JP2009163499A - Storage device, method, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a storage device composed of a plurality of storage media, a method, and a program for reducing access time for commands as a whole. <P>SOLUTION: The storage device includes the plurality of storage media 221 for storing data in a redundant fashion, and a command processing part 21 for analyzing a plurality of access commands that are put in a command queue 222 corresponding to each of the plurality of the storage media and are in an issuance standby state, performing scheduling of issuance order of the access commands so that access time as a single storage medium is reduced, and selecting a combination of the storage media at the access command issuance destination from the plurality of the storage media for which the scheduling has been performed as a single storage medium, so as to accelerate the completion of access for the plurality of the storage media as a whole. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a storage apparatus, and more particularly to an apparatus and a program having a plurality of storage media.

  With the evolution of information communication technology and the expansion of its application fields, the amount of data handled by information processing systems and their importance are increasing. For this reason, a storage system for storing data is required to have high reliability. At the same time, high-speed access is required to process a large amount of data in a short time.

  RAID (Redundancy Array Of Inexpensive Disks) listed in Non-Patent Document 1 is used as a method for simultaneously improving the reliability and performance of a storage having a storage medium in response to these requirements. RAID is a technique for improving reliability and access performance by providing redundancy among a plurality of storage media in a storage. There are a plurality of types of RAID depending on how to take redundancy between storage media.

<RAID1>
RAID1 is a method for mirroring data on a plurality of storage media. In RAID1, since the same data is stored in a plurality of storage media, data is not lost unless all the storage media storing the data fail. For this reason, the reliability with respect to the failure of a storage medium improves.

  In RAID1, since data is stored in a plurality of storage media, parallel processing is possible by performing a plurality of Read accesses to different storage media. For this reason, RAID 1 improves parallel access performance with respect to Read access compared to a single storage medium.

  Furthermore, in RAID1, by performing read access on data that is continuous in parallel, throughput performance during continuous access is improved with respect to read access.

<RAID5>
Next, RAID5 can be cited as a method of having a parity with data among a plurality of storage media. In RAID5, one of the plurality of storage media stores the parity of data on all other storage media. Therefore, in RAID5, when any one storage medium fails, data can be restored using parity. Thus, RAID 5 improves the reliability against storage medium failures.

  In RAID5, when a plurality of read accesses occur simultaneously, if the storage media storing the data are different from each other, the access can be processed in parallel.

  Therefore, when RAID 5 is used, parallel access performance is improved with respect to Read access compared to a single storage medium.

  Further, as in RAID 1, RAID 5 performs read access on continuous data in parallel, thereby improving throughput performance during continuous access with respect to Read access.

  In RAID 5, each storage medium is divided into areas of a certain size, and the parity is calculated and stored in units of areas. Furthermore, the storage medium that plays the role of storing the parity differs for each area.

<RAID 6>
Similar to RAID 5, RAID 6 can be cited as a method having parity between a plurality of storage media. RAID 6 differs from RAID 5 in that the parity calculated by two different methods is stored in different storage media. For this reason, in RAID 6, it is possible to restore data until two storage media fail.

  In RAID 6, the read access can be performed in parallel when the data storage destination storage medium is different, as in RAID 5.

  Recently, not only storage but also storage media themselves are being accessed at higher speeds. When the storage medium is a magnetic disk, the time required for seeking to move the disk head to be accessed to the position where data is stored occupies much of the response time.

  Therefore, in order to efficiently access a plurality of data distributed and stored on the magnetic disk, the movement of the disk head is optimized.

<Elevator algorithm>
The algorithm generally called “elevator algorithm” or “SCAN” listed in Non-Patent Document 2 optimizes the order of disk access so that the head can process in one round trip when processing a plurality of access requests. Algorithm. As a result, the total response time of a plurality of accesses is shortened.

<Reduction of response time by data arrangement>
In the case of a storage medium comprising a magnetic disk, the former has a shorter response time than the latter when accessing data stored at a close position on the disk and when accessing data stored at a distant position. have.

  For this reason, it is possible to shorten the response time at the time of access by storing data to be used continuously at a physically close position on the disk.

  As a related technique, for example, Patent Document 1 discloses a rotation angle detection unit that detects a rotation angle of a disk medium and a track number of a track on which a head is located in each disk device that multiplexes and stores data. When the track number detection means and the calculation of the waiting time are instructed, it is necessary until the head moves to the storage area for which the waiting time is calculated using the rotation angle and the track number detected by the track number detection means. Output means for calculating and outputting a waiting time that is a waiting time, and when the controller receives a read request from an external device, the controller responds to the read request using the disk device with the shortest calculated wait time Is disclosed.

  In Patent Document 2, the processing queue for each head is rearranged in order from the physical or logical address for each head corresponding to the position of the head when processing at the present or the nearest past, and the processing request is handled. A configuration is disclosed in which the head closest to the address is processed.

JP-A-7-141770 JP-A-10-27321 Patterson et al. , “A Case for Redundant Arrays of Inexpensive Disks (RAID)”, Proceedings of the 1988 ACM SIGMOD international conference on management data. 109-116 Silberschatz et al. "Operating System Concepts Fifth Edition", Addison-Wesley, pp. 435-437

  The disclosures of Patent Documents 1 and 2 and Non-Patent Documents 1 and 2 are incorporated herein by reference. The following analysis is given by the present invention.

  By utilizing the redundancy among a plurality of storage media by RAID, it is possible to improve storage reliability and speed up access. Among these, the high speed realized by RAID is realized by improving throughput and improving parallel access performance as described in each RAID.

  However, as an index of access performance, in addition to throughput and parallel access performance, response time to access can be cited.

  When the storage medium is a magnetic disk or the like, the response time for access differs depending on the access destination data and the state of the magnetic disk itself.

  And it is difficult to shorten this response time only by using RAID. That is, there is a problem that the response time is not shortened when redundancy is secured by RAID or the like in a storage constituted by a plurality of storage media.

  For example, in a storage to which RAID 1 (mirror) is applied, there are a plurality of access destination storage media from which data can be acquired. When the access times of the storage media are different, in order to shorten the response time, it is necessary to select a storage medium that shortens the access time.

  However, RAID does not select an access destination storage medium that shortens the access time.

  That is, the response time cannot be shortened by RAID as compared with the case of a single storage medium.

  On the other hand, techniques for shortening the response time for the storage medium itself include the above-described elevator seek and data arrangement on a disk.

  However, these techniques perform local optimization for a single disk, that is, a single storage medium. Optimization across multiple storage media is not targeted (out of range). For this reason, it is difficult to shorten the response time for the entire storage including a plurality of storage media using these known techniques.

  Therefore, an object of the present invention is to provide an apparatus, a method, and a program that can shorten the access time in the entire plurality of storage media in response to an access request in a storage including a plurality of storage media that store data in a redundant form. is there.

  In order to solve the above-described problems, the invention disclosed in the present application is generally configured as follows.

In the storage device according to one aspect of the present invention, for each of a plurality of storage media that store data in a redundant form, at least a plurality of access commands that are waiting to be executed with respect to the storage medium are analyzed. Means for scheduling the execution order of access commands in a single storage medium so as to shorten the access time (scheduling means in the storage medium);
Means for selecting a combination of storage media for issuing an access command so as to expedite access completion of the entire storage media from a plurality of the storage media on which the scheduling of a single storage medium is performed (between storage media) Scheduling means).

  In the present invention, the means for scheduling the execution order of the access commands in the single storage medium is based on the access destination address, the access length, and the access time characteristics of the storage medium for each of the plurality of access commands. The access command issuance order is scheduled so as to shorten the access time.

  In the present invention, the means for selecting a combination of storage media issuing the access command predicts an access time when a read access is made to the storage media based on the execution order of the access commands.

  In the present invention, the means for selecting a combination of storage media for issuing the access command may be a storage medium storing the substance of data or redundant data as a storage medium to which the read command is issued so that the access is completed quickly. One of the stored storage media may be selected.

In the present invention, the means for selecting a combination of storage media for issuing the access command stores the entire entity of the data to be read for the plurality of storage media that have undergone the scheduling as a single storage medium. Obtain the access time when the storage medium is read, and the access time when the entire entity of the data to be read is restored using partial data and redundant data read from each of the storage media,
From the access time calculation result, the access command issue destination storage medium may be selected so that the access is completed earlier.

  In the present invention, the means for selecting a combination of storage media to issue the access command does not need to be executed from among access commands whose issue order is scheduled for each storage medium, and the processing time is relatively long. When the access command is detected, the access command is removed from the issuance target to the storage medium, thereby speeding up the access completion.

In another aspect of the present invention, for each of a plurality of storage media for storing data in a redundant form, an access command input to the storage medium is analyzed prior to issuance to the storage medium,
It is an access command that does not need to be executed due to the relationship with the access command executed on another storage medium among the access commands for at least one storage medium, and the processing time is relative among the access commands for the at least one storage medium A command processing unit that performs scheduling to remove an access command that is large in size from being issued to the storage medium and to speed up access completion.

In the present invention, the command processing means schedules an access command issuance order for each of the plurality of access commands based on an access destination address, an access length, and an access time characteristic of the storage medium.
An access command that does not need to be executed and has a relatively long processing time is detected from the access commands whose issue order is scheduled for each storage medium.

  In the present invention, the command processing means may predict an access time when a read access is made to the storage medium based on the access issue order.

In the present invention, the command processing means, for each of the plurality of storage media, when the issuing order of the plurality of access commands to the storage medium is scheduled,
The total access time between the plurality of storage media is compared, and at least one access command among a plurality of access commands to the storage medium having a relatively long total access time is deactivated. Also good.

  In the present invention, when a new access request arrives, the command processing means activates the deactivated access command to be rescheduled, and reschedules the execution order of a plurality of access commands. May be.

  In the present invention, the command processing means may calculate an access time from an internal state of the storage medium and access destination position information.

  In the present invention, regarding the access to the storage medium, the internal state of the storage medium may be calculated from the access destination address in the preceding access in the storage medium or from the address and the access target data.

  In the present invention, the internal state of the storage medium includes head position information.

  In the present invention, the command processing means accesses the storage medium in consideration of the execution status of the command being executed on the storage medium, the end time of the command being executed, and the internal state of the storage medium at the end. The access time required to complete access may be calculated based on the above.

  In the present invention, a plurality of command queues for storing access commands to be input to the storage medium corresponding to the plurality of storage media are provided.

  In the present invention, the command processing means may estimate the total time of the preceding access commands issued to the storage medium when the number of access commands stored in the command queue is equal to or smaller than a predetermined number. If any of the following is satisfied, the access command is issued to the storage medium. Note that the prediction result of the total time of the preceding access command issued to the storage medium is based on the total access time calculated at the scheduled time before issuing the preceding access command to the storage medium. If the storage medium side has a function of scheduling the execution order of access commands issued to the storage medium from the command processing means, the command processing means means A prediction result of the total time of the preceding access command may be acquired from the medium side.

  In the present invention, a plurality of storage media are provided, the access issued to the storage media is managed, and the processing order of the issued access to the storage media within the storage media is predicted, or the storage media Means for determining (scheduling) the access processing order and the access command processing order for the storage medium.

  The present invention further includes means for predicting an access required time when a read (read) access is made to each storage medium based on the predicted or determined access processing order.

  Furthermore, in the present invention, from the information of the access time required for Read, the access time when the storage medium storing the substance of the data (direct data) is read, and some or all of the data to be read are read targets. The access time when the data to be read is restored using the parity read from the storage medium storing the parity for the data is obtained, and the command processing completion time is accelerated from the calculation result of the time required for command processing. For the command, a command issue destination storage medium is selected. In the present invention, when the requested data is read, processing is performed so as to shorten the access time, thereby realizing a reduction in response time during Read access.

The method according to the present invention analyzes a plurality of access commands waiting to be executed on a medium, schedules the execution order of access commands in a single storage medium so as to shorten the access time,
Selecting each combination of storage media to issue an access command from a plurality of the storage media that have been scheduled in a single storage medium so as to expedite access completion across the plurality of storage media. Including.

The method according to the present invention, for each of a plurality of storage media that store data in a redundant form, analyze an access command input to the storage medium prior to issuance to the storage medium,
It is an access command that does not need to be executed due to a relationship with an access command executed on another storage medium among access commands for at least one storage medium, and processing time is relative among access commands for at least one storage medium The above-described steps are included in which an access command that is too large is excluded from the issuance target to the storage medium, and the access completion is scheduled.

The program according to the present invention analyzes, for each of a plurality of storage media that store data in a redundant form, at least a plurality of access commands that are waiting to be executed with respect to the storage medium, and stores the access commands so as to shorten the access time. A process for scheduling the execution order of access commands for a single medium;
A process of selecting a combination of storage media to issue an access command from a plurality of the storage media subjected to the scheduling in a single storage medium so as to expedite access completion in the plurality of storage media as a whole;
Is made up of a program that causes a computer of the storage apparatus to execute.

The program according to the present invention, for each of a plurality of storage media that store data in a redundant form, analyzes an access command input to the storage medium prior to issuance to the storage medium,
It is an access command that does not need to be executed due to the relationship with the access command executed on another storage medium among the access commands for at least one storage medium, and the processing time is relative among the access commands for the at least one storage medium This is a program that causes a computer of a storage apparatus to execute a process for removing an access command that is large in size from being issued to the storage medium and performing a schedule for quick access completion.

  According to the present invention, in a storage including a plurality of storage media that store data in a redundant form, it is possible to reduce the access time in the entire plurality of storage media in response to an access request.

The above-described present invention will be described below with reference to the accompanying drawings in order to explain in more detail. In the present invention, a function for managing the issued access to each storage medium in the storage and predicting the processing order in the storage medium for issued access to the storage medium (referred to as “storage medium processing order prediction function”). ) Is related to the command issued to the storage medium.
・ Access length,
・ Access time characteristics for storage media,
-Physical and logical configuration information of the storage medium, or
A scheduling algorithm used inside the storage medium,
The execution order of the access commands optimized for shortening the access time in the storage medium is predicted based on the information regarding.

  In the present invention, based on the command execution order predicted by the storage medium processing order prediction function, a function for predicting the required access time when the storage medium is accessed (referred to as “command processing completion time prediction function”). ) If the access time calculation command and the preceding command affect the access time of the storage medium for the storage medium, the access destination address is also passed to the access command for the preceding command. As a result. Alternatively, information on the physical or logical state in the storage medium storing the access destination data may be acquired based on the access destination address, and the time required for accessing the storage medium may be calculated. Alternatively, the physical or logical state in the storage medium storing the data from the access destination address when accessing the storage medium storing the data in the past or the history of the real time required for the access. And the time required for access to the storage medium may be calculated.

  In the present invention, from the access time information required for Read, the access time when the storage medium storing the data entity is directly read and the parity for the Read target data for some or all of the Read target data are calculated. The function for obtaining the access time when the data to be read is restored using the parity read from the stored storage medium (referred to as “redundancy utilization access method enumeration / access time calculation function”) is a command processing completion time prediction function. To calculate the time required to access the data to be accessed or the redundant data for the data to be accessed for each storage medium constituting the storage, and when the data to be directly accessed is accessed or the data to be directly accessed Stored in a storage medium that does not store By using the redundant data is, in the case of restoring the access target data is calculated respectively access duration to the storage medium in the entire storage.

  A function for selecting a command issue destination storage medium for each command (“access destination” so that the command processing completion time is shortened from the calculation result of the time required for processing all commands required for access completion. The storage medium selection function ”) completes command processing from the time required to access the storage medium in the entire storage for each access destination storage medium selection method calculated using the redundancy use access method enumeration / access time calculation function. In order to process each access request to the storage so as to shorten the time, the processing order of the storage medium to be actually accessed and the command queue is determined.

  In storage with redundancy of data in a plurality of storage media, redundant data such as parity is used when a failure occurs in the storage medium, and the data to be directly accessed is stored when no failure occurs Access to. Therefore, for example, when it is desired to complete the access process early, even if the process can be completed earlier if the data is restored from the redundant data such as parity, the redundant data is not used.

On the other hand, according to the present invention, during storage access processing,
-A storage medium that stores direct data (data entity);
A storage medium storing redundant data;
Is used selectively. For this reason, for example, when it is desired to complete the access process early, it is possible to select whether to directly access the storage medium storing the data or to restore the data to be accessed from the redundant data. The access processing is performed by selecting a combination of media, and the average access time is shortened. Embodiments of the present invention will be described below with reference to the drawings.

  FIG. 1 is a diagram showing the configuration of an embodiment of the present invention. Referring to FIG. 1, in the present embodiment, the configuration of the entire system includes one or more hosts 1 that access data, a storage 2 that is composed of a plurality of storage media, and a network 3 that connects the aforementioned 1 and 2. .

  The network 3 may have a form in which the host 1 and the storage 2 are directly connected as well as a form in which a plurality of hosts 1 can be connected to the storage 2.

  FIG. 2 is a diagram illustrating a configuration example of the storage 2 in FIG. Referring to FIG. 2, the storage 2 includes a command processing unit 21, a data storage unit 22, a command processing work memory 23, a storage configuration information storage unit 24, and a storage medium access history storage unit 25. .

  The data storage unit 22 includes a plurality of storage media 221 and a plurality of command queues 222 respectively corresponding to the plurality of storage media 221.

  The command processing work memory 23 includes an access request management table 231 for managing access requests from the host and a command management table 232 for managing commands scheduled to be issued to each storage medium 221. .

  The command processing unit 21 interprets the access request received from the host (1 in FIG. 1), issues a command to the storage medium 221 of the data storage unit 22, and performs actual access processing for the entire storage 2. The command processing unit 21 generally operates as follows.

  The command processing unit 21 generates a command for each storage medium 221 necessary for processing the access request.

  Next, the command processing unit 21 performs scheduling so that the access processing for the storage medium 221 is completed earlier in order to process the access request for the commands generated from the plurality of access requests, For each access request, the command issue destination storage medium 221 is selected.

  Then, the command processing unit 21 issues a command to the storage medium 221 in accordance with the scheduled execution order, and performs processing for transmitting a result for the Read access request when the command is completed.

  When the command is Read and the access target data stored in the access destination storage medium 221 includes redundant data, the command processing unit 21 performs a process of restoring the access target data from the redundant data.

  After the access target data is obtained, the command processing unit 21 returns the result to the access request source host 1.

  The data storage unit 22 holds all data stored in the storage 2 and performs data read and update (write) in accordance with an instruction from the command processing unit 21.

  The command processing work memory 23 stores data read from the storage medium 221 and data requested to be written by the host in accordance with instructions from the command processing unit 21.

  In the example illustrated in FIG. 2, the command processing work memory 23 includes an access request management table 231 and a command management table 232. The command queue 222 included in the data storage unit 22 may be stored in the command processing work memory 23.

The storage configuration information storage unit 24
The correspondence relationship between the storage area of the entire storage 2 and the plurality of storage media 221 constituting the data storage unit 22 that is the substance of the storage area is stored.

The storage configuration information storage unit 24
Information relating to response time characteristics for access to an access position, such as physical and logical configuration information affecting the calculation of the access time of each storage medium 221 is stored. As the access response time characteristic information of the storage medium, for example, when the storage medium 221 is a hard disk,
・ The distribution of blocks on the disk platter,
・ Number of disk platters
-Correspondence between seek distance of disk head and seek time,
Stores the information.

  In the example shown in FIG. 2, the storage configuration information storage unit 24 exists independently in the storage 2, but may be stored in the command processing work memory 23.

  The storage medium access history storage unit 25 stores the address and access length of a command issued and completed for the storage medium 221 included in the data storage unit 22 as a history.

  The command processing unit 21 updates the history in the storage medium access history storage unit 25 every time the storage medium 221 completes command processing.

  FIG. 3 is a diagram showing an example of the configuration information of the access request management table 231 shown in FIG. The access request management table 231 holds information associated with each access request from the host 1 (see FIG. 1).

As information accompanying the access request,
A unique ID for each access request given when the command processing unit 21 receives an access request from the host;
An access type (R / W) indicating whether the access request is Read or Write;
The address of the first block in the address space of the entire storage 2 to be accessed;
・ Access length,
A data storage destination address on the command processing work memory 23;
The requesting host of the access request;
A corresponding command ID that is an ID indicating a command issued to the storage medium 221 in order to process an access request;
Record.

  When a plurality of commands issued to the storage medium 221 to process one access request are issued, a plurality of corresponding command IDs are registered for one access request ID.

The data storage destination address in FIG.
In the case of Read, the address of the area (buffer area) temporarily reserved for storing data to be transmitted as a result to the host 1 (see FIG. 1) by the command processing unit 21 is indicated.
In the case of Write, the address of the area (buffer area) for holding the data requested to be written from the host 1 (see FIG. 1) in the command processing unit 21 is indicated.

  The access request management table 231 is referred to by the command processing unit 21 when information accompanying each access request from the host 1 (see FIG. 1) is necessary.

  FIG. 4 is a diagram showing an example of configuration information of the command management table 232 of FIG. The command management table 232 holds information associated with a command issued to each storage medium 221 in order to process each access request from the host 1. The command management table 232 is individually reserved for each storage medium 221.

As information accompanying the command for the storage medium 221,
-A unique ID for each command,
An access type (represented as R / W in FIG. 4) indicating whether the command is Read or Write;
The address of the first block in the storage medium 221 to be accessed;
・ Access length,
A data storage destination address on the command processing work memory 23;
A corresponding access request ID assigned to the access request on the access request management table 231 that is the cause of command issuance,
-Command ID in the storage medium,
An issue time indicating the time when the command is issued to the storage medium 221;
• Record the command execution status.

Among the information, the data storage destination address is an access request,
In the case of Read, it indicates the address of the area reserved for storing the data read from the storage medium 221.
In the case of Write, it indicates the address of an area that holds data to be written to the storage medium 221.

  The corresponding access request ID is a unique ID (see FIG. 3) for each access request given to the access request on the access request management table 231 that causes the command issuance.

  The in-storage medium command ID is an ID given as tag information for command execution management from the storage medium side when the command processing unit 21 issues a command to the storage medium 221.

  The execution state is information for managing the state of the command, and takes the states of standby, deletion, issue, and completion.

  Waiting is a state in which a command is currently waiting in the command queue 222 for issuance.

  Deletion is a state in which a command currently exists in the command queue 222 but is not actually issued.

  The issue indicates that the command has already been issued in the storage medium 221.

  Completion indicates a state in which processing is completed upon receipt of a command completion notification.

  The command management table 232 is referred to by the command processing unit 21 when the information accompanying the command scheduled to be issued or issued for each storage medium 221 is required.

  FIG. 5 is a diagram showing information constituting the command queue 222 of FIG. The command queue 222 holds the issue order of commands issued to the storage medium 221. That is, the command IDs (command IDs in the storage medium in FIG. 3) are held in the order of commands issued to the storage medium 221. The command queue 222 is secured for each storage medium 221 individually. Note that the command may be fetched from the command queue 222 when the access to the storage medium is executed (command execution status is issued).

  6 to 11 are flowcharts for explaining the operation of the present embodiment described with reference to FIGS. Although not particularly limited, in the following, as a plurality of storage media, parity is stored in a corresponding region of at least one other storage medium with respect to data (an entity of data) in a region of the plurality of storage media, and a plurality of storage media A redundant storage medium that can recover data from the remaining storage medium even when a failure occurs in one of the storage media, or a redundant configuration that stores data redundantly among a plurality of storage media will be described as an example. It should be noted that the function / processing of the command processing unit 21 that executes the processing procedure shown in FIGS. 6 to 11 may be realized by a program executed on the computer of the storage 2.

  First, with reference to the flowchart of FIG. 6, an operation when an access request is made from the host 1 to the storage 2 in this embodiment will be described. Each time an access request is made from the host 1 to the storage 2, the entire system shown in FIGS. 1 and 2 executes the following operation.

  The command processing unit 21 interprets the received access request from the host 1 to the storage 2 and registers the access request in the access request management table 231 (S11 in FIG. 6).

  When the access request received from the host 1 is Write (Yes branch of S12 in FIG. 6), the command processing unit 21 receives the data to be written sent from the host 1 following the access request, and executes the command processing work. The data is stored in an empty area on the memory 23, and the data storage address of the access request on the access request management table 231 is updated.

  When the command processing unit 21 receives the write data from the host 1 and completes storage in the free area on the command processing work memory 23, the command processing unit 21 notifies the host 1 that the processing of the access request is completed (FIG. 6). S13).

  When the access request received from the host 1 is Read (No branch of S12 in FIG. 6), the command processing unit 21 executes Step S14 without executing Step S13.

  The command processing unit 21 generates a command for each storage medium 221 necessary for processing the access request from the host 1. More specifically, a command for performing read / write of data to an area that is an actual data storage destination of an access destination address of a storage 2 designated by the host 1 and an area that stores redundant data It is generated for the storage medium 221.

  When the access request is Read, a Read command for reading the data entity is generated for the storage medium 221 storing the data entity requested by the host 1. At the same time, a Read command for reading the parity data is also generated for the storage medium 221 having the parity data corresponding to the actual data requested by the host 1.

  When the access request is “Write”, a Write command for storing data is generated for all the storage media 221 having an area serving as a data storage destination entity.

  When the parity data is updated in accordance with the write access request process, a Read command for reading out data that is insufficient to regenerate the parity data stored in the storage medium 221, and the storage medium 221. A Write command for updating the parity data stored in is also generated.

  Note that there is no data to be written by the Write command for updating the parity data at this time.

  The command processing unit 21 acquires information on the correspondence between the address space of the entire storage 2 and the storage area of each storage medium 221 from the configuration information of the storage 2 stored in the storage configuration information storage unit 24.

  The command processing unit 21 registers the command generated for the storage medium 221 in the command management table 232. The command processing unit 21 adds the command scheduled to be issued to the storage medium to the end of the command queue 222 (S14 in FIG. 6).

  The command processing unit 21 schedules a command issuance order that is registered on the command queue 222 and is scheduled to be issued to the storage medium 221, and selects a command to be actually issued.

  The command processing unit 21 performs scheduling of commands input to the command queue 222 and selection of commands to be actually issued (S15 in FIG. 6).

More specifically, the command processing unit 21
The command issue order is changed so that processing for all access requests registered in the access request management table 231 is completed earlier.

  Further, when the command processing unit 21 detects an unnecessary command among the commands input to the command queue 222, the command processing unit 21 changes the execution state on the command management table 232 to “deleted”.

  When the command processing unit 21 is issued to the storage medium 221 and the actual access processing to the storage medium 221 is not completed (waiting for execution) is equal to or less than a predetermined number (threshold), The command processing unit 21 performs command issuance processing from the head of the command queue 222 until the number of commands being issued to the storage medium 221 reaches a predetermined number in advance, and the execution status on the command management table 232 is “issued”. The command issuance time column of the corresponding command ID in the command management table 232 is set to the time when the command is issued.

  Alternatively, the command processing unit 21 determines that the number of commands for which the actual access processing to the storage medium 221 has not been completed (waiting for execution) is equal to or smaller than a predetermined number (threshold), or On the other hand, when the predicted result of the total time of the preceding access commands that have been issued satisfies any of the predetermined time or less, the access command may be issued to the storage medium. Note that the prediction result of the total time of the preceding access command issued to the storage medium is based on the total access time calculated at the scheduled time before issuing the preceding access command to the storage medium. It may be calculated by the unit 21, or the storage medium has a function of scheduling the execution order of access commands issued from the command processing unit 21 (for example, the storage medium implements the above-described elevator algorithm). In some cases, the command processing unit 21 may acquire the prediction result of the total time of the preceding access command from the storage medium side.

  The command processing unit 21 changes the execution state column of the corresponding command ID on the command management table 232 to “completed” for the command whose execution state is “deleted”, and does not perform the issue process (FIG. 6). S16).

  The processes from the above steps S11 to S16 are performed, and the process when an access request is made from the host 1 to the storage 2 is completed.

  Next, with reference to FIG. 7, a processing procedure when the command processing unit 21 receives a completion notification for a command being issued in the storage medium 221 in the present embodiment will be described.

  Upon receiving the command completion notification (including the storage medium command ID) from the storage medium 221, the command processing unit 21 refers to the command management table 232, refers to the R / W column from the storage medium command ID, The type of the completed command is determined (S21 in FIG. 7).

  If the completed command is Read (Yes branch of S22 in FIG. 7), the command processing unit 21 performs processing when the Read command is completed (S23 in FIG. 7). Details of step S23 will be described later with reference to FIG.

  If the completed command is Write (No branch in S22 of FIG. 7), the command processing unit 21 performs processing upon completion of the Write command (S24 of FIG. 7). Details of step S24 will be described later with reference to FIG.

  The command processing unit 21 refers to the command issuance time column of the command management table 232 and determines whether the completed command is the earliest issued command being issued. That is, the command processing unit 21 determines whether the command issue order is different from the command completion order.

  If the command issuance order is different from the command completion order, the command processing unit 21 performs, for example, the same processing as step S15 in FIG. 6 (S25 in FIG. 7). That is, the command processing unit 21 reschedules the command input to the command queue 222.

  After execution of step S25, the command processing unit 21 executes the same processing as step S16 in FIG. 6 (S26 in FIG. 7). That is, the command processing unit 21 issues a command when the number of commands that are issued to the storage medium 221 and the actual access processing to the storage medium 221 is not completed (waiting for execution) is equal to or less than a predetermined number. The processing unit 21 performs command issuance processing from the head of the command queue 222 until the number of commands being issued to the storage medium 221 reaches a predetermined number in advance, and sets the execution state on the command management table 232 to “issue”. The command issuance time column of the corresponding command ID in the command management table 232 is set to the time when the command is issued.

  The processes from step S21 to S26 are performed, and the process when the command processing unit 21 receives a completion notification for the command being issued in the storage medium 221 ends.

  FIG. 8 is a detailed flowchart of step S23 in FIG. Next, a more detailed processing procedure in step S23 of FIG. 7 in the command processing unit 21 will be described.

  The command processing unit 21 changes the execution status column of the completed command in the command management table 232 to “completed”. Further, the storage medium access history storage unit 25 is accessed to update the history information of the command executed by the storage medium 221 (S2301 in FIG. 8).

  The command processing unit 21 stores the data read from the storage medium 221 in an area on the command processing work memory 23 indicated by the data storage destination address specified by the command management table 232 (S2302 in FIG. 8).

  The command processing unit 21 refers to the access request management table 231 and the command management table 232, and confirms whether all the commands necessary for processing the access request that is the source of the command have been completed. For example, if the original access request from the host 1 is Read and the storage medium 221 has a RAID 1 configuration, the access request can be processed as long as data reading from at least one storage medium 221 is completed. It is. The access request can be processed only by the read result of the completed command, and it is substantially equivalent that all necessary commands are completed.

  On the other hand, if the original access request from the host 1 is also Read and the storage medium 221 has a RAID 5 configuration consisting of four storage media, all storage media storing the substance of the data to be read are stored. The processing can be performed by either completing the access to 221 or restoring the requested data using any three parity data.

  Therefore, in this case, it is possible to check whether all necessary commands have been completed by checking whether all data has been read or whether data has been read from any three storage media. It is. The command processing unit 21 acquires the configuration information of the storage medium 221 from the storage configuration information storage unit 24 (S2303 in FIG. 8).

  As a result of the confirmation in step S2303, if all the commands necessary for the access request are not completed (No branch in S2304 in FIG. 8), the command processing unit 21 ends the process in step S23.

  As a result of step S2303, when all the commands necessary for the access request have been completed (Yes branch of S2304 in FIG. 8), the command processing unit 21 executes the determination of next step S2305.

  When the access request from which the command is issued is Write (Yes branch in S2305 in FIG. 8), the command processing unit 21 reads the Read result on the command processing work memory 23 and the write for which Write is requested from the host 1. Parity data is recalculated from the data (S2311). That is, the command processing unit 21 refers to the data storage destination address of the command management table 232 from the command processing work memory 23 with respect to the result of the Read command executed to process the access request that has caused the generation of the command. read out. From the read data and the data instructed to be written by the host 1, recalculation processing of the updated parity data generated in response to the write access request that is the command generation source is performed, and the result is stored in the command processing work memory 23. Store in free space.

  Then, the command processing unit 21 updates the command management table 232 so that the recalculated parity data is the write target data, and inputs the write command for the storage medium 221 storing the parity to the command queue 222, Data is written (S2312). That is, the command processing unit 21 updates the data storage destination address for the parity data update command in the command management table 232 so that the data calculated in step S2311 is used in the command generated for parity data update. Then, the process of step S23 is completed.

  When the access request from which the command is issued is Read (No branch in S2305 in FIG. 8), the command processing unit 21 executes the determination in step S2306. In the command necessary for processing the access request that caused the generation of the command, the parity is not read, that is, the command is a Read for the requested data entity (in S2306 of FIG. 8). No branch), the command processing unit 21 executes Step S2308.

  In the command necessary for processing the access request that is the source of the command, when the parity is read (Yes branch of S2306 in FIG. 8), the command processing unit 21 determines the access request of the command generation source. The result of reading the command executed to process the command is read from the command processing work memory 23 with reference to the data storage destination address of the command management table 232. Data requested from the host 1 is calculated using the read parity data (S2307 in FIG. 8). The method of restoring data from parity differs depending on the method of configuring the storage medium 221, for example, when the configuration of the storage medium 221 is RAID 5, it can be restored by an XOR operation between data in a redundant relationship. Therefore, the configuration information of the storage medium 221 is acquired from the storage configuration information storage unit 24.

  The command processing unit 21 converts and configures the data requested or read-read from the host 1 into a form that can be transmitted to the host as it is (S2308 in FIG. 8).

  The command processing unit 21 transmits the result obtained in step S2308 to the access request source host 1 (S2309 in FIG. 8).

  The command processing unit 21 deletes from the command management table 232 the data of all registered commands that are derived from the same access request that is being processed and that has been processed. In addition, the command processing unit 21 deletes the data related to the access request under attention whose processing has been completed from the access request management table 231.

  The command processing unit 21 releases an area on the command processing work memory used by the command deleted from the command management table 232 and the access request deleted from the access request management table 231, and ends the process of step S23 (FIG. 8 S2310).

  The processes from step S2301 to S2312 are performed, and the command processing unit 21 ends the process in step S23.

  Next, a more detailed processing procedure in step S24 of FIG. 7 of the command processing unit 21 will be described with reference to FIG.

  The command processing unit 21 executes the same process as in step S2301 (S241 in FIG. 9).

  Subsequently, the command processing unit 21 executes the same processing as step S2303 (S242 in FIG. 9).

  As a result of the confirmation in step S242, if all the commands necessary for the access request are not completed (No branch in S243 in FIG. 9), the process in step S24 is terminated.

  As a result of the confirmation in step S242, if all the commands necessary for the access request have been completed (Yes branch in S243), the command processing unit 21 executes the same process as in step S2310 and ends the process in step S24. (S244 in FIG. 9).

  The processes from step S241 to S244 are performed, and the process in step S24 of the command processing unit 21 is completed.

  Next, with reference to FIG. 10, a more detailed processing procedure of command issue order scheduling and command selection to be actually issued by the command processing unit 21 in steps S15 and S25 of FIGS. 6 and 7 will be described. .

  The command processing unit 21 arranges the issue order of commands existing in the command queue 222 according to the characteristics of the corresponding storage medium 221.

  For example, when the storage medium 221 is composed of a hard disk, it is possible to arrange the command issuing order by a method called SCAN or elevator algorithm described in Non-Patent Document 2. By arranging the commands, the moving distance of the head of the hard disk is shortened, so that the entire access time when a plurality of commands are issued for the hard disk as the storage medium 221 can be shortened.

  However, when the command queue 222 includes a Write command, the command processing unit 21 determines whether a command generated later accesses the same area as the Write command. When the command generated later and the area accessed by the Write command overlap, if the order of the commands is changed, the processing of the access request becomes inconsistent. For this reason, the command processing unit 21 prohibits the order of these commands from being changed.

  Further, the command processing unit 21 needs information on the type and access characteristics of the storage medium 221 in order to perform command alignment processing according to the access characteristics and physical characteristics of the storage medium 221. The command processing unit 21 acquires these pieces of information from the storage configuration information storage unit 24 (S31 in FIG. 10).

  After the command alignment process for the command queue 222, the command processing unit 21 calculates an estimated time until all the processes are completed when the commands belonging to the command queue 222 are executed in order for each storage medium 221. To do.

  For example, when the storage medium 221 is composed of a hard disk, the time required for processing each command is the head seek time, the disk rotation waiting time required for the block accessed by the command to reach the head position, and the actual access It can be calculated from the total of the three types of processing times for accessing the target block. The time required for processing all commands is represented by the total processing time of each command.

  In order for the command processing unit 21 to calculate the processing completion time of the storage medium 221, information on access characteristics and access time is required. The command processing unit 21 receives these information from the storage configuration information storage unit 24. Obtain (S32 in FIG. 10).

  The command processing unit 21 pays attention to, for example, a storage medium predicted to be the latest processing completion time from the storage medium 221 using the result of the access completion time prediction in step S32.

The command processing unit 21 performs the processing for the storage medium predicted to have the latest processing completion time.
Of the commands to be issued included in the command queue 222,
It is possible to process an access request from the host 1 without issuing a command, and
A command that causes a delay in access processing time to the storage medium is attempted to be changed to a deleted state (S33 in FIG. 10).

  If the command processing unit 21 succeeds in changing the command issued to the storage medium 221 to the deleted state as a result of the processing in step S33 (No branch in S34), the command processing unit 21 returns to step S31 and continues the processing.

  As a result of the processing in step S33, if the command issued to the storage medium 221 has failed to be changed to the deletion state (Yes branch in S34), the command processing unit 21 determines that the processing completion time is the latest in step S33. The command deletion processing that is limited to the predicted storage medium 221 is also performed for the other storage medium 221.

  The command processing unit 21 attempts to change a command that can process an access request from the host 1 to a deletion state without issuing a command among commands to be issued included in the command queue 222.

  Since the criterion for determining whether or not deletion is possible differs depending on the configuration of the storage medium 221 and the storage form of the redundant data, the command processing unit 21 acquires the configuration information of the storage medium 221 from the storage configuration information storage unit 24.

  The following are examples of commands to be deleted.

  If the original access request from the host 1 is Read and the storage medium 221 has a RAID 1 configuration, the access request can be processed if data can be read from one or more storage media 221. Therefore, when a command for this access request exists for a plurality of storage media 221, it is possible to delete other commands except for one command.

  On the other hand, when the original access request from the host 1 is also Read and the storage medium 221 has a RAID 5 configuration including four storage media 221, all the storage media 221 storing the substance of the data to be read are stored. Can be processed by either completing the access to or restoring the requested data using any three parity data. For this reason, when a command for this access request is scheduled to be issued to all four storage media 221, even if one command is in a deleted state, data to be read can be generated from parity. . Therefore, it is possible to delete a command included in any one of the four command queues 222 for the four storage media (S35 in FIG. 10).

  The processes from step S31 to S35 described above are performed, and the command processing unit 21 performs the scheduling of command issue order and the process of selecting a command to be actually issued in steps S15 and S25.

  Next, a more detailed processing procedure in step S33 of FIG. 10 by the command processing unit 21 will be described with reference to FIGS. FIG. 12B is a correspondence list between commands and arm positions, and FIG. 12A is a schematic diagram for explaining a seek operation when the commands in FIG. 12B are arranged by an elevator algorithm.

  The command processing unit 21 pays attention to a command that causes the longest processing time from a command scheduled to be issued to the storage medium 221 included in the command queue 222, and determines whether the command can be deleted in a subsequent step. Next, the command to be judged is selected.

  For example, when the storage medium 221 is configured by a hard disk, the command that causes the longest processing time is that the head position of the disk at the time of access is determined as long as the command execution order is determined by the elevator algorithm described in Non-Patent Document 2. This command is farthest from the initial head position.

  In the example of FIG. 12, since the seek distance is most reduced when the command C is deleted from the commands A, B, and C, the command C is selected as a determination target whether the command C can be deleted.

  In this way, by deleting the command that accesses the block farthest from the initial position of the head, it is possible to reduce the seek distance and seek time required to execute all the commands included in the command queue 222. .

  Further, when the command selection process for the determination target is repeatedly executed in the command processing unit 21, the command that has been the determination target is not selected, but the command that causes the next largest processing time is selected. For example, when the storage medium 221 is composed of a hard disk, a command that is farthest from the initial position of the head other than the command already determined is selected.

  In addition, when the storage medium 221 is a hard disk, in order to process all commands when the command for accessing the block farthest from the initial position of the head and the command for accessing the block farthest from the head cannot be deleted. It is impossible to reduce the seek distance required for the operation.

  For this reason, when these two commands cannot be deleted, step S33 may be terminated at that time (S331 in FIG. 11).

  If there is no command to be determined in step S331 (Yes branch in S332 in FIG. 11), the command processing unit 21 ends the process in step S33.

  If there is a command to be determined in step S331 (No branch in S332 in FIG. 11), the command processing unit 21 determines whether or not the command focused in step S331 can be deleted.

  Whether or not the command can be deleted can be determined by examining whether or not processing for an access request from the host 1 that is the source of the focused command is possible when the focused command is not executed. . Since the criterion for determining whether or not deletion is possible differs depending on the configuration of the storage medium 221 and the storage form of the redundant data, the command processing unit 21 acquires the configuration information of the storage medium 221 from the storage configuration information storage unit 24.

  An example of specific determination as to whether a command can be deleted is as follows. For example, if the original access request from the host 1 that is the generation source of the command to be determined is Read and the storage medium 221 has a RAID 1 configuration, if data can be read from one or more storage media 221 The access request can be processed.

  When the command processing unit 21 makes a determination, the command management table 232 is referred to. If a command other than the deletion determination target is already in the deletion state with respect to the original access request, the determination target command is deleted. Impossible.

  Conversely, when one or more commands other than the deletion determination target exist in the command queue 222 other than in the deletion state, the processing for the access request can be executed regardless of the determination target command, and the command can be deleted ( S333 in FIG. 11).

  If it is determined in step S333 that the command cannot be deleted (No branch in S334 of FIG. 11), the command processing unit 21 returns to step S331 to perform processing.

  If it is determined in step S333 that the command can be deleted (Yes branch in S334 of FIG. 11), the command processing unit 21 ends the process of step S33.

  As a first operation example from steps S11 to S16, steps S21 to S26, and steps S31 to S35, the storage medium 221 included in the data storage unit 22 of the storage 2 is composed of two hard disks. An operation when a read access request is made from the host 1 to the storage 2 in the case of adopting the RAID 1 configuration will be described with reference to FIGS. 2 to 11, 13, and 14.

  As a hard disk constituting the storage medium 221 of the data storage unit 22, a disk having the block arrangement shown in FIG. 13 and the following characteristics is used as an example. FIG. 13A schematically shows the track arrangement on the platter (tracks # 0 to # 1001), and FIG. 13B schematically shows the block arrangement on the track (blocks # 0 to # 999).

  First, it does not have a function to respond to an inquiry about the response time for the access destination address. The number of platters and heads on the hard disk is one.

  The response time to access is equal to the sum of the seek time of the hard disk, the rotation waiting time of the hard disk, and the access time, and the seek distance of the hard disk is proportional to the seek time.

  The seek time from the outermost periphery to the innermost periphery of the hard disk is 10 ms.

  The total of the hard disk rotation waiting time and access time is 4 ms.

  The number of blocks per track in which the blocks are arranged on the circumference is 1000 for all tracks.

  Also, the arrangement interval of the tracks on the platter is equal, and the total number of tracks on the platter is 10001.

  Assume that numbers are assigned to the tracks on the platter starting from number 0 in order from the outermost periphery.

  Address numbers are assigned to the blocks on the hard disk in order from the outermost track.

  For example, the outermost track includes blocks from addresses 0 to 999, and the second track from the outermost track includes blocks from addresses 1000 to 1999.

  Further, the two hard disks constituting the storage medium 221 are referred to as a disk A and a disk B, respectively.

  The correspondence between the address space of the entire storage 2 and the address space of the two hard disks is as shown in FIG. Since the two hard disks have a RAID 1 configuration, they hold common data, and the address space of the entire storage 2 matches the storage area of each of the two disks. That is, the data of the block on an address in the address space of the entire storage 2 matches the data of the block existing at the same address of each disk.

  With respect to the two hard disks A and B constituting the storage medium 221, the initial position of the head of each disk, that is, the immediately preceding access position, is that the disk A is the outermost track (0th track) and the disk B is the innermost track. Suppose that the track is on the track No. 10000 (track no.

  Further, it is assumed that one instruction is simultaneously input to each disk.

  In this state, it is assumed that the host 1 issues a read request for an access length of one block at an interval of 1 ms for the data of the blocks having the addresses 1000000, 8000000, and 4000000 in the storage 2. Read access requests from the host 1 are (A), (B), and (C) in order.

First, the operation when the storage 2 receives the access request (A) will be described.
The command is interpreted by the command processing unit 21 of the storage 2.

  Information of the access type, access head block, access length, and request source host of the access request (A) is registered in the access request management table 231.

  Since the access request received by the command processing unit 21 is Read, the command processing unit 21 next generates a command for the storage medium 221 for processing the access request.

  The correspondence between the address space of the storage 2 and the storage areas of the disks A and B constituting the storage medium 221 is as shown in FIG.

  Therefore, access to the address 1000000 of the storage 2 is realized by the Read command for the block of the address 1000000 on the disks A and B.

  The command processing unit 21 relates to the Read command for the disks A and B, the access type, the access head block, the access length, the storage destination address of the data that is an area secured on the command processing work memory 23, the command generation source Are registered in the command management table 232 corresponding to the disks A and B, respectively, and the ID indicating the access request and the execution state (“standby” state). At the same time, the Read command is input to the end of the command queue 222 corresponding to each of the disks A and B.

  The command processing unit 21 schedules commands included in the command queue 222 corresponding to each of the disks A and B after the command generation is completed. Using the elevator algorithm disclosed in Non-Patent Document 2 for each of the disks A and B, the commands are aligned so that the seek time is optimized.

  Since each command queue 222 has only one command, in each command queue 222, the command derived from the access request (A) comes first in the execution order.

  The command processing unit 21 calculates a time until processing of commands belonging to each command queue 222 is completed. That is, the command processing unit 21 calculates the time until each of the disks A and B completes the Read command derived from the access request (A).

  According to the physical and logical configuration information of the hard disk indicating the access time characteristics with respect to the access destination, the block with the address 1000000 as the access destination exists on the 1001st track, that is, the 1000th track from the outermost periphery. At this time, the head positions of the disks A and B exist on the outermost track of the disk A and the innermost track of the disk B from the head position at the time of the last access. In order for the head of the disk disk to seek on the track where the access destination block exists, seeking for 1000 tracks on the disk A and 10,000-1000 = 9000 tracks on the disk B is necessary.

In other words, of the two disks A and B, the seek time of the head and the seek distance are proportional, so the seek time to the data to be accessed is
With disk A, 10 ms × 1000/10000 = 1 ms,
With disk B, 10 ms x 9000/10000 = 9 ms
It is. Since the total of the rotation waiting time and the access time is 4 ms, the access completion time is
With disk A, 1 + 4 = 5 ms,
With disk B, 9 + 4 = 13 ms,
It becomes.

  Therefore, reading the access target block from the disk A shortens the access processing time.

Here, the command processing unit 21 pays attention to the command queue 222 on the disk B side having a long access processing time,
-Execution is unnecessary and
・ Large processing time,
Delete the command.

  The command queue 222 on the disk B side stores only commands derived from the access request (A).

  The command processing unit 21 recognizes from the information in the storage configuration information storage unit 24 that the disks A and B are in a mirror relationship with each other, and either of the disks A and B is required to complete the processing of the access request (A). Only access to one is required.

  Therefore, it can be seen that the command derived from the access request (A) can be deleted from the command queue 222 for the disk B.

  The command processing unit 21 updates the execution state information in the command management table 232 in order to place the command in the deletion state.

  Since the command processing unit 21 succeeded in deleting the command, the command processing unit 21 tries to delete another command. However, the command on the disk B side is already deleted in a state where only the command derived from the access request (A) is issued. Therefore, no more commands can be deleted.

  Therefore, the command processing unit 21 deletes the command on the disk B side, and the scheduling processing by the command processing unit 21 is completed.

  Since there is no command issued to each of the disks A and B when the scheduling is completed, the command processing unit 21 newly performs a command issuing process for the disks A and B.

  The command to be issued is the first command among the commands registered in the command queue 222. A command derived from the access request (A) is issued to the disk A as it is because it is in a “standby” state. At this time, the command processing unit 21 records the current time and the command ID in the storage medium notified from the disk A in the issue time column of the command management table 232, and changes the execution status column to “issue”. To do.

  However, since the command derived from the access request (A) is in the “deleted” state for the disk B, the command is not issued, and the command processing unit 21 sets the execution status column of the command management table 232 to “ Change to Completed state.

  When 1 ms elapses from the processing, the host 1 issues an access request (B) to the storage 2, that is, a Read request for the address 8000000.

  In the command processing unit 21, the procedure from receiving the access request (B) to scheduling is the same as the processing when the access request (A) is received. In the following, the process of scheduling commands included in the command queue 222 corresponding to each of the disks A and B will be described.

  The command processing unit 21 arranges commands for the disks A and B in the same manner using an elevator algorithm.

  Since a command already derived from the access request (A) does not exist on the command queue 222, only one command exists in each command queue 222. Therefore, in each command queue 222, the command derived from the access request (B) comes to the top of the execution order.

  The command processing unit 21 calculates a time until processing of commands belonging to each command queue 222 is completed. That is, the command processing unit 21 calculates the time until each of the disks A and B completes the Read command derived from the access request (B).

  In the disk A, a command derived from the access request (A) is being processed, and the processing completion time is 1 + 4 = 5 ms from the time of issue.

  Since 1 ms has passed since issuance, disk A requires an extra 5-1 = 4 ms until the completion of the preceding command.

  In the disk A, after the preceding command is completed, the head moves to the position of the 1000th track.

On the other hand, since the disk B has not yet executed the command, the head exists at the position of 10000th. Therefore, the time required for seeking is
With disk A, 10 ms × (8000−1000) / 10000 = 7 ms,
With disk B, 10 ms x (10000-8000) / 10000 = 2 ms
It is.

The total of the rotation waiting time and the access time is 4 ms, and the access completion time is
With disk A, 4 + 7 + 4 = 15 ms,
With disk B, 2 + 4 = 6ms
It becomes.

  Therefore, reading the access target block from the disk B shortens the access processing time.

The command processing unit 21 pays attention to the command queue 222 on the disk A side having a long access processing time,
・ Execution is unnecessary, and
• Delete commands that cause significant processing time.

  Thereafter, similarly to the processing when the access request (A) is received, since the command for the disk A can be deleted, the command processing unit 21 changes the execution status column of the command management table 232 to “deleted”. .

  The command processing unit 21 completes the scheduling process. At this time, there is no command being executed by the disk B, and the command for the disk B is in the “standby” state. A command derived from the access request (B) is issued. At this time, the command processing unit 21 records the current time and the command ID in the storage medium notified from the disk B as the issue time of the command management table 232, and changes the execution status column to “issue”.

  However, no command is issued to the disk A because the command derived from the access request (A) is being executed.

  When a further 1 ms elapses from the above processing, the host 1 issues an access request (C) to the storage 2, that is, a Read request for the address 4000000.

  The procedure from when the command processing unit 21 receives the access request (C) to the scheduling is the same as the processing when the access requests (A) and (B) are received.

  In the following, the process of scheduling commands included in the command queue 222 corresponding to each of the disks A and B will be described.

  Similarly, commands are arranged for each of the disks A and B using an elevator algorithm.

  As for the disk B, since the command derived from the access requests (A) and (B) does not exist on the command queue 222, there is only one command in the command queue 222. Therefore, in the command queue 222 for the disk B, the command derived from the access request (C) comes to the top of the execution order.

  On the other hand, for the disk A, the Read access command for the address 1000000 derived from the access request (A) is being executed.

  The completion time of the command processing is 1 + 4 = 5 ms from the time of issue.

  Since 2 ms has passed since the issue, Disk A requires 5-2 = 3 ms to complete the preceding command.

  Therefore, the Read access command for the address 8000000 derived from the subsequent access request (B) is left in the command queue 222 for the disk A without being executed.

  On the other hand, a Read command for the address 4000000 derived from the access request (C) is added. When command alignment by the elevator algorithm is executed for the command queue 222 for the disk A, the position of the head of the disk A at the time when the head command of the command queue 222 is executed is the access position of the command currently being executed, that is, the address 1000000. As a result, the Read command for the address 4000000 comes first in the execution order.

  That is, the command derived from the later access request (C) is scheduled before the command derived from the preceding access request (B).

  The command processing unit 21 calculates a time until processing of commands belonging to each command queue 222 is completed.

  The disk A is processing a command derived from the access request (A), and the processing completion time is 1 + 4 = 5 ms from the time of issue.

  Since 2 ms has passed since the issuance, the disk A requires an extra 5-2 = 3 ms until the preceding command is completed.

  In the disk A, after the preceding command is completed, the head moves to the position of the track of 1000000/100 = 1000.

  The disk B is processing a command derived from the access request (B), and the processing completion time is 2 + 4 = 6 ms from the time of issue.

  Since 2 ms has passed since the issuance, Disk 6-1 requires an extra 6-1 = 5 ms until the preceding command is completed.

  In the disk A, after the preceding command is completed, the head moves to the track position of 8000000/100 = 8000.

The time required for seeking is 10 ms × ((4000−1000) + (8000−4000)) / 10000 = 7 ms for two commands derived from the access requests (B) and (C) in the disk A.
In the disk B, 10 ms × (8000−4000) / 10000 = 4 ms.

  The total of the rotation waiting time and the access time is 4 ms, and this time is added to the command processing completion time by the number of commands existing in the command queue 222.

Command processing completion time is
With disk A, 3 + 7 + 4 × 2 = 18 ms,
With disk B, 5 + 4 + 4 = 13ms
It becomes.

  Therefore, the processing time of the access request is shortened when the access target block is read from the disk B.

Here, paying attention to the command queue 222 on the disk A side having a long access processing time,
・ Execution is unnecessary, and
• Delete commands that cause significant processing time.

  When the commands included in the command queue 222 are arranged by the elevator algorithm, the command derived from the access request (B) is executed next to the command derived from the access request (C).

  At this time, the read command for the address 8000000 derived from the access request (B) causes the increase in the total seek distance of all commands belonging to the command queue 222.

  The command processing unit 21 confirms whether or not the command can be deleted.

  First, a command derived from the access request (B) has already been issued to the disk B. Here, since the disks A and B are mirrors of each other, it is understood that the Read command for the address 8000000 derived from the access request (B) can be deleted on the disk A side. The command processing unit 21 updates the execution state information in the command management table 232 in order to place the command in the deletion state.

In the state where the Read command for the address 8000000 derived from the access request (B) is deleted on the disk A side, the total command completion time of the disk A is recalculated.
3 + 3 + 4 = 10ms
It becomes.

Therefore, this time, paying attention to the command queue on the disk B side,
・ No need to execute, and ・ Large processing time
Delete the command.

  At this time, only the Read command for the address 4000000 derived from the access request (C) exists in the command queue 222 of the disk B.

  Thereafter, the command derived from the access request (C) for the disk B can be deleted similarly to the processing when the access request (A) is received, so the command on the disk B side is deleted, and the command processing unit 21 Changes the execution status column of the command management table 232 to “deleted”.

  At this time, the command remaining in the command queue 222 is only the Read command for the address 4000000 derived from the access request (C) of the disk A.

  Since the command derived from the access request (C) has already been deleted in the command queue 222 of the disk B, it is indispensable for executing the access request (C). Therefore, the command processing unit 21 completes the scheduling process.

  If there is no command being executed by the disk after scheduling, the command processing unit 21 issues a new command. However, since the command derived from the access request (A) is being executed for the disk A, and the command derived from the access request (B) is being executed for the disk B, the command processing unit 21 No new command is issued.

  When a further 3 ms elapses from the above process, the execution of the command derived from the access request (A) executed on the disk A is completed. The disk A notifies the command processing unit 21 of the completion of the command with the command ID in the storage medium.

  The command processing unit 21 searches the command management table 232 for a command corresponding to the command ID in the storage medium, and determines that the command is Read.

  The command processing unit 21 updates the command management table 232 and changes the execution status column of the completed command to “completed”. In addition, the command processing unit 21 accesses the storage medium access history storage unit 25 and updates the information on the previous access to the disk A with the start address being 1000000 and the access length being 1 block.

  Subsequently, the command processing unit 21 stores the result read upon completion of the command at the data storage destination address designated by the command management table 232 on the command processing work memory 23.

  The command processing unit 21 confirms whether all the instructions necessary for processing the access request (A) that is the generation source of the completed command are completed.

  Since the disks A and B have a mirror configuration, the processing of the access request (A) is completed when the result of Read access is obtained from one of the disks. Further, the access request (A) that is the generation source of the completed command is Read access, and the access destination information is not the parity but the substance of the data requested from the host 1.

  Since the correspondence between the disk A and the address space of the entire storage 2 is one-to-one, the command processing unit 21 sends the read result of the completed command to the host 1 as it is without processing the access request (A). Returns as a result.

  Finally, the command processing unit 21 releases the memory reserved for the completed command processing on the command processing work memory 23, the entry of the access request (A) from the access request management table 231, and the access request ( Delete all command entries originating from A).

  Thereafter, the command processing unit 21 performs processing in the same procedure when the command derived from the access requests (B) and (C) is completed.

Finally,
In the disk A, a command derived from the access request (A), followed by a command derived from the access request (C),
In the disk B, a command derived from the access request (B) is executed.

  In the description of the processing procedure, the read access from the host 1 to the storage 2 has been described. However, the write access is performed in the same manner except for the write-only processing.

  Subsequently, as a second operation example from steps S11 to S16, steps S21 to S26, and steps S31 to S35, the data storage unit 22 of the storage 2 is composed of four hard disks, and these disks adopt a RAID 5 configuration. The operation when the host 1 performs a read access to the storage 2 will be described with reference to FIGS. 2 to 13 and FIG. FIG. 15 is a diagram for explaining an example in which the present invention is applied to RAID5.

  The hard disk constituting the storage medium 221 included in the data storage unit 22 is the same as that in the above embodiment. Further, the four hard disks constituting the storage medium 221 are referred to as disks A, B, C, and D, respectively.

  The correspondence between the address space of the entire storage 2 and the address space of the four hard disks is as shown in FIG.

  Data is continuously recorded in 1000 blocks on each disk. When four disks are divided into areas of 1000 blocks from the beginning of the address, areas existing at the same address of each disk are represented as a set having a redundant relationship with each other. Since the four disks have a RAID 5 configuration, the data entity is stored in an area for three of the set. Parity information of the data of the three areas is recorded in the remaining area.

For example, regarding the address space of the entire storage 2,
-Blocks with addresses 0 to 999 are recorded at addresses 0 to 999 of disk A,
-Blocks with addresses 1000 to 1999 are recorded at addresses 0 to 999 on disk B,
-Blocks with addresses 2000 to 2999 are recorded at addresses 0 to 999 on disk C,
The parity of the block at addresses 0 to 2999 is recorded at addresses 0 to 999 on disk D.

In RAID5, the disk for storing parity is not constant.
For example,
-The parity of the block of addresses 3000 to 5999 is stored in the block of addresses 1000 to 1999 on disk C,
The parity of the block at addresses 6000 to 8999 is stored in the block at addresses 2000 to 2999 on disk B,
The parity of the block with addresses 9000 to 11999 is stored in the block with addresses 3000 to 3999 on disk A.

  Assume that the initial state of each disk for the four hard disks constituting the storage medium 221 is as follows.

  Disks A, B, C, and D are in a command processing state, and command processing for each disk is completed after 5 ms.

The position of the head after command processing is completed is as follows:
Disc A is on track number 1000
Disc B is on track number 2000,
Disc C is on track number 3000,
Disc D is on track No. 4000,
Suppose that

  Further, it is assumed that one instruction is simultaneously input to each disk.

  In this state, the host 1 issues a read request for an access length of one block in order to the data of the block with the address 0, 15000000 in the storage 2. Read access requests from the host 1 are (A) and (B) in order.

  First, the operation when the storage 2 receives the access request (A) will be described.

  The procedure from when the command processing unit 21 receives the access request (A) to scheduling is the same as the processing in the operation example in the case of RAID1.

  However, in RAID1, each disk stores a copy of data, whereas in RAID5, there are disks having the data of the access destination and disks having the parity for the data of the access destination.

  It can be seen that the access request (A) is Read for address 0 in the address space of the entire storage 2, that is, Read access for data stored in the block of address 0 of disk A in the example shown in FIG.

  On the other hand, the parity data corresponding to the block at address 0 on disk A is the block at address 0 on disks B, C, and D.

  Therefore, the Read command of the block with the address 0 is generated for the disks A, B, C, and D, and is input to the command queue 222 for each disk.

  In addition to the read for the disk A, the read is the parity data for the data to be accessed.

  In the following, the process of scheduling commands included in the command queue 222 corresponding to each of the disks A, B, C, and D will be described.

  For each of disks A, B, C, and D, commands are aligned using an elevator algorithm.

  For the disks A, B, C, and D, no command exists on the command queue 222 when the command processing unit 21 receives the access request (A). For this reason, for each disk, only the command derived from the access request (A) exists in the command queue 222, and the command derived from the access request (A) comes first in the execution order.

  The command processing unit 21 calculates a time until processing of commands belonging to each command queue 222 is completed.

  The disks A, B, C, and D are both processing commands, and the processing completion time is 5 ms after the time when the access request (A) is issued.

  The command processing unit 21 calculates the time required for each disk to seek in the processing of the command derived from the access request (A).

  The commands for each disk derived from the access request (A) are all Read commands for the block of address 0.

Therefore,
The seek time of the disk A is 10 ms × (1000-0) / 10000 = 1 ms.

Similarly,
For disk B, 10 ms × (2000-0) / 10000 = 2 ms,
In the disk C, 10 ms × (3000-0) / 10000 = 3 ms,
For disk D, 10 ms × (4000−0) / 10000 = 4 ms
It is.

  The total of the rotation waiting time and the access time is 4 ms, and this time is added to the command processing completion time by the number of commands existing in the command queue 222.

Therefore, the command processing completion time is
With disk A, 5 + 1 + 4 = 10 ms,
With disk B, 5 + 2 + 4 = 11 ms,
With disk C, 5 + 3 + 4 = 12 ms,
With disk D, 5 + 4 + 4 = 13 ms
It is.

  Since only the commands derived from the access request (A) are input to the command queues 222 of the disks A, B, C, and D, the processing completion time of all commands existing in the command queue 222 is It is equal to the command processing completion time.

Here, paying attention to the command queue 222 of the disk D having the longest access processing completion time,
• Delete commands that do not need to be executed and that cause large processing time.

  In the command queue 222 of the disk D, only commands derived from the access request (A) are stored.

  Here, it can be seen from the information in the storage configuration information storage unit 24 that each disk as the storage medium 221 has a relationship of having parity information with each other as shown in FIG.

  The command input to the disk D is a Read command for parity data with respect to data for which the host 1 has requested read access.

  Therefore, the command processing unit 21 knows that the command derived from the access request (A) can be deleted from the command queue 222 for the disk D. Therefore, the command processing unit 21 updates the execution state information in the command management table 232 in order to place the command in the deletion state.

  The command processing unit 21 again arranges the commands in the command queue 222 and calculates the command processing completion time for each disk.

  As described above, only the command derived from the access request (A) is input to the disk D having the latest processing completion time, and the access processing completion time of the disk C becomes the longest because the command is in a deleted state.

Similar to the processing performed on the disk D, the command processing unit 21 pays attention to the command queue 222 of the disk C, and
・ Execution is unnecessary, and
• Delete commands that cause significant processing time.

  Since the command derived from the access request (A) input to the disk C is also a Read command for the parity data for the data requested for read access from the host 1, the command processing unit 21 deletes the command in the same manner as the disk D. I do.

  The command processing unit 21 again calculates the command processing completion time for each disk. The command derived from the access request (A) input to the disk B is also deleted for the same reason as the disks C and D.

  However, among the disks A, B, C, and D, the command input to the disk A, which is the only disk that remains in the command queue 222, is the data requested to be read from the host 1 by the access request (A). Access to the entity.

  Since the Read command for parity data that already exists for another disk is in a deleted state, it is impossible to delete the Read command from the command queue 222 of disk A. Therefore, the command processing unit 21 completes the scheduling process.

  The command processing unit 21 issues a new command when there is no command being executed by the disk after scheduling. In this case, the command is being executed by the disks A, B, C, and D. No new command is issued.

  Subsequently, an access request (B) is sent from the host 1 to the storage 2, that is, a Read request for the address 15000000.

  First, the operation when the storage 2 receives the access request (B) will be described.

  The procedure from when the command processing unit 21 receives the access request (B) to scheduling is the same as the processing for the access request (A).

  In the following, scheduling processing of commands included in the command queue 222 corresponding to each of the disks A, B, C, and D will be described.

  The command processing unit 21 arranges commands for each of the disks A, B, C, and D using an elevator algorithm.

  For all of the disks B, C, and D, there is no command on the command queue 222 when the command processing unit 21 receives the access request (B). Therefore, for the disks B, C, and D, only the command derived from the access request (B) exists in the command queue 222, and the command derived from the access request (B) comes first in the execution order.

  On the other hand, a command derived from the access request (A) is input to the command queue 222 in the disk A. This command is a Read command for the block of address 0.

  On the other hand, the command derived from the access request (B) is that the address space of the entire storage 2 and the address space of each disk have the correspondence shown in FIG. This is a Read command for this block.

  The command currently being processed on disk A is for the 1000th track. Therefore, when the commands are arranged on the command queue 222 of the disk A using the elevator algorithm, the commands of the access request (A) and the access request (B) are arranged in this order.

  The command processing unit 21 calculates a time until processing of commands belonging to each command queue 222 is completed.

  Commands are being processed for all of the disks A, B, C, and D, and the processing completion time is 5 ms after the time when the access request (A) is issued.

  Next, the command processing unit 21 calculates the time required for each disk to seek.

  The commands for each disk derived from the access request (A) and the access request (B) are Read commands for the block at address 0 (0th track) and address 5000000 (5000th track), respectively.

  The command processing unit 21 first calculates the seek time for the disk A.

Since the disk A processes in the order of the command derived from the access request (A) and the command derived from the access request (B), the seek time is
10 ms × ((1000-0) + (5000-0)) / 10000 = 6 ms
It is.

  On the other hand, in the disks B, C, and D, since the command derived from the access request (A) is in a deleted state, only the command derived from the access request (B) is processed.

So seek time is
In disk B, 10 ms × (5000−2000) / 10000 = 3 ms,
For disk C, 10 ms × (5000-3000) / 10000 = 2 ms,
For disk D, 10 ms × (5000-4000) / 10000 = 1 ms
It is.

  The total of the rotation waiting time per command and the access time is 4 ms, and the command processing completion time is added to the number of commands existing in the command queue 222.

Therefore, all command processing completion times existing in the command queue 222 are
With disk A, 5 + 6 + 4 × 2 = 19 ms,
With disk B, 5 + 3 + 4 = 12 ms,
With disk C, 5 + 2 + 4 = 11 ms,
With disk D, 5 + 1 + 4 = 10 ms
It is.

Here, the command processing unit 21 pays attention to the command queue 222 of the disk A having the longest access processing completion time.
・ Execution is unnecessary, and
Remove commands that cause significant processing time.

  The command queue 222 of the disk A stores commands derived from the access request (A) and the access request (B).

  When the command processing unit 21 arranges the commands included in the command queue 222 by the elevator algorithm, the command derived from the access request (B) is executed next to the command derived from the access request (A).

  At this time, the read command for the address 5000000 derived from the access request (B) causes the increase in the total seek distance of all commands belonging to the command queue 222.

  The command processing unit 21 confirms whether this command can be deleted.

  Each disk constituting the storage medium 221 has a RAID 5 relationship, and it is possible to obtain a Read result by restoring from the parity without accessing one disk.

  Since the command derived from the access request (B) other than the disk A is not deleted, the command derived from the access request (B) of the disk A can be deleted. The command processing unit 21 updates the information in the execution state column of the command management table 232 in order to place this command in the deletion state.

  The command processing unit 21 arranges the commands in the command queue 222 again, and calculates the command processing completion time for each disk.

  As for the disk A that has the latest processing completion time, the command processing completion time is 5 + 1 + 4 = 10 ms from the calculation result at the time of the access request (A) because the command derived from the access request (B) is in the deleted state. . Therefore, 12 ms of the access completion time of the disk B is the longest.

The command processing unit 21 pays attention to the command queue 222 of the disk B in the same manner as the process performed on the disk A.
・ Execution is unnecessary, and
• Delete commands that cause significant processing time.

  Only commands derived from the access request (B) are input to the disk B. Here, it can be seen that the substance of the data requested to be read by the access request (B) exists in the disk A from the relationship between the address space of the entire storage 2 and the address space of each disk shown in FIG.

  Since the command derived from the access request (B) in the disk A has already been deleted, and since each disk has a RAID 5 relationship, all the data to obtain the target data without using the actual data is obtained. Parity, that is, the result of the Read command for all disks B, C, and D is required. Therefore, it is impossible to delete the Read command from the command queue 222 of the disk B. Since the command processing unit 21 has failed to delete the command from the disk B, the scheduling processing is completed.

  The command processing unit 21 issues a new command when there is no command being executed by the disk after scheduling. However, since the command is currently being executed in all the disks A, B, C, and D, no new command is issued.

Finally,
In disk A, a command derived from the access request (A) is executed,
In the disks B, C, and D, a command derived from the access request (B) is executed.

  The procedure for subsequent command completion processing in the disk is the same as that in the case where the storage medium 221 is composed of a RAID 1 disk.

  In the above description of the processing procedure, the description has focused on the Read access from the host 1 to the storage 2, but the same processing is performed for Write access except for Write-only processing.

  Within the scope of the entire disclosure (including claims) of the present invention, the examples and the examples can be changed and adjusted based on the basic technical concept. Various combinations and selections of various disclosed elements are possible within the scope of the claims of the present invention. That is, the present invention of course includes various variations and modifications that could be made by those skilled in the art according to the entire disclosure including the claims and the technical idea.

It is a figure which shows the whole system configuration | structure in one Example of this invention. It is a figure which shows the structure of the storage in one Example of this invention. It is a figure which shows the access request management table for managing the information accompanying the access request with respect to the storage from the host in one Example of this invention. It is a figure which shows the command management table for managing the information accompanying the command which is issuing with respect to each storage medium in one Example of this invention, or is issuing. It is a figure which shows the command queue which shows the execution order of the command which is scheduled to issue with respect to each storage medium in one Example of this invention. It is the flowchart which showed the process sequence at the time of the storage in one Example of this invention receiving the access request from the host. It is the flowchart which showed the process sequence when the command processing part in one Example of this invention is notified of the completion of command processing from the storage medium in a storage. It is the flowchart which showed the process sequence of the command processing part when the storage medium in the storage in one Example of this invention completed the process of Read command. It is the flowchart which showed the process sequence of the command processing part when the storage medium in the storage in one Example of this invention completed the process of Write command. It is the flowchart which showed the processing procedure of the scheduling with respect to the command which is due to issue to each storage medium in the storage of the command processing part in one Example of this invention, and issuing command selection. It is the flowchart which showed the process sequence which selects and deletes the command which causes increase in processing time from the command which is due to issue to each storage medium in the storage of the command processing part in one Example of this invention. It is the figure which showed the example which selects the command of the object which performs deletion determination from the command which is due to issue to each storage medium in the storage in one Example of this invention. It is a figure which shows arrangement | positioning of the track | truck and block on the platter of the hard disk in the Example of this invention. It is a figure which shows the correspondence of the block address of each hard disk of RAID1 structure and the address space of the storage 2 in 1st Example of this invention. It is a figure which shows the correspondence of the block address of each hard disk of RAID5 structure and the address space of the storage 2 in 2nd Example of this invention.

Explanation of symbols

1 Host 2 Storage 3 Network 21 Command processing unit 22 Data storage unit 221 Storage medium 222 Command queue 23 Command processing work memory 231 Access request management table 232 Command management table 24 Storage configuration information storage unit 25 Storage medium access history storage unit 31 In the hard disk Platter 32 outermost track 33 on platter innermost track 34 on platter track 35 on platter block on track

Claims (42)

For each of a plurality of storage media that store data in a redundant form, analyze at least a plurality of access commands waiting to be executed with respect to the storage medium, and change the access command of the storage medium alone so as to shorten the access time. Means for scheduling execution order;
Means for selecting a combination of storage media to issue an access command so as to expedite access completion of the entire storage media from the plurality of storage media subjected to the scheduling in a single storage medium;
A storage apparatus comprising: Means for scheduling the execution order of access commands in the single storage medium,
For each of a plurality of access commands, the execution order of the access commands is scheduled so as to shorten the access time of the single storage medium based on the access destination address, the access length, and the access time characteristics of the storage medium. The storage device according to claim 1. Means for selecting a combination of storage media for issuing the access command,
3. The storage according to claim 1, wherein an access time when a read access is made to the storage medium is predicted based on an execution order of access commands scheduled in units of the storage medium. apparatus. Means for selecting a combination of storage media for issuing the access command,
4. The storage medium storing the substance of data or the storage medium storing redundant data is selected as the storage medium to which the read access command is issued so that the access is completed earlier. The storage device according to any one of the above. Means for selecting a combination of storage media for issuing the access command,
The access time when reading the storage medium storing the entire entity of the data to be read out from the plurality of storage media subjected to the scheduling in a single storage medium, and portions read from the plurality of storage media, respectively Obtain the access time when restoring the entire data to be read using data and redundant data,
The storage apparatus according to any one of claims 1 to 3, wherein a storage medium to which an access command is issued is selected so that an access is completed earlier from a calculation result of the access time. Means for selecting a combination of storage media for issuing the access command,
When an access command that does not need to be executed and has a relatively long processing time is detected from access commands that are scheduled to be issued in units of storage media, the access command is issued from the subject to be issued to the storage medium. The storage apparatus according to any one of claims 1 to 5, wherein the access completion is accelerated by removing the storage device. For each of a plurality of storage media that store data in a redundant form, an access command input to the storage medium is analyzed prior to issuance to the storage medium,
It is an access command that does not need to be executed due to the relationship with the access command executed on another storage medium among the access commands for at least one storage medium, and the processing time is relative among the access commands for the at least one storage medium A storage apparatus comprising: command processing means for performing scheduling to remove an access command that is large in size from being issued to the storage medium and to speed up access completion. The command processing means schedules the execution order of the access commands for each of the plurality of access commands based on the access destination address, the access length, and the access time characteristics of the storage medium, in units of storage media,
8. The storage apparatus according to claim 7, wherein an access command that does not need to be executed and has a relatively long processing time is detected from access commands scheduled in units of storage media.   The storage apparatus according to claim 7 or 8, wherein the command processing means predicts an access time when a read access is made to the storage medium based on the access command execution order. The command processing means, for each of the plurality of storage media, when the execution order of the plurality of access commands is scheduled for the storage medium,
Comparing the total access time between the plurality of storage media and deactivating at least one access command among the plurality of access commands to the storage medium having a relatively long total access time The storage apparatus according to claim 7, wherein the storage apparatus is excluded from the issue target.   The command processing means activates the deactivated access command to be rescheduled when a new access request arrives, and reschedules the execution order of a plurality of access commands. The storage device according to claim 10.   The storage apparatus according to claim 9, wherein the command processing means calculates an access time from an internal state of the storage medium and location information of an access destination.   The command processing means calculates the internal state of the storage medium from the access destination address in the preceding access in the storage medium or the address and the access target data with respect to the access to the storage medium. The storage device according to claim 9.   14. The storage apparatus according to claim 12, wherein the internal state of the storage medium includes head position information.   The command processing means considers the execution status of the command being executed on the storage medium, the end time of the command being executed, and the internal state of the storage medium at the end, and completes the access based on the access command to the storage medium The storage apparatus according to claim 9, wherein an access time required until the time is calculated.   The storage apparatus according to any one of claims 1 to 15, further comprising a plurality of command queues that temporarily store access commands to the storage medium corresponding to the plurality of storage media.   The command processing means determines in advance whether the number of access commands stored in the command queue is equal to or less than a predetermined number, or a prediction result of the total time of preceding access commands issued to a storage medium. The storage apparatus according to claim 16, wherein an access command is issued to the storage medium if it is less than or equal to a predetermined time.   When the issuing order of the access commands to the storage medium and the completion order of the access commands issued to the storage medium in the storage medium are different, the command processing means retransmits the access commands in the command queue. The storage apparatus according to claim 16, wherein scheduling is performed.   19. The access command according to claim 1, wherein the storage device analyzes an access request input from a host and is generated as an access command for the plurality of storage media. The storage device described. For each of a plurality of storage media that store data in a redundant form, analyze at least a plurality of access commands waiting to be executed with respect to the storage medium, and change the access command of the storage medium alone so as to shorten the access time. Schedule execution order,
Selecting a combination of storage media to issue an access command from a plurality of the storage media subjected to the scheduling in a single storage medium so as to expedite access completion in the plurality of storage media as a whole. Storage access method.   For each of a plurality of access commands, the execution order of the access commands is scheduled so as to shorten the access time of the single storage medium based on the access destination address, the access length, and the access time characteristics of the storage medium. The storage access method according to claim 20.   The storage access method according to claim 20 or 21, wherein an access time when a read access is made to the storage medium is predicted based on the access command execution order.   23. The storage medium storing the substance of data or the storage medium storing redundant data is selected as the storage medium to which the read command is issued so that the access is completed earlier. The storage access method according to claim 1. For a plurality of the storage media subjected to the scheduling in a single storage medium,
When the entire data of the data to be read is restored using the access time when the storage medium storing the entire data of the data to be read is read, and partial data and redundant data respectively read from the plurality of storage media Seeking access time,
The storage access method according to any one of claims 20 to 22, wherein a storage medium to which an access command is issued is selected so that access is completed earlier from a calculation result of the access time.   When an access command that does not need to be executed from access commands scheduled to be issued in units of storage media and has a relatively long processing time is detected, the access command is detected from the issue target to the storage medium. The storage access method according to any one of claims 20 to 24, wherein the access completion is accelerated by removing the access. For each of a plurality of storage media that store data in a redundant form, an access command input to the storage medium is analyzed prior to issuance to the storage medium,
It is an access command that does not need to be executed due to the relationship with the access command executed on another storage medium among the access commands for at least one storage medium, and the processing time is relative among the access commands for the at least one storage medium A storage access method characterized in that an access command that is large is excluded from being issued to the storage medium, and scheduling is performed to accelerate access completion. For each of a plurality of access commands, the access command execution order is scheduled for each storage medium based on the access destination address, access length, and access time characteristics of the storage medium.
27. The storage access according to claim 26, wherein an access command that does not need to be executed and has a relatively long processing time is detected from among access commands whose issue order is scheduled for each storage medium. Method.   28. The storage access method according to claim 26, wherein an access time when a read access is made to the storage medium is predicted based on the access command execution order. For each of the plurality of storage media, when the order of issuing the plurality of access commands to the storage medium is scheduled,
Comparing the total access time between the plurality of storage media, and deactivating at least one access command among the plurality of access commands to the storage medium having a relatively long total access time. 27. A storage access method according to claim 20 or 26.   The command processing means activates the deactivated access command to be rescheduled when a new access request arrives, and reschedules the execution order of a plurality of access commands. 30. The storage access method according to claim 29.   The storage access method according to claim 22 or 28, wherein an access time is calculated from an internal state of the storage medium and location information of an access destination.   23. The access to the storage medium, wherein the internal state of the storage medium is calculated from the access destination address in the preceding access in the storage medium or from the address and the access target data. 28. The storage access method according to 28.   The storage access method according to claim 31 or 32, wherein the internal state of the storage medium includes head position information.   In consideration of the execution status of the command being executed on the storage medium, the end time of the command being executed, and the internal state of the storage medium at the end, the access time required to complete access based on the access command to the storage medium is calculated. The storage access method according to claim 22 or 28, wherein the storage access method is calculated.   35. The access command according to claim 20, wherein the storage device analyzes an access request input from a host, and is generated as an access command for the plurality of storage media. The storage access method described. For each of a plurality of storage media that store data in a redundant form, analyze at least a plurality of access commands waiting to be executed with respect to the storage medium, and change the access command of the storage medium alone so as to shorten the access time. A process for scheduling the execution order;
A process of selecting a combination of storage media to issue an access command from a plurality of the storage media subjected to the scheduling in a single storage medium so as to expedite access completion in the plurality of storage media as a whole;
For causing a computer of the storage device to execute the program.   For each of a plurality of access commands, a process for scheduling the execution order of the access commands so as to shorten the access time of the single storage medium based on the access destination address, the access length, and the access time characteristics of the storage medium. 37. The program according to claim 36, which is executed.   37. The program according to claim 36, which causes the computer to execute a process of predicting an access time when a read access is made to the storage medium based on the access command execution order.   37. The program according to claim 36, which causes the computer to execute a process of selecting a storage medium storing the substance of data or a storage medium storing redundant data as a storage medium to which a read command is issued so that access is completed earlier. . For each of a plurality of storage media that store data in a redundant form, an access command input to the storage medium is analyzed prior to issuance to the storage medium,
It is an access command that does not need to be executed due to the relationship with the access command executed on another storage medium among the access commands for at least one storage medium, and the processing time is relative among the access commands for the at least one storage medium A program that causes a computer of a storage apparatus to execute a process for performing scheduling to remove an access command that is large in size from being issued to the storage medium and to speed up access completion. For each of a plurality of access commands, the access command issuance order is scheduled for each storage medium based on the access destination address, access length, and access time characteristics of the storage medium.
41. The program according to claim 40, wherein the computer executes a process of detecting an access command that does not need to be executed and has a relatively long processing time from among access commands whose issue order is scheduled for each storage medium. .   42. The access command according to claim 36, wherein the storage device analyzes an access request input from a host and is generated as an access command for the plurality of storage media. The listed program.

Images