given priority over the prefetching process. When the command is issued from the host computer 10 (step S24), the prefetching process is stopped in consideration of the seek time and the rotation waiting time for executing the next command as described above (step S25). When the prefetching process is stopped, the address at which the prefetching process is restarted and the remaining length are registered in the prefetching interruption information table 35 (step S26).

The prefetch processing is continued for the length registered in the prefetch suspension information table 35 (step S2).
7, when the prefetching process is completed (step S27), the prefetching process that has just been completed is deleted from the prefetching interruption information table 35 (step S28). If the prefetch processing that must be restarted is registered in the prefetch suspension information table 35 (step S29), the next prefetch processing is restarted.

As described above, in the magnetic disk device 20 according to the present embodiment, when executing a command with a random access tag, the command execution order is reduced by rearranging the command execution order so that the command execution order becomes efficient. However, by performing a pre-read process for transferring the data on the magnetic disk 40 to the cache memory 33 during the remaining rotation waiting time, when the host computer 10 issues a read command for the pre-read address, By transferring data from the cache memory 33 without accessing the disk 40, the processing performance of executing the tagged command can be improved.

That is, when a command with a random access tag is executed, even if the command execution order is reduced by rearranging the command execution order so that an efficient execution order is obtained, the remaining rotation waiting time is effectively used. By pre-reading the data into the cache memory 33, the probability that the data requested to be read from the host computer 10 exists in the cache memory 33 (hit rate) is improved, and the data transfer performance from the magnetic disk 40 to the host computer 10 is improved. Can be improved.

Although the invention made by the present inventor has been specifically described based on the embodiments, the invention is not limited to the above-described embodiments, and can be variously modified without departing from the gist thereof. Needless to say, there is.

For example, in the description of the above embodiment,
Although a case where the present invention is applied to a magnetic disk device as an example of a rotary storage device has been described as an example, the present invention can be widely applied to control of a general rotary storage device.

[0037]

According to the control method of the rotary storage device of the present invention, the read-ahead processing is performed by utilizing the rotation waiting time which remains even after the execution order of the plurality of commands is rearranged. The effect that the processing performance of the apparatus can be improved can be obtained.

According to the control method of the rotary storage device of the present invention, the rotation waiting time between the plurality of command execution processes is effectively used for the prefetch process to the cache memory, and the data read process via the cache memory is performed. The effect that the throughput can be improved is obtained.

[Brief description of the drawings]

FIG. 1 is a conceptual diagram showing an example of a configuration of a magnetic disk device including a magnetic disk control device that executes a method of controlling a rotary storage device according to an embodiment of the present invention.

FIG. 2 is a flowchart showing an example of an operation of a control method for a rotary storage device according to an embodiment of the present invention, together with FIG. 3;

FIG. 3 is a flowchart showing an example of an operation of a control method for a rotary storage device according to an embodiment of the present invention, together with FIG. 2;

FIG. 4 is a conceptual diagram showing an example of control information used in a method for controlling a rotary storage device according to an embodiment of the present invention.

FIG. 5 is a conceptual diagram illustrating an example of an operation of a method of controlling a rotary storage device according to an embodiment of the present invention.

FIG. 6 is a conceptual diagram showing an example of an operation of a method of controlling a rotary storage device according to an embodiment of the present invention, in comparison with a reference technique;

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 10 ... Host computer (host device), 20 ... Magnetic disk device (rotary storage device), 30 ... Magnetic disk controller, 31 ... Control processor, 32 ... Internal memory, 3
3 ... cache memory, 34 ... command queue, 35 ...
Prefetch interruption information table, 35a... Request number, 35b.
Address at which prefetching is resumed, 35c: remaining length for prefetching, 36: host request data, 37: data read ahead, 40: magnetic disk (rotary storage medium), 50 ...
Host interface, 60: Disk interface, 70: Read command A, 71: Read command B

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Yuji Yamane 2880 Kozu, Kodawara-shi, Kanagawa Pref.Hitachi, Ltd.Storage Systems Division, Hitachi, Ltd. (72) Inventor Tsuneo Hirose 2880 Kozu, Kozu, Hitachi, Ltd. Within the System Division (72) Inventor Keiichiro Hirata 6-81 Onoecho, Naka-ku, Yokohama-shi, Kanagawa Prefecture In-house Hitachi Software Engineering Co., Ltd. (72) Inventor Koji Horiuchi 6-81 Onoecho, Naka-ku, Yokohama-shi, Kanagawa Prefecture Hitachi Software Engineering Stock Association Internal F-term (Reference) 5B065 BA01 CA03 CA15 CC08 CH05 5D044 AB02 BC01 CC04 DE03 DE12 EF03 HL01 5D066 AA02 AA03

Claims (3)

[Claims] A rotary storage medium for storing data exchanged with a host device; and a data transfer path between the rotary storage medium and the host device via a data transfer path. Reads or writes the data in the rotary storage medium from among a plurality of commands queued in a command queue, including a cache memory that is temporarily held, and a plurality of commands whose execution order can be changed among the plurality of commands queued in a command queue. A method of controlling a rotary storage device having a function of rearranging the execution order so that the sum of the seek time and the rotation waiting time for the rotation is minimized. A prefetching process of the data from the rotary storage medium to the cache memory using the storage device. Your way. 2. A command for controlling a rotary storage device according to claim 1, wherein a command which must be processed with priority over other commands, such as a read command, is not registered in the command queue. When all the data transferred by the write command are read into the cache memory, only the write command that specifies that the write completion status may be reported to the higher-level device is registered. The pre-reading process specified by the higher-level device or the pre-reading process determined to be necessary by the rotary storage device is performed during the rotation waiting time before writing, which remains even after being reduced by the rearrangement. A method for controlling a rotary storage device. 3. The method of controlling a rotary storage device according to claim 1, wherein when executing a command involving a plurality of read and write processes with a single command, the first read or write process is completed and the next read or write process is completed. Alternatively, the read-ahead process is performed during a rotation waiting time until a write process is performed.