JPH0477915A - Magnetic disk controller - Google Patents

Abstract

PURPOSE:To improve the throughput of a whole computer system by providing a command storage means storing plural write/read request commands from CPU and saving execution when there are overlapped commands. CONSTITUTION:When the command storage means 131 storing plural write/read request commands transferred from CPU 11 intends the timing buffer of the processing of CPU 11 and the processing of a magnetic disk device 14 and when the write/read request commands for the same storage area are overlapped and stored, the execution of several commands which are overlapped is omitted. Thus, the number of the physical input/output operations of the magnetic disk device 14 can effectively be suppressed and the throughput of the whole computer system can be improved.

Description

[Detailed Description of the Invention] [Object of the Invention] (Industrial Application Field) The present invention relates to a magnetic disk control device, and particularly relates to a magnetic disk control device for suppressing the number of physical input/output operations to a magnetic disk device. .

(Prior Art) In general, input/output operations, that is, data write/read operations for a hard disk drive (HDD) are controlled by a hard disk control bag [(
HDC).

That is, the hard disk control device interprets the data read/write request command transferred from the CPU, specifies the data read/write start position and its length to the hard disk drive, and thereby controls the hard disk drive. Control output. In this case, the operation of specifying the data reading/writing start position and its length is executed for each command, so for example, C
Upon receiving ten read/write request commands from the PU, the hard disk control device performs ten input/output controls on the hard disk drive device.

Actual input/output operations to the magnetic disk in a hard disk drive device include seek operations and magnetic disk rotation operations to position the head in the section (page) indicated by the specified read/write start position and its length. I need. Mechanical operations such as seek operations and magnetic disk rotation operations require a much longer time than electrical processing times for inputting and outputting data in an IC memory or the like. As described above, physical input/output operations of the hard disk drive device take time, so if the hard disk control device controls the input/output of the hard disk drive device for each command, the throughput of the entire computer system will decrease. .

Therefore, in order to prevent the throughput from decreasing,
By providing a cache memory consisting of an IC memory inside the hard disk control device and registering page information read from the hard disk drive device in the cache memory, the hard disk drive device can be used in response to subsequent read requests for the same page. Many techniques are used to replace this with reading from the cache memory without any control. In addition, a store-in method has begun to be used in which not only read requests but also write requests are completed by writing to the cache memory, and the hard disk drive device is controlled for writing only when there is no free storage space in the cache memory.

However, although this method of using cache memory is more effective as the storage capacity of the cache memory increases, in reality there is a limit to its storage capacity, so even if cache memory is used, throughput cannot be sufficiently increased. It is difficult to do so. In addition, although static RAM, which has a fast operating speed, is normally used as cache memory, static RAM is volatile memory, so if data is written to cache memory using the store-in method, even if the battery is Even if a backup mechanism is provided, there is a risk of data loss due to various failures.

(Problems to be Solved by the Invention) Conventionally, physical input/output operations in a hard disk drive device take time, which has the disadvantage of causing a reduction in the throughput of the entire computer system.

This invention was made in view of these points,
It is an object of the present invention to provide a magnetic disk control device that can suppress the number of physical input/output operations of a magnetic disk device and improve the throughput of the entire computer system.

[Structure of the Invention (Means and Effects for Solving the Problems)] A magnetic disk control device according to the present invention includes a command storage means for storing a plurality of write/read request commands from a CPU to a magnetic disk device, and detecting means for detecting whether write/read request commands for the same storage area of the device are stored redundantly in the command storage means; Prohibits execution of read request commands following the first stored read request command, and when write request commands are stored redundantly, writes preceding the write request command stored last in the command storage means. and means for prohibiting execution of a request command, thereby suppressing the number of physical input/output operations of the magnetic disk device.

In this magnetic disk control device, timing buffering between CPU processing and magnetic disk device processing is achieved by a command storage means that stores a plurality of write/read request commands transferred from the CPU, and moreover, there are duplicate commands. In some cases, execution of some of the duplicate commands is omitted. Therefore, the number of physical input/output operations of the magnetic disk device can be effectively suppressed, and the throughput of the entire computer system can be improved.

(Example) Hereinafter, the present invention will be described in detail with reference to the drawings.

FIG. 1 shows a magnetic disk control device according to an embodiment of the present invention. This magnetic disk control device 13
is a hard disk controller that controls input/output operations, that is, data write/read operations of a plurality of hard disk drive devices (HDD) 14 according to instructions from the CPU II, and includes a stack memory 131, an execution command management section 132, and an input/output control section. It is equipped with 138.

The stack memory 131 stores multiple write request commands and read request commands transferred from the CPU II.
This stack memory 131 is for temporary storage, and is configured with a FIFO (first-in first-out) buffer.

The execution command management unit 132 selects a command whose execution can be omitted from among the write/read request commands stacked in the stack memory 131, and prohibits execution of the selected command. A command that specifies the same storage area as an access target is selected as a command that can be omitted to be executed.

That is, when read request commands specifying the same storage area as an access target are stored redundantly in the stack memory 131, the execution command management unit 13
2, the read request command first stored in the stack memory 131 is treated as a valid command and its execution is permitted, and the read request command that follows the valid command is treated as an invalid command and its execution is prohibited. Furthermore, when write request commands specifying the same storage area as an access target are stored redundantly in the stack memory 131, the execution command management unit 1
32 makes the write request command stored last in the stack memory 131 a valid command and permits its execution, and treats the write request command preceding the valid command as an invalid command and prohibits its execution.

The input/output control section 133 controls the input/output of the hard disk drive device 14 according to instructions of valid commands selected by the execution command management section 132.

Next, the selection process of valid commands and invalid commands executed by the execution command management unit 132 in the hard disk controller 13 will be described with reference to the flowchart in FIG.

First, m write/read request commands transferred from the CPU II are sequentially stored in the stack memory 181. At this time, the entry address of the first write/read request command stored in the stack memory 131 is set as the parameter m for specifying the entry address of the collation command, and the entry address of the first write/read request command stored in the stack memory 131 is set as the parameter "1" is set as the parameter n (step Al). For example, when the number of commands that can be temporarily stored in the stack memory 131 is 10, the entry address of the first command stored at the first position of the stack memory 131 is "10".

Next, the command specified by the parameter m and (m
-n), that is, the first command whose entry address is "10", and the contents of the subsequent command whose entry address is rlO-1-9J are compared (step A2). Then, it is determined whether the page specified by the first command and each of the first subsequent commands, that is, the hard disk drive device specified as the access target and the storage area on the disk are common. (Stayab A3
).

If the first command and the first subsequent command do not specify the same storage area on the disk that is input/output by the same hard disk drive device, the parameter n is updated by +1 (step A4). Then, it is determined whether or not the values of parameters m and n are equal (step A5).
The processes of steps A2 to A4 are repeatedly executed until the values of the first command and all the commands subsequent to the first command are compared with each other.

On the other hand, if the first command and the first command that follows it specify the same storage area on the disk that is input/output by the same hard disk drive device, then the first command and the first command It is determined whether the th subsequent commands are both read request commands or write request commands (Step A8. Step A8).

If both the first command and the first subsequent command are read request commands, the first command is selected as a valid command, and the first subsequent command is selected as an invalid address (step A7). Then, the process moves to step A4, and this time the first command is compared with the second command following it.

On the other hand, if both the first command and the first subsequent command are write request commands, the command is set (step A9). Then, in order to use the first command following the first command as the verification command instead of the first command, (m-n) is set as the parameter m (step A10), and after this, in the process of step A4. Then, a comparison process is performed using the first command following the first command as the collation command.

In this way, even if the first command and the first subsequent command are both write request commands or read request commands, the comparison process is performed with the command selected as the valid command as the new verification command. be exposed. Then, by sequentially updating the collation commands,
A process of collating all commands in the stack memory 13i is performed.

In this way, the hard disk controller 1 of this embodiment
In step 3, it is detected whether commands specifying the same page as an access target are stored in duplicate in the stack memory 131, and if the duplicate command is a read request command, the stack memory 131! Only the first read request command stored in $1 is executed, and the execution of subsequent read request commands is omitted. Also, if the duplicate command is a write request command,
Only the write request command stored last in the stack memory 131 is executed, and the execution of the preceding write request commands is omitted.

Therefore, the number of times the input/output of the hard disk drive M14 is controlled can be reduced, and the throughput of the entire computer system can be improved.

[Effects of the Invention] As described above, according to the present invention, it is possible to suppress the number of physical input/output operations to a magnetic disk device, and it is possible to improve the throughput of the entire computer system.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of a disk control device according to an embodiment of the present invention, and FIG. 2 is a flowchart explaining the operation of the disk control device shown in FIG. 1. 11...CPU, 12...System bus, 13...
-Hard disk controller, 14...hard disk drive device. Applicant's agent Patent attorney Takehiko Suzue

Claims (1)

[Scope of Claims] Command storage means for storing a plurality of write/read request commands from a CPU to a magnetic disk device; and a command storage means for storing a plurality of write/read request commands for the same storage area of the magnetic disk device in the command storage means. a detecting means for detecting whether or not the read request command is stored in the command storage means; and a detecting means for detecting whether or not the read request command is stored in the command storage means, and prohibiting execution of a read request command subsequent to the read request command first stored in the command storage means when the read request command is stored redundantly. and means for prohibiting execution of a write request command that precedes the write request command last stored in the command storage means when write request commands are stored redundantly; A magnetic disk control device characterized by suppressing the number of input/output operations.