JP2002175156A - Data storage array device and data access method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To solve such a problem that the guarantee of response delay time has been difficult when applying a command cueing to data storage devices in a data storage array device. SOLUTION: This data storage array device has the plural data storage devices Disk 1, 2, 3, 4 and 5 having the redundant constitution, and a controller for controlling the data storage devices. The controller successively and multiply issues plural access requests A, B and C to the data storage devices as a command group, successively resets a timer with every command group, guarantees the response delay time by control of the timer, and performs real-time picture data processing.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a redundant data storage array device for storing continuous media data such as video information.

[0002]

2. Description of the Related Art With the development of multimedia technology in recent years, opportunities for recording continuous media data such as digitized moving image and audio data in a data storage device (random access device) such as a hard disk drive are increasing. . In the case of video data, for example, MPE
If G, the transfer rate of one stream is about 1.5 Mbps to 30 Mbps. In addition, since continuous media data has a large data size, it is becoming common to form an array using a plurality of random access devices and to make the external virtual recognition as a single large-capacity device.

In accessing continuous media data, continuous access to data in a data storage device is required, and even if a part of the data storage device fails, the entire server system does not stop. Reliability is required. Further, if a plurality of data storage devices have a redundant configuration, even if a part of the data storage device fails, correct data can be restored by the redundant data.

In a data storage array device having a redundant configuration, when data is written from a host, redundant data is added to data from the host and then written to the data storage device. In the case of a data read request from the host,
When all the data storage devices are normal, the read data is returned to the host as it is. When one data storage device fails due to the redundant configuration, the data of the failed portion can be reconstructed from the remaining data from the normal data storage device and the redundant data and returned to the host.

It is possible to respond to a partial failure of the data storage device by the data storage array device having such a redundant configuration. If a certain delay or more occurs, the video is interrupted, so that a large delay for an access request from the host may not be allowed.

In order to guarantee a delay in response time in a data storage array device for a server system corresponding to such a continuous medium, there is a method of guaranteeing response time using a timer.

As a conventional processing method using a timer at the time of reading, Japanese Patent Laid-Open No. 2-81123, "Parallel Data Transfer Method and Apparatus", detects a response delay due to a failure in a data storage device, and regenerates data from redundant data. A method of configuring is disclosed. As a method of detecting the delay, in the case of a configuration of N data disks and one redundant disk, a timer is started after a read response from any one of the disks is received in response to a read request to N + 1. , If there is no response of the last one after a predetermined time has elapsed,
With data from normal data disks and redundant disks,
Recover data from a disk whose response has been delayed. As a result, the delay after the disk starts responding can be kept within a certain range.

Another conventional example is disclosed in Japanese Patent Application Laid-Open No. 9-69.
No. 027 “Recording system for multimedia”
A method of delay guarantee by a timeout table for guaranteeing response time to a plurality of users is disclosed. The timeout table is set for each user,
Set a timer when receiving a request from the host,
Issues a read / write request to the disk. If there is a disk whose response is delayed, a timeout process is executed. If the necessary data access is not completed even after the time stored in the timeout table elapses, the redundant data is accessed and the data is restored, and a response is returned to the host. After responding to the host, the process returns to a state of waiting for the next request from the host, and when a new request comes, the above-described request to the disk and timeout processing are repeated. With the timer value of this timeout table, a delay time from when a request comes to when a response is returned can be guaranteed.

[0009]

However, the conventional data storage array device has the following problems.

[0010] The timer processing as in the conventional device has the following problems. An ordinary disk device requires an access wait time such as a seek time of a magnetic head and a disk rotation wait time. However, in order to optimize disk access and increase total transfer efficiency, a command (read or write) by command queuing is used. Instruction)
Is performed.

[0011] By optimizing the execution of a plurality of commands inside the disk device by command queuing,
It is possible to greatly improve the transfer rate as compared with processing commands separately without queuing. However, in the conventional example, since a timer process in which a request and a response are paired is performed, a plurality of requests from the host cannot be queued. For this reason, in the disk device, the processing is performed separately for each access request, so that the performance of the disk device cannot be exhibited, and data cannot be supplied at a high transfer rate as a data storage array device.

In response to the host issuing a plurality of requests one after another, if the data storage array device also supports the command queuing, the access to the disk device is queued to operate the disk device. Although it is possible to improve the efficiency, it is difficult to guarantee the response delay time by the conventional method using the timer only by performing queuing to the disk device. For example, if the access time for the maximum number of requests that can be accepted at the same time is guaranteed, a long timer is used. As a result, the response delay time increases, and real-time performance is not guaranteed. Furthermore, when the number of users is small, the number of requests that are being accepted at the same time is small, and wasteful waiting time occurs.

In order to solve the above-mentioned problems, the present invention provides a data storage array device and the like which guarantees the access response performance of continuous media data and multiplexes a plurality of access requests to achieve high transfer efficiency. The purpose is to:

[0014]

Means for Solving the Problems The first invention (claim 1)
Is a redundant configuration, a plurality of data storage devices, and a data storage array device including a controller that controls the plurality of data storage devices, the controller, for the plurality of data storage devices, Instruction issuing means for sequentially issuing an access instruction group composed of a plurality of divided access instructions, response detecting means for detecting a response to an instruction executed in the data storage device, and timer means for counting time Monitoring a response group to the predetermined one access command group, and when responses from all the data storage devices are completed within a time set by the timer means, at the time of the completion, the predetermined Completes the response process to one of the access instruction groups, and ends the time set by the timer means. At this time, if the responses from all the data storage devices have not been completed, at the time when the timer expires, only the response from the already completed data storage device is used, and the predetermined one access command is used. And a completion processing means for completing a response process for the group.

According to a second aspect of the present invention (corresponding to claim 2), for a predetermined one access instruction group, the start of the timer is performed by setting a response group to a predetermined one access instruction group immediately before that. The data storage array device according to the first aspect of the present invention is started at a timing at which the response ends the latest.

According to a third aspect of the present invention (corresponding to claim 3), for a predetermined one access instruction group, the start of the timer is performed by setting a response group to a predetermined one access instruction group immediately before that. The data storage array device according to the first aspect of the present invention is started at the timing when the response is completed earliest.

According to a fourth aspect of the present invention (corresponding to claim 4), for a predetermined one access instruction group, the start of the timer is performed by setting a response group to a predetermined one access instruction group immediately before that. The data storage array device according to the first aspect of the present invention, wherein the data storage array device is started at an intermediate timing between the timing at which the response ends early and the timing at which the response ends.

According to a fifth aspect of the present invention (corresponding to claim 5), for a predetermined one access instruction group, execution of the timer is started earliest among the predetermined one access instruction group. 1 is a data storage array device according to the first aspect of the present invention, which is started at a predetermined timing.

According to a sixth aspect of the present invention (corresponding to claim 6), the timer is started for the predetermined one access instruction group at the latest start of the predetermined one access instruction group. 1 is a data storage array device according to the first aspect of the present invention, which is started at a predetermined timing.

According to a seventh aspect of the present invention (corresponding to claim 7), for a predetermined one access instruction group, execution of the timer is started earliest among the predetermined one access instruction group. The data storage array device according to the first aspect of the present invention, wherein the data storage array device is started at a timing intermediate between the timing at which execution is started and the timing at which execution is started latest.

According to an eighth aspect of the present invention (corresponding to claim 8), the completion processing means is configured to determine whether or not responses from all data storage devices have not been completed at the end of the time set by the timer means. In a limited case, the data recording array device according to any one of the first to seventh aspects of the present invention restarts the timer from the time when the timer expires.

According to a ninth aspect of the present invention (corresponding to claim 9), the completion processing means is configured to execute the processing when all of the responses from the data storage devices have not been completed at the end of the time set by the timer means. Performs data restoration processing using data of a predetermined number of data storage devices including a data storage device that has already completed and has stored redundant data, using data of a data storage device that has not completed a response. , A data storage array device according to any one of the first to seventh aspects of the present invention.

According to a tenth aspect of the present invention (corresponding to claim 10),
In a data access method for a data storage array device having a redundant configuration, a plurality of data storage devices, and a controller for controlling the plurality of data storage devices, the controller may be configured to access the plurality of data storage devices. Instruction groups are sequentially issued in a multiplex manner, a response to an instruction executed in the data storage device is detected, time is counted by a timer means, and a response group to a multiplexed predetermined one access instruction group is generated. Monitoring, when the responses from all the data storage devices are completed within the time set by the timer means, at the time when the responses are completed, complete the response process for the predetermined one access instruction group; At the end of the time set by the timer means, all the data storage devices If the response has not been completed, at the end of the timer, the response processing for the predetermined one access instruction group is completed using only the response from the data storage device that has already been completed. This is a data access method for the array device.

[0024]

Embodiments of the present invention will be described below in detail with reference to the drawings. FIG. 1 is a configuration diagram of a server system including a data storage array device and a host device according to an embodiment of the present invention.

The data storage array device 20 is connected to the host device 10 and responds to a data access request from the host device 10. The host device 10 and the plurality of data storage devices 31 to 35 are connected to the controller 21, and the controller 21 controls access to the data storage array device 20.

The data storage devices 31 to 35 are data storage devices for recording data and redundant data. Data access requests are sequentially issued from the external host device 10 to the data storage array device 20. Between the external host device 10 and the data storage array device 20, for example,
Fast Wide SCSI standard interface and F
It is connected by an external interface 11 such as an interface of ibreChannel standard.

The controller 21 receives one data access request issued from the external host device 10 and divides it into a plurality of data access requests, which is used as a data access instruction group, and the data storage devices 31 to 35
To perform data access to the data storage devices 31 to 35. For example, if it is a read request,
The data in the data storage devices 31 to 35 is read and transferred to the host device 10. When all of the plurality of access requests to the data storage devices 31 to 35 are completed, a transfer completion notification is sent to the external host device 10 as one data access response.

Next, a method of writing and reading data in a redundant configuration will be described. FIG. 2 is a diagram schematically illustrating a data storage format. As shown in FIG. 2, in the case of five data storage devices 31 to 35 (Disks 1 to 5),
A set of data (cluster) forms a redundant group and is divided into five Disks 1 to 5 and stored. One of them is redundant data, and for example, parity is used. Four of the Disks 1 to 4 are data storage devices for data, in which the data itself is stored. The remaining one Disk 5 is a data storage device for parity, and stores the parity calculated from the four data.

When data is written from the external host device 10 to one cluster, the parity value is calculated and stored in the five data storage devices 31 to 35. The parity is set so that, for example, the sum of the number of 1s of the four data values becomes an even number including the parity bit. For example, if the data value is 0, 1, 0, 0, the parity value is 1
And this value is stored in the parity Disk5.

When a data read request for one cluster is received from the external host device 10, data from the four normal data storage devices 31 to 34 are normally transferred to the external host device 10 collectively. . When a failure occurs in a part of the data storage devices 31 to 35, for example, when a failure occurs in the data storage device 34, the data in the three data storage devices 31 to 33 and the parity data storage device 35 can be read. For example, data in the data storage device 34 can be restored by performing parity calculation. For example, in the case described above, the values 0, 1, 0 of Disks 1 to 3 and Dis
It can be seen from the parity value 1 of k5 that the remaining one data is 0.

(Embodiment 1) Next, the operation of the data storage array device will be described with reference to FIGS.

FIG. 3 is a time chart from the access request of the host device 10 to the completion of the response when the command queuing number is 3. Access requests from the host device 10 are distinguished by symbols A, B, and C in FIG. FIG. 5 shows steps from reception of an access request to issuance of an instruction to Disks 1 to 5. Check that the request from the host does not exceed the queuing upper limit (3 in this case), and if not, accept the request and
Generate instructions to access ~ 5. In the case of writing, redundant data is generated to generate a plurality of write instructions for data and redundant data, and in the case of reading, Disks 1 to 3 constituting a cluster belonging to one instruction group are generated.
5, a plurality of read instructions are generated. The generated instruction is issued to each of the Disks 1 to 5 in the next step.

As shown in FIG. 3, a plurality of requests A, B, C
Are arranged in respective instruction groups and are issued in a multiplexed manner. When an instruction of the instruction group A is issued to the Disks 1 to 5, the instruction is executed in each Disk.
Due to the nature of the Disk, the seek time of the magnetic head, the rotation waiting time of the disk, the retry time, and the like occur, and the response time differs depending on the Disk. In the case of FIG. 3, for the instruction group A, the response of the disk 2 is completed first, then the responses from the disk are sequentially detected, and finally, the detection of the response of the instruction group A is completed when the response of the disk 1 is completed.

Next, the timer setting and the instruction completion processing will be described with reference to FIG. When the response detection process is started, first, a timer is set (at the very beginning, as shown in FIG. 3, the execution start timings are all set, and the operation is started in accordance therewith). Thereafter, the responses of the Disks 1 to 5 are checked, and it is checked whether or not there is a response from all the Disks 1 to 5 belonging to the instruction group A. Before the time is up, all Disks 1 belonging to instruction group A
If there is a response of 5 to 5 (see FIG. 4A),
A response is made to the host 10 (instruction completion processing). Of course, as shown in FIG. 3, each disk starts the instruction processing of B after the processing of A is completed. Then, the reset of the timer is started from the point in time when the response from the latest Disk1 is completed.

If the timer has timed out before the responses from all of the Disks 1 to 5 have yet to be prepared (see FIG. 4B), the host 10 is sent to the host 10 at that time as described later. Respond (instruction completion processing). In this case, the reset of the timer is started from the point in time when the response from Disk 1, which has the longest response, ends.

By executing a plurality of the response detection processes for each instruction group, the response detection processes can be performed independently of each other.

In addition, one response detection process may be commonly executed repeatedly to sequentially process a plurality of instruction groups. When processing is performed sequentially, processing does not overlap between groups as shown in response detection in FIG. 3, but when this is actually realized, a timer may be exclusively used using, for example, a semaphore.

Next, a more specific description will be given. The instruction completion processing will be described with reference to FIG. In the response detection process, when the responses of all the Disks 1 to 5 belonging to the instruction group are returned earlier than the detection of the timeout of the timer, as shown in FIG. Perform completion processing. At this time, if the access request from the host device 10 is a read, the command completion processing is performed by using the read data (cluster) in the host device 10.
Transfer to When the timer times out before responses of all the Disks 1 to 5 belonging to the instruction group, FIG.
As shown in (b), an instruction completion process is performed at the time of timeout. In the instruction completion process at this time, if the data storage device returning the response is, for example, Disks 1 to 4,
Since the necessary data is available, the data is transferred to the host device 10. One of Disks 1 to 4
If no response is returned from k, the data is restored using the data of Disk 5, which is redundant data, and transferred to the host device 10 as described above.

Next, an operation for sequentially detecting a response from a plurality of instruction groups will be described. When the response detection of the instruction group A in FIG. 3 is completed, the detection of the response of the next instruction group B is started (timer start). At the start of detecting the response of the instruction group B, the Disks 2 to 5 have already started executing instructions. In the response detection process of the instruction group B, the response detection is terminated when the timer times out or when the responses from all the Disks 1 to 5 belonging to the instruction group B are detected, similarly to the group A. After this,
For the instruction group B, the instruction completion processing is executed similarly to the instruction group A.

Hereinafter, the same processing is performed for the instruction group C. In the processing of each instruction group, when a new access request is input from the host device 10, if a queue is free, it is accepted, and the same processing sequence is subsequently performed.

By the above operation, the data storage devices 31 to 31
In 35, since the operation is performed in the command queued state, the operation can be performed so that the head movement and rotation waiting time can be minimized.
Since 1 to 35 operate efficiently, the transfer rate can be increased. Further, since the timer is a timer value according to the execution time of each instruction group, it may be a constant value regardless of the number of queued instructions (commands). Further, the timer value evaluates the instruction execution time of the data storage devices 31 to 35, so that the maximum transfer performance of the data storage devices 31 to 35 can be obtained and the response delay time can be minimized. . In addition, since the response detection shown in FIG. 3 only needs to use one timer exclusively, there is no need to multiplex the management of the timer.

The start of the next timer is not limited to the above example. The next start may be performed at the timing when the response is completed earliest. Alternatively, in FIG.
The earliest Disk and the slowest D, like 3,4,5
It may be started when the response of the disk in the middle of the disk ends.

Further, if no timeout occurs,
As described above, only when a timeout occurs, the next timer may be started after the timeout.

(Embodiment 2) Next, referring to FIG. 7, an example of operation for making the start of response detection independent for each instruction group will be described. In FIG. 7, the timing of the start of the response detection (start of the timer) is different. For each instruction group,
When the instruction belonging to the instruction group first starts executing in one of the Disks 1 to 5, the response detection (start of the timer) of the instruction group is started. The response detection process in this case is executed independently for each instruction group, and the timer is also separately counted for each instruction group. In the case of the instruction group B, execution of the instruction group is first started in the Disk 2, so that the response detection is started from here.
The content of the response detection processing is the same (FIG. 6). This operation also has the same effect as the operation of FIG. 3, but furthermore, since the response delay from the data storage device which responds the fastest can be evaluated for each instruction group, the application is more demanding for the response delay. , You can make a quicker response. Further, when a delay occurs in one data storage device on a regular basis, the data storage device is always timed out so that a quick response is always possible.

As a modified example, the timer may be started at the latest execution start timing, or may be started at an intermediate timing between the earliest execution start timing and the latest execution start timing. You can start it.

As described above, in the data storage array device according to the present embodiment, there may be a case where a failure such as a delay temporarily occurs due to the multiple issue of the instruction and the response detection processing of the instruction group. Normal data can be supplied within a fixed delay time. At this time, access inside the data storage device can be efficiently performed by issuing multiple instructions.

In FIG. 7, it is assumed that when any one of the data storage devices starts executing the instruction, the response detection of the instruction group is started. Instead, the instruction of the immediately preceding instruction group is started in any one of the data storage devices. Response detection may be started when the execution ends. Further, instead of the time at which the execution of an instruction is started in any one of the data storage devices, the response detection of the instruction group may be started at the time when the execution of the instruction belonging to the instruction group is started in the entire data storage device. In FIG. 3, after the response detection processing is completed, the instruction completion processing of the instruction group in which the response is detected and the response detection processing of the next instruction group are performed. May be started.

In this embodiment, instructions are multiplexed in response to an access request from the host device.
In the data storage array device, access to the data storage device such as data reconstruction processing may be multiplexed and issued.

In this embodiment, the total number of data storage devices and the number of data storage devices forming an instruction group are the same. However, the number of data storage devices forming an instruction group is one or more arbitrary number. May be.

The value of the timer for detecting the response delay for each instruction group may be different for each instruction group, and may be changed according to the number of command queuing.
Response time according to the presence or absence of queuing can be guaranteed.

[0051]

As is apparent from the above description, according to the present invention, an instruction can be efficiently transferred to a data storage device by multiplexing instructions by command queuing, and a response can be detected for each instruction group. Timer management) can guarantee a response delay.

Further, the timer value in this response detection can be set to a constant value irrespective of the number of command queuing, and in this case, the response delay time can be made uniform.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a configuration of a data storage array device according to an embodiment of the present invention.

FIG. 2 is a block diagram showing a management area of the data storage array device.

FIG. 3 is an execution timing diagram of an instruction group.

FIG. 4 is a timing chart of an instruction completion process.

FIG. 5 is a flowchart of an instruction issuing process.

FIG. 6 is a flowchart of a response detection process.

FIG. 7 is an execution timing diagram of an instruction group.

[Explanation of symbols]

 Reference Signs List 10 host device 11 external interface 20 data storage array device 21 controller 31, 32, 33, 34, 35 data storage device

Continued on the front page F term (reference) 5B065 BA01 CA15 CA30 CC08 CE12 CH05 CH13 5D044 AB07 BC01 CC04 CC09 DE03 DE12 DE72 DE92 EF02 GK05 GK08 HL02 HL11

Claims (10)

[Claims] 1. A data storage array device comprising: a plurality of data storage devices in a redundant configuration; and a controller that controls the plurality of data storage devices. The controller: Instruction issuing means for sequentially issuing an access instruction group composed of a plurality of divided access instructions, response detecting means for detecting a response to an instruction executed in the data storage device, and timer means for counting time. Monitoring a response group to the predetermined one access command group, and when responses from all the data storage devices are completed within a time set by the timer means, at the time of the completion, the predetermined Completes the response process to one of the access instruction groups, and is set by the timer means. Between the end point in time,
If the responses from all the data storage devices have not been completed, the response to the predetermined one access command group is performed at the time when the timer ends by using only the responses from the already completed data storage devices. A data storage array device, comprising: completion processing means; and completion processing means. 2. The timer for a predetermined one access instruction group is started at a timing at which a response is completed latest among a response group to a predetermined one access instruction group immediately before the timer. The data storage array device according to claim 1. 3. The timer for a predetermined one access instruction group is started at a timing at which a response ends earliest among a response group to a predetermined one access instruction group immediately before the access instruction group. The data storage array device according to claim 1. 4. The start of the timer for one predetermined access instruction group includes the timing of the earliest end of the response group among the response groups for the one predetermined access instruction group immediately before that. 2. The data storage array device according to claim 1, wherein the data storage array device is started at a timing intermediate to the timing at which the response ends late. 5. The data storage according to claim 1, wherein the timer of a predetermined one access instruction group is started at a timing at which execution of the timer is started earliest in the predetermined one access instruction group. Array device. 6. The data storage according to claim 1, wherein the timer of a predetermined one access instruction group is started at the latest timing of execution of the predetermined one access instruction group. Array device. 7. The timer of a predetermined one access instruction group is started at the timing at which execution is started earliest and the timing at which execution is started latest at the predetermined one access instruction group. 2. The data storage array device according to claim 1, wherein the data storage array device is started at an intermediate timing. 8. The completion processing means, if only responses from all data storage devices have not been completed at the end of the time set by the timer means, from the end of the time of the timer. The data recording array device according to any one of claims 1 to 7, wherein the timer is restarted. 9. When the responses from all the data storage devices are not completed at the end of the time set by the timer device, the completion processing unit is configured to determine whether the response has not been completed. The data restoration process according to any one of claims 1 to 7, wherein data restoration processing is performed by using data of a predetermined number of data storage devices including a data storage device storing redundant data which has already been completed. Data storage array device. 10. A data access method for a data storage array device comprising: a plurality of data storage devices having a redundant configuration; and a controller that controls the plurality of data storage devices, wherein the controller comprises: the plurality of data storage devices , An access instruction group is sequentially issued in a multiplexed manner, a response to an instruction executed in the data storage device is detected, time is counted by a timer means, and a multiplexed predetermined one access instruction is issued. A response group for the group is monitored, and when responses from all data storage devices are completed within the time set by the timer means, a response to the predetermined one access instruction group is made at the time when the response is completed. When the processing is completed and the time set by the timer means ends,
If the responses from all the data storage devices have not been completed, the response to the predetermined one access command group is performed at the time when the timer ends by using only the responses from the already completed data storage devices. A data access method for a data storage array device, which completes processing.