JP2000056932A - Disk controller - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a disk controller which satisfies the requested delay time according to the deterioration of performance of a disk device and also executes a request with high efficiency. SOLUTION: A request management part 13 holds the requests as a queue, and a delay time management part 12 holds the end time when the execution should be ended for each request included in the queue. An access time management part 15 holds the access time that is acquired by giving a request to a hard disk 16. A rearrangement control part 11 rearranges the requests every time a request is inputted so as to shorten the seek distance and to end the execution of a request before its end time. A disk IO part 14 measures the execution time of each request and updates the access time. Thus, it's possible to execute a request before its end time according to the deterioration of performance of the disk 16 due to its aging and also to secure the real time nature of animation and voice data.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a disk control device used in a media server which needs to process requests generated at a constant speed in real time, and more specifically to randomly generated requests. The present invention relates to a disk controller in which a seek time is reduced to improve a request processing capability.

[0002]

2. Description of the Related Art In recent years, a disk device for storing multimedia data including audio data and video data has been provided, and data has been supplied to a plurality of users connected via a communication line or the data has been transferred to a disk device. A media server that writes data to the Internet has been put to practical use. The user can watch a desired video when he / she wants to watch, and can hear a desired voice when he / she wants to listen, using the data supplied from the media server. Similarly, the user can record video and audio on the media server by supplying data to the media server.

This media server needs to output data to a plurality of users in real time. This is because if the output timing of data exceeds a predetermined allowable delay time, there is a problem that the sound is interrupted or the video is disturbed (missing). Similarly, media servers need to record data from multiple users in real time. This is because when audio and video data is converted from a terminal to digital data in real time and stored in a media server,
This is because if the completion of the process of storing in the media server exceeds a predetermined allowable delay time, a problem occurs in that the memory of the terminal or the media server overflows.

To cope with this problem, a disk controller for reading data from a disk in real time in a media server has been disclosed in Japanese Patent Laid-Open No. 8-5 / 1996.
There is one disclosed in Japanese Patent No. 5055 (hereinafter referred to as a conventional document). Hereinafter, a disk control device described in this conventional document will be briefly described with reference to FIG.

FIG. 17 is a block diagram showing a configuration of a disk control device described in this conventional document. 17, the conventional disk control device includes a rearrangement control unit 301, an allowable delay data management unit 302, a request management unit 303, and a hard disk 306.

[0006] The hard disk 306 divides and stores multimedia data representing a plurality of types of video and audio for each fixed size. The allowable delay data management unit 302
For each read request, a delay time allowed until reading of data specified by the request is completed is stored. The request management unit 303 holds each read request.
For each read request, the rearrangement control unit 301 calculates and determines the time until the read is completed.
The read requests are rearranged so that the reading is completed within the allowable delay time held by the allowable delay data management unit 302 and the seek distance on the hard disk 306 is shortened.

[0007] The time until the reading is completed is calculated by the sum of the waiting time in the queue and the execution time of the request. The waiting time in the queue is calculated as the sum of the execution times of the requests stored in the queue before the request for calculating the waiting time. The execution time of the request is the time required for seeking calculated from the cylinder position of the data accessed by the request, the average rotation waiting time calculated from the rotation speed of the hard disk, the number of blocks per cylinder, and the rotation speed of the hard disk. And the data transfer time calculated from the above. As described above, the execution time of the read request is calculated from a value described in a product manual or the like. As described above, the conventional disk control device improves the real-time processing by rearranging the read requests based on the completion time of the requests.

[0008]

However, in the disk control device described in the above-mentioned conventional document, a seek time, an average rotation waiting time, and a data transfer time are calculated, and the sum of these is calculated as a read request execution time. When the values described in the product manual and the like are the same, the calculated execution times are equal. Therefore, the calculated execution time does not reflect the difference in the execution time based on the number of defective sectors on the disk surface or the difference in the operation of the control IC in the disk. Also, the effect of disk performance degradation that occurs when the same disk is used for a long period of time was not reflected in the execution time calculation.

Thus, in the conventional disk control device, the requests are rearranged so that the requests can be executed within the allowable delay time from the calculated execution time. However, when the requests are actually executed, the request is executed. Completion exceeded the allowable delay time.

[0010] Therefore, an object of the present invention is to reduce the seek time by rearranging the order of a plurality of requests input continuously and to execute the requests within an allowable delay time. The goal is to provide a disk controller that is complete.

[0011]

According to a first aspect of the present invention, the requests are rearranged and issued to the disk device so that the execution of each request is completed within the delay time individually provided by each request. A disk control device, comprising: a request management means for holding one or a plurality of requests; a disk IO for taking out the request held by the request management means, issuing the request to the disk device, and receiving a response to the request from the disk device Means, access time management means for holding the time required for processing of each request in the disk device as access time, delay time management means for holding, for each request, a completion time at which the processing of the request should be completed; A rearrangement control unit that rearranges the requests held by the request management unit based on the time and the access time.

As described above, according to the first aspect, each time a request is input, the execution time of the request is predicted based on the access time held by the access time management means, and the delay specified in each request is estimated. Reorder requests so that they complete execution within the available time. Since the prediction of the execution time of the request only refers to the stored access time, no special operation is required to obtain the predicted value at the time of rearrangement. Thus, the requests are rearranged each time a request is input, so that access efficiency is improved and the requests can be executed within the delay time. In particular, when accessing data that requires real-time properties such as moving pictures and audio data, it is possible to improve access efficiency and secure real-time properties.

A second invention is characterized in that, in the first invention, the access time is determined based on a time from issuing a request to the disk device to receiving a response to the request from the disk device.

As described above, according to the second aspect, the first aspect
According to the invention, the access time is obtained from a value obtained by measuring a time required for issuing a request to the disk device and receiving a response to the request, so that a value corresponding to the performance of the connected disk device can be obtained.

In a third aspect based on the second aspect, the access time divides the area on the disk device accessible to the request into a plurality of groups, which are continuous areas, and the request accesses each request. The group to which the position on the disk device belongs is obtained, and the group is determined by the order of the group corresponding to the execution order of each request.

As described above, according to the third aspect, the second aspect
In the invention, the access time can be determined by the order of the group to which the position accessed by each request belongs.

In a fourth aspect based on the third aspect, the access time is individually determined for each type of request.

As described above, according to the fourth aspect, the third aspect
In the invention, the access time can also be determined according to the type of each request, and more accurate prediction can be made for each type of request.

In a fifth aspect based on the third and fourth aspects, the access time is individually determined for each data size accessed by the request.

As described above, according to the fifth aspect, the third aspect
In the fourth and fourth aspects of the present invention, the data size can be determined by the data size accessed by the request, and more accurate prediction can be performed for each data size.

According to a sixth aspect based on the fifth aspect, the data size is a fixed length.

As described above, according to the sixth aspect, the fifth aspect
According to the invention, the amount of memory required to hold the access time can be reduced.

According to a seventh aspect based on the third aspect, the area on the disk device is composed of a plurality of fixed-length blocks, and each group has an equal number of blocks. .

According to an eighth aspect based on the third aspect, the disk device is a hard disk composed of a plurality of cylinders, and each group has an equal number of cylinders.

As described above, according to the seventh and eighth aspects, in the third aspect, the fluctuation of the execution time for each group is reduced, and a more accurate predicted value can be obtained.

According to a ninth aspect, in the third to eighth aspects, the order of the groups is a permutation or combination of two groups.

As described above, according to the ninth aspect, the third aspect
In the eighth to eighth aspects, the execution time can be obtained by a permutation or combination of two groups, and prediction can be performed with a smaller amount of memory without changing the accuracy of the predicted value.

In a tenth aspect based on the fourth aspect, the request types are a read request for reading data from the disk device and a write request for writing data to the disk device.

As described above, according to the tenth aspect of the present invention, in the fourth aspect of the present invention, the request for rearranging is limited to a read request and a write request. Sorting can be performed while keeping time.

According to an eleventh aspect based on the first aspect, the completion time is a time obtained by adding a delay time which the request has to the time when the request is inputted, and the rearrangement control means comprises:
So that the execution of each request completes before its completion time,
It is characterized in that each request is rearranged.

As described above, according to the eleventh aspect, in the first aspect, it is not necessary to constantly update the completion time of each request, and the amount of calculation can be reduced.

In a twelfth aspect based on the first aspect, the delay time management means has a predetermined constant, and the completion time is a time obtained by adding the constant to the time at which the request was input; The rearrangement control means rearranges the requests so that the execution of each request is completed before each completion time.

As described above, according to the twelfth aspect, in the first aspect, it is not necessary to add a delayable time to each request, and the calculation of the completion time is simplified.

In a thirteenth aspect based on the third to sixth aspects, the disk IO means measures an execution time from when the request is issued to the disk device to when a response to the request is received from the disk device. An execution time measurement unit is further provided, and the access time management unit is further provided with an access time calculation unit for updating the execution time as a new access time.

As described above, according to the thirteenth aspect, in the third to sixth aspects, the access time is updated based on the measured execution time while measuring the execution time of the request. The access time corresponding to the performance degradation of the disk can be obtained.

According to a fourteenth aspect, in the thirteenth aspect,
The access time calculation means calculates the execution time in the order of the group,
It is characterized in that the request is classified for each type and data size, and the access time corresponding to the classification is updated.

As described above, according to the fourteenth aspect, in the thirteenth aspect, the access time is updated for the group order, the type of request, and the data size.
It is possible to obtain an access time corresponding to the performance deterioration of the disk with the passage of time.

In a fifteenth aspect based on the thirteenth and fourteenth aspects, the access time management means updates the access time for each of the classifications each time the execution of a certain number of requests is completed. I do.

In a sixteenth aspect based on the thirteenth and fourteenth aspects, the access time management means updates the access time of each of the classified classes each time the execution of the fixed number of requests is completed. Is performed.

As described above, according to the fifteenth and sixteenth inventions, in the thirteenth and fourteenth inventions, the access time is always updated for every fixed number of requests. The access time corresponding to the performance degradation of the disk can be obtained.

In a seventeenth aspect based on the thirteenth to sixteenth aspects, the access time is obtained by an average value of the classified execution times.

As described above, according to the seventeenth aspect, in the thirteenth to sixteenth aspects, it is possible to suppress the influence on the required access time in the case where there is a request whose execution time is increased by accident. It becomes possible.

In an eighteenth aspect based on the thirteenth to seventeenth aspects, the disk IO means further comprises a command generating means for generating a request to be issued to the disk device, and the access time calculating means comprises an initial value of the access time. Is obtained.

As described above, according to the eighteenth aspect, in the thirteenth to seventeenth aspects, it is possible to obtain the initial value of the access time when using the connected disk control device.

In a nineteenth aspect based on the first to eighteenth aspects, the access time is stored by a write command.
A storage device that outputs the read command; and a command issuing unit that issues the write command and the read command to the storage device.The access time management unit generates the write command and stores the access time in the storage device. The method is characterized in that a read command is generated and an access time is obtained from a storage device.

According to a twentieth aspect, in the nineteenth aspect,
The storage device is configured as a disk device, and the command issuing unit is configured as a disk IO unit.

As described above, according to the nineteenth and twentieth aspects of the present invention, in the first to nineteenth aspects, the access time is stored in the storage device and read out at the time of use to set the initial value of the access time. The desired operation can be simplified.

In a twenty-first aspect based on the first aspect, the request management means includes, for each request, a flag indicating that a direction in which the disk device seeks changes, and a request in the queue at the time the request is input. A first time obtained by adding a time to wait for completion of another request and an access time of the request is held. The request input to the rearrangement control means is regarded as a first request, and the request management means Last request with the flag set, or disk I if the flag is not set.
When the last request extracted from the request management means by the O means is the second request, the rearrangement control means determines that the first request is a logical address accessed by the first request. Temporarily inserting the logical addresses accessed from the last request to the second request in a position that does not disturb the order,
When the order of the logical addresses is changed by the temporary insertion, the flag of the first request is set, and after the temporary insertion, the first time is calculated again for each request after the first request in the request management unit. If the recalculated first time does not exceed the completion time, the provisionally inserted position is set as the insertion position of the first request. If the recalculated first time exceeds the completion time, the request management means is used. Is the insertion position of the first request.

As described above, according to the twenty-first aspect, in the first aspect, the requests are arranged in the order of the logical addresses to be accessed so that the execution of the requests is completed within the delay time of each request. By changing, the access efficiency can be improved.

According to a twenty-second invention, in the twenty-first invention,
When a request for accessing the same data as the first request is set as a third request, the rearrangement control unit sets the first request and the third request
Is a read request, and the first request and the third request
If there is no other request to write the data accessed by the first or third request between the first request and the third request, the access time of the first request is set to 0, and the disk IO means executes the third request. The result is an execution result of the first request.

As described above, according to the twenty-second aspect, in the twenty-first aspect, a read request for accessing the same data can be efficiently executed.

According to a twenty-third aspect, in the twenty-first aspect,
When a request to write data at the same position as the first request is a third request, the rearrangement control unit determines that the first request is a write request and the first request and the third request If there is no other request to read data accessed by the first or third request in the meantime, the first request is inserted immediately after the third request, and the access time of the third request is set to 0. The disk IO unit does not issue the third request, and sets the execution result of the first request as the execution result of the third request.

As described above, according to the twenty-third aspect, in the twenty-first aspect, a write request for accessing the same data can be efficiently executed.

According to a twenty-fourth aspect, in the twenty-first aspect,
When a request for reading data at a position that is continuous with the data to be read in the first request is a third request, the rearrangement control unit determines that the first request is between the first request and the third request. If there is no other request to write to the data to be accessed, the first request is inserted immediately before or after the third request according to the sequence of consecutive positions, and the access time of the first request or the third request is reduced. The access time is determined according to the data size read by the first request and the third request, and the disk IO unit combines the first and third requests and sets the disk IO as one request for reading data at consecutive positions. It is characterized by being issued to a device.

As described above, according to the twenty-fourth aspect, in the twenty-first aspect, a plurality of read requests for accessing continuous data can be efficiently accessed.

According to a twenty-fifth aspect, in the twenty-first aspect,
When a request to write data at a position that is continuous with the data to be written by the first request is a third request, the rearrangement control unit determines that the first request is between the first request and the third request. If there is no other request to read in the data to be accessed, the first request is inserted immediately before or immediately after the third request in the order of the consecutive positions, and the access time of the first request or the third request is reduced. The access time is determined according to the data size read by the first request and the third request, and the disk IO unit combines the first and third requests and sets the disk IO as one request for reading data at consecutive positions. It is characterized by being issued to a device.

As described above, according to the twenty-fifth aspect, in the twenty-first aspect, a plurality of write requests for accessing continuous data can be efficiently accessed.

A twenty-sixth invention is a disk control apparatus for rearranging requests and issuing them to a disk device such that the execution of each request is completed within a delay time individually provided by each request. A request management means for holding a plurality of requests, a disk IO means for taking out the requests held by the request management means and issuing the requests to the disk device, and receiving a response to the requests from the disk device; A delay time management unit for obtaining the number of insertable requests indicating the number of requests that may be executed earlier in addition to the number of requests already held in the request management unit when the request is received; A reordering control unit for reordering the requests held by the request management unit based on the number.

As described above, according to the twenty-sixth aspect, every time a request is input, the number of insertable requests is calculated from the delayable time held by the request, and more requests than the number of insertable requests are determined. Reorder the requests so that they are not inserted before the request. As a result, access efficiency is improved, and execution of the request can be completed within the delay time.

According to a twenty-seventh aspect, in the twenty-sixth aspect,
The delay time management means includes: a reference execution time holding means for holding a reference execution time, which is a time from when a request is issued to the disk device to when a response to the request is received from the disk device; And an insertable request number calculating means for obtaining an insertable request number from the delay possible time designated in (1).

As described above, according to the twenty-seventh aspect, in the twenty-sixth aspect, the reference execution time, which is a value obtained by measuring the execution time of the request issued to the disk device, is measured. By obtaining the number of insertable requests from the delay possible time of, the more accurate number of insertable requests can be obtained.

According to a twenty-eighth aspect, in the twenty-seventh aspect,
The disk IO means includes a command generating means for generating a request to be issued to the disk device, an execution time measuring means for measuring an execution time from when the request is issued to the disk device to when a response to the request is received from the disk device. And a reference execution time calculating means for obtaining an initial value of the reference execution time from the execution time measured by the execution time measuring means.

As described above, according to the twenty-eighth aspect, in the twenty-seventh aspect, a request to be issued to a disk device connected to the disk control device is generated, and the execution time of the request is measured. The initial value of the reference execution time for the connected disk device can be obtained.

According to a twenty-ninth aspect, in the twenty-seventh aspect,
The disk IO means comprises: an execution time measuring means for measuring an execution time from when a request is issued to the disk device to when a response to the request is received from the disk device; and a new reference based on the execution time measured by the execution time measuring means. And a reference execution time calculating means for obtaining the execution time and updating the reference execution time held by the reference execution time holding means.

As described above, according to the twenty-ninth aspect, in the twenty-seventh aspect, by measuring the execution time of the request and updating the reference execution time, the insertion corresponding to the deterioration of the performance of the disk device due to aging can be achieved. The number of possible requests can be obtained.

According to a thirtieth aspect, in the twenty-seventh to twenty-ninth aspects, the reference execution time is obtained by an average value of the measured execution times.

As described above, according to the thirtieth aspect, in the twenty-seventh to twenty-ninth aspects, there is a request in which the execution time of the request is increased accidentally by taking an average value of a plurality of values. In this case, the influence on the reference execution time can be suppressed.

According to a thirty-first aspect, in the twenty-sixth aspect,
The request management means holds, for each request, a flag indicating that the direction of seeking the disk device is changed or the number of insertable requests is 0, and the number of insertable requests. Is the first request, the last request with the flag set in the request management means, or the disk IO if the flag is not set.
Means that the last request retrieved from the request management means
In this case, the rearrangement control means determines that the first request is a logical address accessed by the first request, and that each logical address accessed by the last request to the second request in the request management means. If the order of the logical addresses is changed by the insertion, the flag of the first request is set, and the number of requests existing before the insertion position of the first request is set to the first number. Is subtracted from the number of insertable requests of the first request, and 1 is subtracted from the number of insertable requests of all requests after the insertion position of the first request. , A flag is set.

As described above, according to the thirty-first aspect, in the twenty-sixth aspect, the execution of the requests is completed within the delayable time of each request, and in the order of the logical addresses accessed by each request. By rearranging requests, access efficiency can be improved.

According to a thirty-second aspect, in the first to thirty-first aspects, when the disk device has a cache function, the disk IO means issues a request to the disk device without using the cache function. , And a response to the request.

As described above, according to the thirty-second aspect, in the first to thirty-first aspects, even when the disk device has a cache function, the request is processed without using the cache function. . As a result, the useful effects of the first to thirty-first inventions can be obtained in any disk device.

[0072]

DESCRIPTION OF THE PREFERRED EMBODIMENTS (First Embodiment) FIG. 1 is a block diagram showing a configuration of a disk control device according to a first embodiment of the present invention. 1, a disk control device according to the first embodiment includes a rearrangement control unit 11, a delay time management unit 12, a request management unit 13, a disk IO
It includes a unit 14, an access time management unit 15, and a hard disk 16.

The rearrangement control section 11 receives a processing request from a user management section (not shown). Then, based on the information provided from the delay time management unit 12 and the access time management unit 15, the rearrangement control unit 11 arranges a queue of processing requests (queues) managed by the request management unit 13 so as to shorten the seek distance. ) Inserts the received request at an arbitrary position.

Here, the user management section receives a read or write request for a specific video or audio from a plurality of users, and sequentially issues data processing requests relating to the request to the rearrangement control section 11. . The processing request issued by the user management unit includes a type, a logical address,
It includes the data size and the allowable delay time. The type indicates whether the request is a read request or a write request. The logical address is a logical address indicating the head position on the hard disk 16 of the data to be processed. The data size is the size of data to be processed. The permissible delay time is the maximum delay time (for example, for the user to play back as continuous video or audio) until the processing of the request is completed.

The delay time management unit 12 calculates and holds the completion time of the request managed by the request management unit 13.
Here, the completion time indicates a time at which the processing of the request must be completed. The request management unit 13 holds a plurality of requests continuously input from the rearrangement control unit 11 as a queue. The disk IO unit 14 sequentially reads requests from the queue of the request management unit 13 and outputs a corresponding command to the hard disk 16. This command is a command for reading or writing data of a predetermined size. The access time management unit 15 includes the disk IO unit 1
4 holds a time from issuing the command to receiving a response from the hard disk 16 as an access time. The hard disk 16 stores a plurality of types of multimedia data by dividing the data by a certain size.

FIG. 2 is a schematic diagram of a queue managed by the request management unit 13. In FIG. 2, R (5), R (1
0), R (23)... Represent a read request or a write request. The request R (10) at the head of the queue is taken out by the disk IO unit 14 when the immediately preceding request R (5) on the hard disk 16 is completed. The request R (x) represents a request newly output from the rearrangement control unit 11 (x is a positive integer; the same applies hereinafter).

First, the disk control device according to the first embodiment will be specifically described with reference to FIGS. FIG.
FIG. 3 is a block diagram showing a more detailed configuration of the disk control device according to the first embodiment. FIG. 4 is a diagram showing an example of block division and group division on the hard disk 16. FIG. 5 shows the completion time table 122.
FIG. 3 is a diagram showing an example of the configuration of FIG. FIG. 6 shows the request table 1
FIG. 3 is a diagram illustrating an example of a configuration of a first embodiment. FIG. 7 is a diagram illustrating an example of the configuration of the command management table 141a. FIG.
Is a diagram showing an example of the configuration of an access time table 152.

First, in the following description, the hard disk 16 is previously divided into areas in arbitrary byte units (hereinafter, referred to as blocks) as shown in FIG. 4, and the group g (1) To g (m) (m is a positive integer). In the first embodiment, a description will be given assuming that the block is divided into 64 Kbyte units.

Here, as an example of a method of dividing blocks into groups, there is a dividing method in which each group has an equal number of blocks. In this method, the hard disk 16
Blocks are allocated to each group by the number of quotients obtained by dividing the total number of blocks in the group by the maximum group number m. Group g with the number of the above quotients in order from the block with the first logical address
(1), and the subsequent blocks are similarly assigned to group g.
(2) is sequentially allocated to the group g (m) below. As another example of the division method, a quotient obtained by dividing the number of cylinders on the hard disk 16 by the maximum number m of groups is obtained, and the group belonging to the first cylinder is assigned so that each group is assigned the number of cylinders of the number of quotients. There is a method of allocating blocks in order from the block that exists.

Referring to FIG. 3, delay time management unit 12
A completion time calculation unit 121 and a completion time table 122 are provided. In addition, the disk IO unit 14 includes the request selection unit 14
1, a command issuing unit 142, and a command generating unit 143.
And The access time management unit 15 includes a group conversion unit 151, an access time table 152, and an access time calculation unit 153. The request management unit 13
Manages the queue in the form of a request table 131.

The completion time calculator 121 calculates the completion time f (x) based on the allowable delay time delay (x) given from the rearrangement controller 11. The completion time table 122 includes, as shown in FIG.
Is stored for each request R (x).

The request table 131 is a table for managing requests as described above. Request table 131
Stores a type k, a logical address a, an access time t, a scheduled completion time e, a prohibition flag flag, and an access type for each request R, as shown in FIG. Note that the order P in FIG. 6 indicates the processing order of the requests, that is, the order of the requests extracted by the request selection unit 141. 1 indicates the head of the queue, and the maximum value n indicates the end of the queue. This n is also the queue length.

The type k indicates whether the request is a read request or a write request. The logical address a is the request R
Is a logical address indicating the start position on the hard disk 16 of the data to be processed. Access time t
Is the execution time exe of request R (or up to request R)
Is a predicted value. The execution time exe of the request R is the time required to read and output the corresponding data from the hard disk 16 when the request R is a read request, and to store the corresponding data in the hard disk 16 when the request R is a write request. This is the time required from writing to receiving a response. The access time t in the request table 131 is given from the access time table 152. The scheduled completion time e is a time at which the execution of the request R is predicted to be completed. The prohibition flag “flag” indicates that a new request cannot be inserted before the request R due to the flag ON, or that the seek direction is reversed before and after the request R. The access type “type” indicates a relationship between the request R and the request R processed immediately before the request R, and indicates one of “independent”, “identical”, and “continuous”. For example, if the request immediately before the request R (x) is R (x ₋₁ ), type (x) = “independent” indicates that the request R (x) is executed alone, and type (x) = “Identical” means that the request R (x) has the same logical address a (x) as the request R (x ₋₁ ).
_-1 ), and type (x) = "continuous" means that the request R (x) accesses a logical address continuous to the logical address a (x _-1 ) of the request R (x _-1 ). It represents that.

The request selection section 141 has a request table 131
One or a plurality of requests (in the order of P) are taken out from the beginning or consecutively, and a command C to be issued to the hard disk 16 is generated. In generating the command C, the request selecting unit 141 has a command management table 141a, and registers the relationship between the extracted request R and the generated command C in the command management table 141a. As shown in FIG. 7, the command management table 141a stores a type K, a leading logical address A, a data size len, a request R, and a data position d for each command C.

The type K indicates whether the command C is a read process or a write process, and stores the same type as the type k described above. The head logical address A indicates the head logical address on the hard disk 16 accessed by the command C. The data size len is a value indicating how many blocks are continuously accessed. The data position d is a value that specifies the number of the block from the beginning that the request R accesses, and is set for each request. FIG.
In the first example shown in FIG. 1, a command C (1) is generated for three requests R (1) to R (3) for reading a continuous block of 3 × 64 kbytes. In this case, the data size len (1) is 3, the data positions d (1) to d (3) are 1 to 3, and the start logical address A (1) is the start logical address of the block to be accessed.

The command generator 143 generates a command C for obtaining an initial value of the access time t stored in the access time table 152. Command issuing unit 142
Receives the command C generated by the request selector 141 or the command generator 143 and issues the command C to the hard disk 16. Further, the command issuing unit 142 receives the execution result of the issued command C from the hard disk 16 and notifies the request selecting unit 141 of the execution result. Further, the command issuing unit 142 measures the execution time exe from the command issuance to the reception of the execution result, and obtains the access time calculation unit 153.
Output to

The access time calculation unit 153 stores the access time t in the access time table 152 based on the execution time exe received from the command issuing unit 142.
The group conversion unit 151 has information on grouping in the hard disk 16 in advance. The group conversion unit 151 receives the logical address a from the rearrangement control unit 11 and obtains a group g to which the logical address a belongs.

The access time table 152 is shown in FIG.
As shown in (a), using the type k of the request R and the data size len accessed by the request R as parameters, the access times t between the groups (see FIG. 4) in the hard disk 16 are permutated, respectively. Remember. FIG.
(A) shows an access time table of a read request of 64 kbytes when all blocks on the hard disk 16 are divided into m groups. FIG.
In (a), for example, t {1, 3} is the group g
It represents the time required to access group g (3) from (1). Therefore, in this case, the access time t = t {1,3}.

The access time management table 15
2 can arbitrarily set various tables corresponding to the data size to be accessed (for example, 128 k
Bytes or 256k bytes). Also, a table corresponding to a read request and a table corresponding to a write request are created separately. If the permutation of the group g is not considered, the access time management table 152 may be a table corresponding to the combination as shown in FIG. 8B. In this case, the amount of data to be stored can be reduced. Further, the group g that determines the access time may take a value of 3 or more. When determining from the permutation of three or more groups g, the access time is determined by the group g obtained from the logical addresses accessed by three or more requests.

Next, the processing performed by the disk control device according to the first embodiment will be specifically described with reference to FIGS. FIG. 9 is a flowchart showing an initial setting procedure of the access time table 152. FIG. 10 is a flowchart illustrating a procedure for inserting a new request R (x). FIG. 11 is a flowchart showing the calculation processing procedure of step S214 in FIG.

First, a method of obtaining an initial value of the access time t stored in the access time table 152 during operation will be described with reference to FIG. Here, 64
A case where a table corresponding to a read request of k bytes is obtained will be described. The command generator 143 selects at least one or more blocks belonging to the groups g (1) to g (m) in the hard disk 16 (step S101). For example, for each group, the first, center, and last three blocks in the group are selected. Next, the command generation unit 143 generates all permutations that can be generated when two blocks are selected from all the blocks selected in step S101 (step S102). For example, group g (1)
, The center block of group g (1), the head block of group g (1) and the last block of group g (1), the head block of group g (1) and the head block of group g (2),. It is so. Next, the command generation unit 143 generates a command C for accessing blocks according to the permutation for all permutations generated in step S102 (step S103).
The head logical address specified in each command is a logical address indicating the head of each block, the type of command is a read request, and the data size is 64 kbytes.

The command issuing section 142 receives the command C generated by the command generating section 143, issues the command C to the hard disk 16, and executes the command C for an execution time e until the response is received.
xe is measured (step S104). The access time calculation unit 153 acquires the logical address accessed according to the command C from the command issuing unit 142 and the execution time exe. Then, the access time calculation unit 153 refers to the group conversion unit 151 to determine which group the command C is to be processed between, and sets the execution time exe in a corresponding part of the access time table 152 as the access time t between the groups. It is stored (step S105). The processing of steps S103 to S105 is performed in step S102.
(Step S10).
6).

When a plurality of execution times exe are obtained for the same group (for example, the execution time exe from the head block of the group g (1) to the head block of the group g (2) and the group g (1) In the case where both the execution time exe for the last block of the group g (2) and the execution time exe for the last block of the group g (2) are obtained), the access time calculation unit 153
May calculate the average value of the execution time exe for the same group and store the average value in the access time table 152.

Next, the rearrangement control unit 11 sends a new request R
The operation after receiving (x) will be described with reference to FIG. Upon receiving the new request R (x) from the user management unit, the rearrangement control unit 11 transmits the allowable delay time delay (x) of the request R (x) to the completion time calculation unit 121. The completion time calculation unit 121 calculates the completion time f (x) by adding the allowable delay time delay (x) to the current time (step S201). Completion time calculation unit 121
Calculates the calculated completion time f (x) in the completion time table 12
2 and returned to the rearrangement control unit 11.

Further, upon receiving the request R (x) from the user management unit, the rearrangement control unit 11
Are scanned in order from the end of the queue, and a search is made as to whether there is a request R having the same type k and logical address a as the request R (x) (step S202). The target of this search processing and comparison processing described later is the request table 131.
Not only the request existing in the request selection unit 141 but also the request extracted by the request selection unit 141 immediately before. Step S20
2, when the same request R (h) is found, the reordering control unit 11 temporarily inserts the request R (x) into a position immediately after the request R (h), and the access type ty of the request R (h).
Pe (h) is set to “identical”, and the scheduled completion time e (x) of the request R (x) is set to “the same value as e (h)” (step S203). On the other hand, if no equal request R is found in step S202, the reordering control unit 11
Next, the request table 131 is scanned in order from the end of the queue, and a request in which the request R (x) and the type k are equal and the logical address a is set before and after by one block of the logical address a (x). It is searched whether or not R exists (step S204).

When the corresponding request R (i) is found in step S204, the rearrangement control unit 11
The logical address a (x) is compared with the logical address a (i) (step S205). If a (x)> a (i) in step S205, the rearrangement control unit 1
1 temporarily inserts the request R (x) into the position immediately after the request R (i), and sets the access type type (i) of the request R (i) to “continuous” (step S206). On the other hand, if a (x) <a (i) in step S205,
The rearrangement control unit 11 temporarily inserts the request R (x) into a position immediately before the request R (i), and sets the access type type (x) of the request R (x) to “continuous” (step S20).
7).

If the corresponding request R is not found in step S204, the reordering control unit 11
The request table 131 is changed from the end of the queue to the prohibition flag f.
It is searched whether there is a request R whose lag is ON (step S208). If the corresponding request R (j) is found in step S208, the sorting control unit 1
1 is the logical address a of all requests R from the request R (j) to the end of the request, and the logical address a of the request R (x).
(X) (Steps S209, S21)
1). If the corresponding request R is not found in step S208, the rearrangement control unit 11 compares the positions of the logical addresses a of all the requests R with the logical addresses a (x) of the requests R (x) (step S208). S210, S2
11).

As a result of comparing the positions of the logical addresses in step S211, if the logical address a (x) is located between the compared logical addresses, the order of the logical addresses a is not disturbed (ascending order or descending order). The request R (x) is temporarily inserted into the position, and the access type type (x) of the request R (x) is set to “independent” (Step S21)
2). Here, if the request R (q) for accessing the logical address a (x) is found in the comparison in step S211, the request R (x) and the request R (q) are of the type k for accessing the same logical address. Are different requirements. In this case, in order to preserve the execution order, the request R (x) is temporarily inserted immediately after the request R (q), and the access type type (q) of the request R (q) is set to “continuous”. On the other hand, as a result of comparing the positions of the logical addresses in step S211, if the logical address a (x) is not located between the compared logical addresses, the request R (x) is temporarily inserted at the end of the queue, and the request R (x) is inserted. (X) access type
(X) is set to “independent” (step S213). At this time, when the logical address a (x) is different from the logical address a of the last request R or the order of the logical address a of the request R immediately before the insertion position, the request R immediately before the last or insertion position is prohibited. The flag is set to ON.

Next, the rearrangement control unit 11 sends the request R
The expected completion time e (x) of (x) is calculated (Step S)
214). The processing performed in step S214 for calculating the estimated completion time e (x) is shown as a subroutine in FIG. In the following description, a request n times before the request R (x) is expressed as a request R (x− _n ), and a request n times after the request R (x) is expressed as a request R (x _{+ n} ). .

Referring to FIG. 11, sorting control unit 11
Follows the processing of steps S201 to S213.
After the request R (x) is temporarily inserted,
Request R (x_-n1) to the counter n indicating
The counter ds indicating whether continuous processing can be performed as much as 1
And initialization is performed (step S301). Next, average
The switching control unit 11 determines that the request n times before the request R (x)
That is, at first, the immediately preceding request R (x_-1) Access type
type (x_-1) Is determined (step S302). S
In step S302, the access type is determined.
(X_-1) Is “independent”, the data size is set to 1 (=
ds) and the first logical address is a (x_-1)
Step S303). In step S302, the access
Ip type (x _-1) Is "identical"
Add 1 to n and change the judgment target and go to step 302.
It returns (step S304). On the other hand, in step S302
The access type type (x_-1) Is "continuous"
In this case, the object to be determined is changed by adding 1 to the counter n.
At the same time, add 1 to the counter ds and return to step 302
(Steps S304 and S305). And sort
The control unit 11 executes steps S302 to S305 described above.
Data size (= ds) and the logical address at the beginning
A (x_-n), The type k (x) of the request R (x) and the argument
To the group conversion unit 151 together with the physical address a (x).
Put out.

The upper limit of the data size (= ds) is limited to the table held in the access time table 152. For example, the access time table 152
, Holds only a table up to 4 × 64 kbytes, the data size is 4 at the upper limit. Therefore, by holding the access time table 152 corresponding to a larger data size, it is possible to set this numerical limit to a larger value.

The group converter 151 converts the logical addresses a (x− _n ) and a
Corresponding groups g (x _−n ) and g based on (x)
(X) is calculated (step S306). Then, the access time table 152 shows that the access time t ｛g (x− _n ) from the group g (x− _n ) to the group g (x) is obtained from the table corresponding to the type k (x) and the data size.
g (x)} is extracted (step S307).

When extraction is performed in step S307,
The rearrangement control unit 11 sets the access time t {g (x− _n ), g (x)} extracted by the access time table 152 as the access time t (x) of the request R (x), and the scheduled completion time e ( x) is calculated based on the following equation (1) (step S308). e (x) = e (x− _n ) + t ｛g (x− _n ), g (x)｝ = e (x− _n ) + t (x) ‥‥ (1)

Next, the rearrangement control unit 11 sends the request R
Scheduled completion time e (x _{+ n} ) of the request R (x _{+ n} ) after (x)
Is updated based on the following equation (2) (step S30).
9). Note that the scheduled completion time e in equation (2)
(X _{+ n} ) ′ is the scheduled completion time that the request R (x _{+ n} ) had before the insertion of the request R (x). e (x _{+ n} ) = e (x _{+ n} ) '+ t (x) + t {g (x), g (x _{+ n} )}-t {g (x- _n ), g (x _{+ n} )}} ‥ (2)

Referring again to FIG. 10, as described above, calculation of expected completion time e (x) and expected completion time e (x _{+ n} )
Are completed, the rearrangement control unit 11 returns all newly calculated scheduled completion times e (x) and e (x _{+ n} ).
Is the completion time f stored in the completion time table 122.
It is checked whether they are within (x) and f (x _{+ n} ), respectively (step S215). If all of the confirmations in step S215 are within the completion time f, the request R
The rearrangement process ends with the temporary insertion position of (x) as the main insertion position. On the other hand, if all of the confirmations in step S215 are not within the completion time f, the temporary insertion position is determined to be inappropriate, and the change in the scheduled completion time e of each request R performed so far is returned to the value before the temporary insertion processing. The request R (x) is inserted at the end of the queue, and the access type of the request R (x) is type
(X) is set to “independent”, and the sorting process ends (step S216). At this time, if the logical address a (x) of the request R (x) is different from the order of the logical addresses of the request that was originally at the tail, the prohibition flag flag of the request that was originally at the tail is set to ON. With the above procedure, the queue reordering process of the request table 131 into which the new request R (x) has been inserted ends.

Next, the request R in the request table 131
Will be described to the hard disk 16. The request selection unit 141 sequentially starts from the top of the request table 131
That is, the requests R (r) are extracted in the order of the order P. At this time, the access type type of the first request R (r)
When (r) is “same” or “continuous”, R (r)
The following request R (r _{+ 1} ) is also taken out. The request R (r ₊₁ ) also has an access type type (r ₊₁ )
And the request R is continuously taken out until the access type “independent” becomes “independent”. The number of requests R that can be continuously taken out is equal to the above-mentioned data size (= d except for those having the same access type “type”).
As in the case of the upper limit of s), the access time table 15
2 is limited to the table it holds. Therefore, the access time table 1 corresponding to a larger data size
By holding 52, it is possible to increase the number of requests R to be continuously taken out.

The request selecting section 141 performs the above-mentioned single request R
(R) or requests R (r) to R (r)
₊₁ ) is generated on the command management table 141a (see FIG. 7). The command issuing unit 142 issues the command C (s) generated by the request selecting unit 141 to the hard disk 16. Further, the command issuing unit 142 measures an execution time exe (s) from issuing the command C (s) to receiving a response indicating that the processing has been completed in the hard disk 16. Then, the command issuing unit 142 sets the execution time exe (s) to the type K
(S), the data size len (s), and the leading logical address A (s) are notified to the access time management unit 153.

The command issuing section 142 notifies the request selecting section 141 of the result of receiving the response. Then, the request selecting unit 141 receives this notification and returns to the user that the processing of the request has been completed by an arbitrary method. On this occasion,
The request selecting unit 141 returns, as a result of execution of a plurality of requests by one command C (s), an execution result for each request. In this case, in the case of a read request, the read data is divided into 64 kbyte units from the beginning,
The data indicated by the data position d of each request is returned as a response from the command management table 141a. With the above procedure, the processing on the hard disk 16 for the request R included in the request table 131 ends.

Finally, an operation of updating the access time t stored in the access time table 152 during operation will be described. The operation of updating the access time t is the same as the operation for setting the initial value of the access time t described above. Access time calculation unit 153
Obtains information on the command C (s) just executed from the command issuing unit 142 as described above. And
Command C (s _-1 ) executed immediately before command C (s)
Based on the logical address A (s _-1 ) and the logical address A (s) of the group g.
(S) and g (s _-1 ) are obtained. Thereafter, the access time calculation unit 153 calculates the access time t ｛g (s ₋₁ ), g for the table in the access time table 152 where the type K (s) and the data size len (s) match.
(S) Rewrite the value of｝ to the value of the execution time exe (s). Such a process is performed every time the command C is executed, the latest access time t is always obtained, and the value in the access time table 152 is updated.

Note that the rearrangement control unit 11 sets the access time table 1 regardless of the presence or absence of a new request R (x).
Every time the value in 52 is updated, the requests in the request table 131 may be rearranged.

As described above, the disk control apparatus according to the first embodiment of the present invention predicts the execution time of a request each time a request is input, and sets the limit value of the delay time specified for each request. Reorder the requests so that execution of the request is completed within. Further, since the prediction of the execution time of the request only refers to the stored access time, no special operation is required to obtain the predicted value at the time of rearrangement. Therefore, the access efficiency is improved, and the request can be executed within the limit value of the delay time. In particular, when accessing data that requires real-time properties such as moving pictures and audio data, it is possible to improve access efficiency and secure real-time properties.

Although the first embodiment deals with a read request and a write request, the present invention can be executed for other requests. In this case, an access time table corresponding to the type of the request to be added may be created and the access time of the added request may be stored. Further, only a read request or a write request may be used.

Although each request is a request for accessing fixed-length data of 64 Kbytes, the data length to be accessed may be specified in the request, and the variable data length may be accessed for each request. In this case, a method of referring to an access time table corresponding to the data length, an access time table corresponding to two types of data lengths are prepared, and the access time of the corresponding data length is determined using the above two types of access times. There is a way to calculate. Specifically, the data length L1
, The access time of the data length L2 (L1 <L2) is T2, and the calculated access time is T,
Assuming that the data length is L, the access time T is determined by the following equation (3), for example, by interpolation using a linear equation. T = (T1−T2) / (L1−L2) × L + T1 / L1 × (L1−L2) / (T1−T2) (3)

In the first embodiment, the allowable delay time delay is specified in the request R. However, several allowable delay times are prepared in advance, and the identification number is assigned to each allowable delay time. And the value specified in the request R may be its identification number. In this case, the correspondence table of the permissible delay time values and the identification numbers thereof is stored in the delay time management unit 12, and the completion time calculation unit 121 converts the permissible delay time values into the permissible delay time values based on the correspondence table and calculates the completion time f. May be. Further, the allowable delay time may be a fixed value of the system, and the completion time calculating unit 121 may calculate the completion time f using the fixed value. Further, the completion time calculation unit 121 holds a fixed value of the system corresponding to the request type k,
The permissible delay time is obtained by referring to the request type k,
The completion time f may be calculated.

In the first embodiment, requests for accessing consecutive blocks can be collectively executed by one command C. However, this function need not be particularly supported. . That is, the request selecting unit 141 does not need to execute the operation when the access type is “continuous”. In this case, in the first embodiment, the data size for accessing the hard disk 16 is always equal to the size accessed by each request. Therefore, the data length accessed by each request may be only the fixed length, and the access time table 152 may hold only the value corresponding to the fixed length. Similarly, a number of data lengths to be accessed by each request may be determined in advance, and an access time table 152 may be prepared according to the data length to determine the access time t.

In the first embodiment, the initialization value of the access time t is obtained before the operation of the disk control device. However, the command generation unit 143, the command issuing unit 142,
The access time t acquired by another device having the same configuration as the access time management unit 15 is stored in the access time table 1.
52 may be set. The set values are recorded on the hard disk 16 and the command generation unit 143
Generates a command C for reading the set value, obtains it from the hard disk 16 through the command issuing unit 142,
It may be stored in the access time table 152. Also,
The command generation unit 143 sets the access time table 152
, A value stored in the access time table 152 may be received, a write command using the value as write data may be generated, and may be written to a predetermined position on the hard disk 16 via the command issuing unit 142. The written value can be read out by the command generation unit 143 generating a command C for reading this value and issuing the command C from the command issuing unit 142.

In the first embodiment, another storage device capable of writing and reading the access time t is provided in addition to the hard disk 16, and a command generation unit for generating a command C for the storage device, May be newly provided to the storage device, and a command issuing unit that receives a response may be newly provided, and the access time t may be stored in the storage device by writing it, read, and stored in the access time table 152.

In the first embodiment, the access time calculation unit 153 updates the access time t based on the execution time exe of a certain number of requests R for each permutation of the two groups g. For all the requests R performed, the access time t may be calculated and updated for all the permutations of the two groups g for every fixed number, or for all the executed requests R, the permutations of the two groups g may be The access time t of the permutation of the group g for which the execution of the request has been completed may be calculated and updated every time the execution of the request is completed for each fixed number.

Further, in the first embodiment, the delay time management unit 12 calculates and holds the completion time f for each request, but the delay time management unit 12 sets the allowable delay time delay to the initial value for each request. The permissible time, which is the time allowed for the execution of the request to be completed, is held, and the permissible time is updated each time the rearrangement is performed. Access time t
May be calculated and the sum of the allowable time and the access time a may be compared. If the sum of the access times is small, the scheduled completion time e is considered to be equal to or less than the completion time f, and the rearrangement may be performed.

(Second Embodiment) FIG. 12 is a block diagram showing a configuration of a disk control device according to a second embodiment of the present invention. 12, the disk control device according to the second embodiment includes a rearrangement control unit 21, a delay time management unit 22, a request management unit 23, and a disk IO unit 2.
4 and a hard disk 26.

The rearrangement control section 21 receives a processing request from a user management section (not shown). Then, based on the information provided from the delay time management unit 22, the rearrangement control unit 21 places an arbitrary position in the queue (queue) of the processing waiting requests managed by the request management unit 23 so that the seek distance becomes short. Insert the accepted request. The user management unit is the same as that described in the first embodiment, and a description thereof will not be repeated.

The delay time management unit 22 calculates and holds the allowable delay time of the request managed by the request management unit 23. The request management unit 23 holds a plurality of requests continuously input from the rearrangement control unit 21 as a queue. The schematic diagram of the queue managed by the request management unit 23 is the same as that shown in FIG. The disk IO unit 24 includes the request management unit 2
Requests are sequentially read from the queue No. 3 and a corresponding command is output to the hard disk 26. This command is a command for reading or writing data of a predetermined size. The hard disk 26 stores a plurality of types of multimedia data by dividing the data by a certain size.

First, a disk controller according to the second embodiment will be specifically described with reference to FIGS. FIG. 13 is a block diagram showing a more detailed configuration of the disk control device according to the second embodiment. FIG.
Is a diagram showing an example of a configuration of a request table 231.

Referring to FIG. 13, delay time management unit 22
Includes an insertable request number calculation unit 221, an average execution time calculation unit 222, and an average execution time holding unit 223. Further, the disk IO unit 24 includes a request selecting unit 241, a command issuing unit 242, and a command generating unit 243. The request management unit 23 stores the queue in the request table 2
31 are managed.

The number-of-insertable-requests calculating section 221 receives the allowable delay time delay given from the rearrangement control section 21.
The number of insertable requests i (x) is calculated from (x) and the average execution time of the command held by the average execution time holding unit 223. The average execution time calculation unit 222 calculates an average execution time based on the execution time received from the command issuing unit 242. The average execution time holding unit 223 holds the average execution time of the command calculated by the average execution time calculation unit 222.

The request table 231 is a table for managing requests as described above. Request table 231
As shown in FIG. 14, the type k, the logical address a, the number i of insertable requests and the prohibition flag
g is stored. The order P in FIG.
Indicates the processing order of requests, that is, the order of requests retrieved by the request selection unit 241. 1 is the head of the queue,
The maximum value n indicates the end of the queue. This n is also the queue length.

The type k indicates whether the request is a read request or a write request. The logical address a is the request R
Is a logical address indicating the head position of the data to be processed on the hard disk 26. The number i of requests that can be inserted indicates the number of requests R that can be inserted before the request R (x) in the queue. The prohibition flag “flag” indicates that a new request cannot be inserted before the request R due to the flag ON, or that the seek direction is reversed before and after the request R.

The request selection section 241 includes a request table 231
One request is taken from the beginning of the command and a command C to be issued to the hard disk 26 is generated. Command generator 243
Generates a command C for obtaining an initial value of the average execution time of the command held by the average execution time holding unit 223. The command issuing unit 242 receives the command C generated by the request selecting unit 241 or the command generating unit 243,
It is issued to the hard disk 26. The command issuing unit 242 receives the execution result of the issued command C from the hard disk 26 and notifies the request selecting unit 241 of the execution result. Further, the command issuing unit 242 measures the execution time from the command issuance to the reception of the execution result, and outputs it to the average execution time calculation unit 222.

The reordering unit 21 sets the newly received request R (x) so as not to exceed the number of insertable requests i (x) of each request, and sets the logical address a to which each request accesses.
Rearrange them so that the order of (x) is as constant as possible.

Next, with reference to FIGS. 15 and 16, the processing performed by the disk control device according to the second embodiment will be specifically described. FIG. 15 shows the average execution time holding unit 223.
6 is a flowchart showing an initial setting procedure of FIG. FIG.
13 is a flowchart showing a procedure for inserting a new request R (x).

First, it is assumed that the average execution time holding unit 2
The method for obtaining the initial value of the average execution time held in 23 is as follows.
This will be described with reference to FIG. The command generation unit 243
A command C for accessing 64 Kbytes of data starting from a random logical address on the hard disk 26 is generated. When generating the command C, the command type k
(Corresponding to the type of request) is also randomly selected (step S401).

The command issuing section 242 receives the command C generated by the command generating section 243, issues the command C to the hard disk 26, and measures the execution time until the response is received (step S402). Average execution time calculation unit 2
22 indicates the type k of the command C from the command issuing unit 242
And the execution time. Then, the average execution time calculation unit 222 classifies the execution times according to the type k, and calculates the sum of the execution times for each classification (step S403). Further, the average execution time calculation unit 222 counts the number of execution times received for each type k (step S40).
3). The average execution time calculation unit 222 determines whether the number of execution times received for a certain type k has reached a predetermined value (for example, 10,000) (step S404). The average execution time is calculated and stored in the average execution time holding unit 223 (Step S)
405). With the above processing, the initial value of the average execution time is held in the average execution time holding unit 223.

Next, the rearrangement control unit 21 determines that the new request R
The operation after receiving (x) will be described with reference to FIG. Upon receiving the new request R (x) from the user management unit, the reordering control unit 21 sends the allowable delay time delay (x) and the type k (x) of the request R (x) to the insertable request number calculation unit 221. Send. Insertable request number calculation unit 2
21 divides the permissible delay time delay (x) by the average execution time held by the average execution time holding unit 223 to calculate the number i (x) of insertable requests (step S501). The average execution time to be referred to is the type k of the request R (x).
(X). Then, the number-of-insertable-requests calculating unit 221 returns the calculated number of insertable requests i (x) to the rearrangement control unit 21.

Next, the reordering control unit 21 scans the request table 231 sequentially from the end of the queue, and searches for a request R for which the prohibition flag flag is ON (step S502). The targets of the search processing and the comparison processing described later include not only the request existing in the request table 231 but also the request extracted by the request selection unit 241 immediately before. If a corresponding request R (j) is found in step S502, the reordering control unit 21 determines the logical addresses a of all the requests R from the request R (j) to the end and the logical addresses of the requests R (x). Compare the position with address a (x) (steps S503, S
505). If no corresponding request R is found in step S502, the rearrangement control unit 21 compares the positions of the logical addresses a of all the requests R with the logical addresses a (x) of the requests R (x) (step S502). S504,
S505).

As a result of comparing the positions of the logical addresses in step S505, if the logical address a (x) is located between the compared logical addresses, the order of the logical addresses a is not disturbed (ascending order or descending order). The request R (x) is inserted into the position (step S506). On the other hand, as a result of comparing the positions of the logical addresses in step S505, if the logical address a (x) is not located between the compared logical addresses, the request R (x) is placed at the end of the queue.
Is inserted (step S507). At this time, when the logical address a (x) is different from the logical address a of the last request R or the order of the logical address a of the request R immediately before the insertion position, the request R immediately before the last or insertion position is prohibited. The flag is set to ON.

Next, the rearrangement control unit 21 sends the request R
The number of requests R arranged before the position where (x) is inserted is counted, and the counted number of requests R is subtracted from the number of insertable requests i (x) (step S508). Then, the rearrangement control unit 21 calculates 1 from the number of insertable requests i (x _{+ n} ) from the request next to the request R (x) to the last request.
Is subtracted (step S509). As a result, the prohibition flag flag (x _{+ n} ) is set to ON for the request R (x _{+ n} ) in which the value of the number of insertable requests i (x _{+ n} ) is 0 (step S510). Thus, the rearrangement of the request R (x) is completed.

Next, the request R stored in the request table 231
Is issued to the hard disk 26. The request selecting unit 241 sequentially starts from the top of the request table 231
That is, the requests R are taken out in the order of the order P. The request selector 141 generates a command C relating to the extracted request R. Command C includes type k, logical address a, and data size. The command issuing unit 242 transmits the command C generated by the request selecting unit 241 to the hard disk 2
Issue to 6. In addition, the command issuing unit 242 measures an execution time exe from issuing the command C to receiving a response indicating that the process has been completed in the hard disk 26.
Then, the command issuing unit 242 notifies the execution time calculation unit 222 of the execution time exe along with the type K. Also,
The command issuing unit 242 notifies the request selection unit 241 of the result of receiving the response. Then, the request selection unit 241
Upon receiving this notification, the user is notified that the processing of the request has been completed. With the above procedure, the processing on the hard disk 16 for the request R in the request table 231 ends. Needless to say, each time the processing of one request is completed, the value of the insertable request number i of each request existing in the request table 231 is decremented by one.

Finally, an operation of updating the average execution time held in the average execution time holding unit 223 during operation will be described. The operation for updating the average execution time is the same as the operation for setting the initial value of the average execution time described above. The average execution time calculation unit 222 obtains the execution time and the type k of the command C executed from the command issuing unit 242 as described above. Then, the average execution time calculation unit 222 classifies the execution times according to the types that receive the execution times, and calculates the sum of the execution times for each classification. Further, the number of execution times is counted for each type k. Then, when the number of received execution times exe of a certain type k reaches a predetermined value, the average execution time calculation unit 222
The average execution time is calculated and stored in the average execution time holding unit 223. Such processing is performed every time the command C is executed, the latest access time t is always obtained, and the value in the average execution time holding unit 223 is updated.

As described above, the disk controller according to the second embodiment of the present invention obtains the number of insertable requests from the limit value of the delay time held by a request every time a request is input, and Reorder requests so that no more requests are inserted before the number of requests. Therefore, the access efficiency is improved, and the execution of the request can be completed within the limit value of the delay time.

Although the read request and the write request are handled in the second embodiment, other requests may be added. In this case, the average execution time holding unit 223
Holds the average execution time corresponding to the type k of the added request, and calculates the number i of insertable requests from the average execution time.

In the second embodiment, the average execution time is updated by measuring the execution time of the command in operation. However, the average execution time need not be updated. Further, the initial value of the average execution time is measured. The initial value may be obtained from another device having the same configuration as that of the second embodiment, and may be stored in the average execution time storage unit 223. Good.

In the second embodiment, the command generation unit 243 sets the average execution time on the hard disk 26.
A command to write to the above predetermined position and a command to read from the hard disk 26 are generated, the average execution time is stored on the hard disk 26 by the write command, the average execution time is acquired from the hard disk 26 by the read command, and the average execution time is obtained. It may be stored in the time holding unit 223.

Further, in the second embodiment, a storage device other than the hard disk 26 capable of storing the average execution time, a command generation unit for generating write and read commands for the storage device, and issuing the generated command The average execution time may be stored in the storage device by a write command, the average execution time may be read from the storage device by a read command, and stored in the average execution time holding unit 223.

Further, in the second embodiment, the request selecting section 241 always selects only one request. However, the request selecting section 241 combines two requests in the same or continuous data area on the hard disk 26 into one. A command may be issued to the hard disk 26. In this case, the request selection unit 24
When extracting a request from the request table 231, the request 1 refers to the logical address a, the request type k, and the number of insertable requests i for the extracted request and the first request in the request table 231. If the request types k are the same and the logical addresses a indicate the same value or different values by 64 kbytes, and if the number i of insertable requests having a smaller logical address a is other than 0, the request is further extracted. . This operation may be repeated, and the requests taken out to the hard disk 26 may be collected and issued as one command.

By the way, as described above, the disk control devices according to the first and second embodiments of the present invention provide the disk device (the hard disk 16, hard disk 16,
The execution time in step 26) is measured, and the optimal requests are rearranged based on this time. However, in a disk device having a cache function, when the cache function is operated, a read / write operation performed on a cache is performed for an arbitrary command instead of a read / write operation performed on the disk device. A response of the write operation may be returned. For this reason, when the cache function is operated, it is conceivable that the execution time in the disk device cannot be measured accurately. To cope with this, it is only necessary to use the cache function of the disk device. There are the following two methods for eliminating the use of the cache function. One is a method of turning off the setting of the cache function itself. The other is F in the read / write request.
In this method, a request is issued by designating a UA (Force Unit Access) bit. The FUA bit is a bit for designating forcible access to the device. When the FUA bit is designated, the operation is the same as when the setting of the cache function is turned off. By setting the cache function to OFF or issuing a request by designating the FUA bit, the above-described useful effects of the present invention can be achieved for any disk device.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of a disk control device according to a first embodiment of the present invention.

FIG. 2 is a diagram showing a schematic diagram of a queue managed by a request management unit 13;

FIG. 3 is a block diagram showing a more detailed configuration of the disk control device according to the first embodiment of the present invention.

FIG. 4 is a diagram showing an example of block division and group division on a hard disk 16;

FIG. 5 is a diagram showing an example of the configuration of a completion time table 122.

FIG. 6 is a diagram showing an example of a configuration of a request table 131.

FIG. 7 is a diagram illustrating an example of a configuration of a command management table 141a.

FIG. 8 is a diagram showing an example of the configuration of an access time table 152.

FIG. 9 is a flowchart showing an initial setting procedure of an access time table 152;

FIG. 10 shows a new request R in the sorting control unit 11.
It is a flowchart which shows the insertion processing procedure of (x).

FIG. 11 is a flowchart illustrating a calculation processing procedure in step S214 in FIG. 10;

FIG. 12 is a block diagram illustrating a configuration of a disk control device according to a second embodiment of the present invention.

FIG. 13 is a block diagram showing a more detailed configuration of a disk control device according to a second embodiment of the present invention.

FIG. 14 is a diagram showing an example of a configuration of a request table 231.

FIG. 15 is a flowchart illustrating an initial setting procedure of the average execution time holding unit 223;

FIG. 16 shows a new request R in the sorting control unit 21.
It is a flowchart which shows the insertion processing procedure of (x).

FIG. 17 is a block diagram showing a configuration of a conventional disk control device.

[Explanation of symbols]

 11, 21, 301 ... rearrangement control unit 12, 22 ... delay time management unit 13, 23, 303 ... request management unit 14, 24 ... disk IO unit 15 ... access time management unit 16, 26, 306 ... hard disk 121 ... completion Time calculation unit 122 ... Completion time table 131,231 ... Request table 141,241 ... Request selection unit 141a ... Command management table 142,242 ... Command issuing unit 143,243 ... Command generation unit 151 ... Group conversion unit 152 ... Access time table 153 access time calculation unit 221 insertion request count calculation unit 222 average execution time calculation unit 223 average execution time holding unit 302 allowable delay data management unit

 ──────────────────────────────────────────────────続 き Continued on the front page (72) Inventor Yukiko Ito 1006 Kazuma Kadoma, Kadoma City, Osaka Prefecture Matsushita Electric Industrial Co., Ltd. F-term (reference) 5B065 BA01 CA12 CC06 CC08 CH15 EK10

Claims (32)

[Claims] 1. A disk controller for rearranging requests and issuing them to a disk device so that execution of each request is completed within a delay time individually provided by each request, the disk control device comprising: Request management means for holding the request, disk IO means for taking out the request held by the request management means, issuing the request to the disk device, and receiving a response to the request from the disk device, and the request in the disk device Access time management means for holding a time necessary for the processing as an access time; for each of the requests, a delay time management means for holding a completion time at which the processing of the request should be completed; and for the completion time and the access time, And a rearrangement control unit for rearranging the requests held by the request management unit based on the request. 2. The disk according to claim 1, wherein the access time is determined based on a time from when a request is issued to the disk device to when a response to the request is received from the disk device. Control device. 3. The access time divides an area on the disk device accessible to a request into a plurality of groups, which are continuous areas, and determines a position on the disk device to which the request accesses for each request. 3. The disk control device according to claim 2, wherein the groups to which each of the groups belongs are determined, and are determined by the order of the groups corresponding to the execution order of the requests. 4. The disk control device according to claim 3, wherein the access time is individually determined for each type of the request. 5. The disk control device according to claim 3, wherein the access time is individually determined for each data size accessed by the request. 6. The disk control device according to claim 5, wherein the data size is a fixed length. 7. The disk drive according to claim 3, wherein the area on the disk device is composed of a plurality of fixed-length blocks, and the groups each have an equal number of the blocks. Disk control unit. 8. The disk control device according to claim 3, wherein the disk device is a hard disk including a plurality of cylinders, and the groups each have an equal number of the cylinders. 9. The method according to claim 3, wherein the order of the groups is a permutation or combination of two groups.
9. The disk control device according to any one of items 1 to 8. 10. The disk control device according to claim 4, wherein the types of the requests are a read request for reading data from the disk device and a write request for writing data to the disk device. 11. The completion time is a time obtained by adding the delayable time of the request to the time at which the request is input, and the reordering control means executes each of the requests at each of the completion times. 2. The disk control device according to claim 1, wherein the requests are rearranged so as to be completed before the time. 12. The delay time management means has a predetermined constant, the completion time is a time obtained by adding the constant to a time at which the request is input, and the rearrangement control means 2. The disk control device according to claim 1, wherein the requests are rearranged such that execution of each of the requests is completed before each of the completion times. 13. The disk IO unit further includes an execution time measurement unit that measures an execution time from when a request is issued to the disk device to when a response to the request is received from the disk device. 7. The disk control device according to claim 3, wherein the access time management unit further includes an access time calculation unit that updates the execution time as a new access time. 14. The access time calculation means classifies the execution time according to the order of the group, the type of the request, and the data size, and updates the access time corresponding to the classification. 14. The disk control device according to claim 13. 15. The disk according to claim 13, wherein the access time management means updates the access time for each of the categories each time the execution of a certain number of requests is completed. Control device. 16. The access time management means updates the access time of the completed classification every time the execution of a fixed number of requests for each of the classifications is completed. 15. The disk control device according to 13 or 14. 17. The disk control device according to claim 13, wherein the access time is obtained by an average value of the classified execution times. 18. The apparatus according to claim 18, wherein said disk IO means further comprises a command generation means for generating a request issued to said disk device, and said access time calculation means obtains an initial value of said access time. Item 18. The disk control device according to any one of Items 13 to 17. 19. A storage device for storing the access time by a write command and outputting the read time by a read command; and a command issuing unit for issuing a write command and a read command to the storage device. 2. The method according to claim 1, wherein the write command is generated to store the access time in the storage device, and the read command is generated to obtain the access time from the storage device.
19. The disk control device according to any one of claims 18 to 18. 20. The storage device is configured in the disk device, and the command issuing unit is configured to execute the disk IO
20. The disk control device according to claim 19, wherein the disk control device is configured as a unit. 21. The request management means, for each request,
A flag indicating that a direction in which the disk device seeks is changed, a time to wait for completion of another request in a queue at a time when a request is input, and a first access time obtained by adding an access time of the request. A request input to the rearrangement control unit as a first request, and a last request in which the flag is set in the request management unit, or if the flag is not set. When the last request retrieved from the request management unit by the disk IO unit is a second request, the rearrangement control unit replaces the first request with a logical address accessed by the first request. Tentatively inserting the logical addresses to be accessed from the last request to the second request in the request management means so as not to disturb the order of the logical addresses. In the case where the request is set, the flag of the first request is set, and after the provisional insertion, the first time is calculated again for each request after the first request in the request management unit. If the recalculated first time does not exceed the completion time, the provisionally inserted position is set as the insertion position of the first request. If the recalculated first time exceeds the completion time, 2. The disk control device according to claim 1, wherein a last position of the request management unit is an insertion position of the first request. 22. When a request for accessing the same data as the first request is a third request, the rearrangement control unit determines that the first request and the third request are read requests. And when there is no other request to write data accessed by the first or third request between the first request and the third request, the access time of the first request is set to 0. 22. The disk control device according to claim 21, wherein the disk IO unit sets an execution result of the third request as an execution result of the first request. 23. When a request to write data at the same position as the first request is a third request, the rearrangement control unit determines that the first request is a write request and the first request is a write request. If there is no other request for reading data accessed by the first or third request between the first request and the third request, the first request is inserted immediately after the third request. In addition, the access time of the third request is set to 0, and the disk IO unit does not issue the third request,
22. The disk control device according to claim 21, wherein an execution result of the first request is an execution result of the third request. 24. When the first request is a request to read data at a position that is continuous with the data to be read, the rearrangement control unit sets the third request to the first request and the third request. And if there is no other request to write to the data accessed by the first request, the first request is inserted immediately before or immediately after the third request according to the sequence of the consecutive positions; The access time of the first request or the third request is an access time corresponding to a data size read by the first request and the third request, and the disk IO means includes: 22. The disk control device according to claim 21, wherein the requests of the third position are collected and issued to the disk device as one request for reading data at the consecutive positions. 25. When a request to write data located at a position contiguous with the data to be written in the first request is a third request, the rearrangement control means: the first request and the third request And if there is no other request to read the data accessed by the first request, the first request is inserted immediately before or after the third request according to the sequence of the consecutive positions; The access time of the first request or the third request is an access time corresponding to a data size read by the first request and the third request, and the disk IO means includes: 22. The disk control device according to claim 21, wherein the requests of the third position are collected and issued to the disk device as one request for reading data at the consecutive positions. 26. A disk controller for rearranging requests and issuing them to a disk device so that execution of each of the requests is completed within a delay time individually provided by each of the requests. Request management means for holding the request, disk IO means for taking out the request held by the request management means and issuing the request to the disk device, and receiving a response to the request from the disk device; and Possible time, in addition to the number of requests already held in the request management means when the request was received, a delay time management means for obtaining the number of insertable requests indicating the number of requests that may be executed earlier, A rearrangement control unit that rearranges the requests held by the request management unit based on the number of insertable requests,
Disk control unit. 27. A reference execution time holding means for holding a reference execution time which is a time from issuing a request to the disk device to receiving a response to the request from the disk device. 27. The apparatus according to claim 26, further comprising: an insertable request number calculating unit that calculates the insertable request number from the reference execution time and the delay time specified in each of the requests. Disk control unit. 28. The disk IO means, comprising: a command generating means for generating a request to be issued to the disk device; and executing the command from issuing the request to the disk device until receiving a response to the request from the disk device. 28. The method according to claim 27, further comprising: an execution time measurement unit that measures time; and a reference execution time calculation unit that obtains an initial value of the reference execution time from the execution time measured by the execution time measurement unit. The disk control device according to the above. 29. The disk IO means, comprising: an execution time measuring means for measuring an execution time from when a request is issued to the disk device to when a response to the request is received from the disk device; And a reference execution time calculating means for obtaining a new reference execution time from the measured execution time and updating the reference execution time held by the reference execution time holding means. 2
8. The disk control device according to 7. 30. The disk control device according to claim 27, wherein the reference execution time is obtained by an average value of the measured execution times. 31. The request management means, for each request,
A flag indicating that the seek direction of the disk device changes or that the number of insertable requests is 0;
The number of insertable requests is held, and the request input to the rearrangement control unit is the first request, and the last request in which the flag is set in the request management unit, or the flag is If the request has not been set, if the last request retrieved from the request management means by the disk IO means is the second request, the rearrangement control means sets the first request as the first request. The logical address to be accessed is inserted at a position that does not disturb the order of the logical addresses accessed from the last request to the second request in the request management means, and the order of the logical addresses is changed by the insertion. In the case, the flag of the first request is set, and the number of requests existing before the insertion position of the first request is subtracted from the number of insertable requests of the first request. Required Subtracting 1 from the number of insertable requests of all requests after the insertion position, and setting the flag for a request in which the number of insertable requests becomes 0 by the subtraction, The disk control device according to claim 26. 32. When the disk device has a cache function, the disk IO means issues the request to the disk device and receives a response to the request without using the cache function. 32. The disk control device according to claim 1, wherein the control is performed.