JPH07244562A - Method for controlling access to mirror disk - Google Patents

Abstract

PURPOSE:To provide a mirror disk access controlling method capable of improving throughput independently of the existence of deflection of a reading disk area from an application program. CONSTITUTION:An operating system 10 connected between an application program and a mirror disk executes following access control. Namely a reading request out of access requests sent from the application program is allocated to one disk device 21 or 22 in which the number of processing queuing access requests in a mirror disk management table 109 is relatively small. When an access request is a writing request, the contents of an access request block 108 in a completion queue 104 are recognized, whether an objective area to be accessed is overlapped between the writing request be accessed is overlapped between the writing request and a reading request in a processing end queuing state or not is judged and a non-overlapped writing request is allocated to both the disk devices 21, 22. On the other hand, an overlapped writing request is queued until the reading request in the processing end queuing state ends its processing.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION The present invention relates to a mirrored disk for protecting the data in a disk by writing the same data in two series of disk devices.
The present invention relates to an access control method, and more particularly, to a mirror disk access control method for increasing the number of access processes to a mirror disk from a higher-order program per unit time in an operating system that controls a mirror disk by an interrupt transmission / reception function from a plurality of disk devices.

[0002]

2. Description of the Related Art As a mode of accessing a mirror disk from an operating system, there are a data write request and a data read request. In this case, the write request must be issued to both disk devices, but the read request is sufficient to be issued to either one of the disk devices. A method for speeding up data read processing from a mirror disk by utilizing such characteristics has been devised conventionally.

A typical method will be described below with reference to FIGS.
This will be described with reference to (c). In FIG. 7, R represents the data read time and W represents the data write time, and it is assumed that there is no overlap in the respective access target areas (areas corresponding to the access size from the access destination start address of the disk device).

(1) Fixed Read Method As shown in FIG. 7A, a method of fixing the read request disk device to any one of the disk devices constituting the mirror disk, in the illustrated example, the primary disk device. Is. This method is described in, for example, “Office Computer with advanced fault countermeasure functions”, Nikkei Computer, 1991.7.
You can refer to 29 articles.

(2) Simultaneous request issuing method As shown in FIG. 7B, a mirror disk is constructed 2
Issue read requests to disk devices of the same series at the same time,
This is a method of adopting the one of the two issued read requests that has been completed first and canceling the other. For the method of issuing requests simultaneously, see, for example, Furukawa et al., "I / O Performance in Duplex Volume Method", Information Processing Society of Japan 37th (late 1988) National Convention, 4P-4, pages 236-237, etc. Can be used as a reference.

(3) Split seek method Most of the disk access time is the time (seek time) until the read / write head of the disk device reaches the target cylinder. The longer the moving distance of the read / write head of the disk device, the longer the seek time. Therefore, in order to reduce the seek time, as shown in FIG. 7C, half of the cylinder is read for one disk device of the mirror disk, and the other half of the cylinder is read for the other disk device. The method of performing access control is the split seek method. For this split seek method, see, for example, Gray et al., “Fault Tolerant System”, McGraw-Hill Book, Inc., 1.
You can refer to the articles on pages 07-109.

[0007]

However, each of the conventional methods has the following problems. (1) Fixed Read Method The data read speed is exactly the same as the data read speed in the case of a single disk device, and the characteristics of the mirror disk using two series of disk devices are not utilized.

(2) Simultaneous request issuing method Since a single read request occupies two series of disk devices at the same time, the response time per read request seen from the host program is short, but conversely per unit time. It is not possible to increase the number of access processing cases (throughput) to the mirror disk from the higher-level program.

(3) Split seek method When the read requests from the host program are not evenly distributed over the entire area of the mirror disk, the read requests are concentrated on either one of the disk devices, and vice versa. In addition, the other disk device has a lot of free time. As a result, the throughput is reduced as a whole.

SUMMARY OF THE INVENTION It is an object of the present invention to solve the above problems and provide a mirror disk access control method for improving throughput regardless of whether or not there is a bias in the disk device to be read from the host program.

[0011]

The mirror disk access control method of the present invention provided by the present invention comprises a mirror disk access control unit for controlling the reading or writing of data with respect to a mirror disk in accordance with an access request from a host program. The mirror disk access control unit detects the type of read or write of the access request, the unit for recognizing the access target area, and the number of pending access requests for each disk device constituting the mirror disk. The operating system including the means for performing the following processes.

That is, of the access requests sent from the upper-level program, the read request is assigned to one of the disk devices that has a relatively small number of access requests waiting to be processed. In this case, the number of access requests waiting for processing is
The state of the load on each disk device is determined by decreasing each interrupt reception from the disk device.

When the access request sent from the upper-level program is a write request, it is determined whether or not the access target area overlaps with the read request waiting for processing completion, and the write request that does not overlap is determined. Is assigned to both disk devices, and duplicate write requests are waited until the processing for the read request in the processing completion waiting state is completed.

[0014]

According to the mirror disk access control method of the present invention, the disk device on the mirror disk side that is occupied when the access request from the higher-order program is a read request is always the one with the lighter load.

In the case of a write request, first, it is determined whether or not the access target area overlaps with the read request in the processing end waiting state, and write requests that do not overlap are sent to both disk devices. Write requests that are allocated and overlapped are waited until the processing for the read request in the processing completion waiting state is completed.

Therefore, the number of access requests from the host program to each disk device is constantly the same regardless of whether the access target area is biased or not. It should be noted that the number of access requests decreases each time an interrupt occurs from each disk device. This interrupt indicates that the processing is completed in the disk device.

[0017]

Embodiments of the present invention will now be described in detail with reference to the drawings. FIG. 1 is a conceptual configuration diagram for realizing the method of the present invention, 10 is an operating system, 2
Reference numerals 1 and 21 denote primary and secondary disk devices constituting a mirror disk, and 30 denotes an application program (upper program).
There are a plurality of application programs 30, and the operating system issues a mirror disk access request, that is, a write request for writing data in the primary or secondary disk device 21 or 22, or a read request for reading data from either disk device. Issue to 10.

At this time, in any case, the head address of the data to be accessed in the mirror disk and the size of the data to be accessed are specified. In this embodiment, data starting from this head address and having a data size corresponding to the data size or its recording area will be referred to as an access target area.

The operating system 10 stores the read / write request type (request type) of the mirror disk access request, the access target area, and the number of access requests waiting to be processed for each of the disk devices 21 and 22. The application program 3 includes a storage unit, for example, a management table.
Upon receiving a mirror disk access request from 0, access control to the disk devices 21 and 22 is performed according to the request type. In this way, a mirror disk is configured by passing through the operating system 10.
From the application program 30, the two disk devices 21 and 22 appear as one disk device as indicated by the broken line.

FIG. 2 is an explanatory diagram showing the relationship between the operating system 10 having the above-mentioned configuration and the mirror disk. In FIG. 2, 101 is a processing queue, 102 is a primary disk queue, 103 is a secondary disk queue, and 104.
Is a completion waiting queue, 105 to 108 are access request blocks stored in each queue, and 109 is a mirror disk management table.

In the mirror disk management table 109, the number of access request blocks to the primary disk device 21, that is, the number of access request blocks 106 in the primary disk queue 102 and the number of access request blocks to the secondary disk device 22,
That is, the number of access request blocks 107 in the secondary disk queue 103 is stored.

The contents of the access request block 106 are shown in FIG.
, The request type (for example, numerical information) that indicates whether the mirror disk access request from the application program 30 is a read / write request, the disk access position (the head of the data to be accessed in the mirror disk). Address), access size (data size to be accessed), disk access completion count, and access result. The disk access completion number has a meaning when the request type is a write request as described later, and its value is 0, 1, or 2. The other access request blocks 105, 107, 108 also have the same configuration.

The above-mentioned "number of access requests waiting to be processed"
Means the number of access request blocks 106 and 107 in the waiting state in each of the disk queues 102 and 103, and the “read request in the processing end waiting state” is
Among the access request blocks 108 in the waiting state in the completion waiting queue 104, the request type is “read”.

Next, a concrete flow of the mirror disk access control, that is, the data read request process and the data write request process by the operating system 10 having the above setting will be described with reference to FIGS.

(Data Read Request Process) FIG. 4 is a procedure diagram of a process of separating an access request from the upper program 30 and an interrupt from the mirror disk. Normally, the access request from the upper program 30 (data read) is performed. (/ Write request), or waiting for an interrupt from each disk 21, 22 (S301).

Here, for example, when an access request block for a data read request to the mirror disk is issued from the upper program 30, it is not an interrupt from the disk side (S302), so that the processing waiting queue 101 is empty first. It is confirmed whether or not (S303), and if it is free, the access request block is connected to the head of the processing queue 101 (S).
304). If the waiting queue 101 is not empty,
The access request block is added to the end of the process waiting queue 101 (S305), and the process returns to the first process (S301).

FIG. 5 is a procedure diagram of the access request processing for the mirror disk. As described above, when the access request block is connected to the head of the processing waiting queue 101 (S
304) confirms whether or not the access request block is a read request (S401). In this example, since it is a read request, the access request block is transferred to the completion waiting queue 104.
And another access request block (disk access request block 106) is created (S402).

After that, referring to the number of access request blocks recorded in the mirror disk management table 109, if the number of access request blocks of the primary disk device 21 is small, the above-mentioned disk access request block 106 is set at the end of the primary disk queue 102. Connect (S403). At this time, if the disk access request block 106 is the head of the primary disk queue 102, an access request is immediately issued to the primary disk device 21 (S405), and if not, the process returns to the first processing (S301). By this processing, the loads on the primary and secondary disk devices 21 and 22 are equalized,
It is possible to improve the read processing from the upper program 30 to the mirror disk and the throughput.

Next, the interrupt processing from the mirror disk to the operating system 10 will be described with reference to FIG. As described above, a read request is issued to the primary disk device 21, and the access area is accessed (data read).
Is completed, the primary disk device 21 notifies the operating system 10 of an access completion interrupt.

The operating system 10 detects the occurrence of this interrupt (S302), and decrements the access request block number of the primary disk device (primary disk queue 102) recorded in the mirror disk management table 109 by 1 (S50).
1). In this case, since the completed access request block is a read request (S502), the access result of the disk access request block is determined (S503), and if it is normally completed, the disk access request block is released (S504). The access request block is returned to the upper program 30 (S505), and the process returns to the first process (S301).

On the other hand, when the access result of the disk access request block is abnormally terminated in S503, the subsequent use of the primary disk device 21 is stopped (S506), and the secondary disk device 22 in the mirror disk management table 109 is stopped.
The number of access request blocks is increased by 1, the disk access request block 107 (106) is connected to the end of the secondary disk queue 103 (S507), and the process returns to S404.

(Data Write Request Process) Next, the data write request process will be described. When the access request block for the write request to the mirror disk is issued from the upper program 30, the operating system 10 sends S
Until 401, processing is performed in the same flow as the data read request described above.

In this case, the access request block at the head of the processing queue 101 is not a read request (S40).
1) It is determined whether there is a read request in which the access target areas overlap in the completion waiting queue 104 (S406).
If it exists, the process returns to the first process (S301). If it does not exist, 0 is set to the disk access completion number of the access request block 105 (S407).

Then, this access request block is connected to the end of the completion waiting queue 104 to create two disk access request blocks (S408), and the disk access request block 106 and the disk access request block 106 are respectively created in the primary disk queue 102 and the secondary disk queue 103. Connect 107. At this time, the disk access request blocks 106 and 107
Is not the head of the disk queues 102 and 103 respectively, the process returns to the first processing (S301) (S410), and if the head is the head, an access request, that is, a write request is issued to each of the disk devices 21 and 22 (S411). . As a result, data is written in the corresponding access area of each disk device 21, 22.

When the writing in the primary disk device 21 is completed, the primary disk device 21 notifies the operating system 10 of an access completion interrupt. The operating system 10 detects this interrupt (S301), and as shown in FIG. 6, decrements the access request block count of the primary disk device of the mirror disk management table 109 by 1 (S501). Then, the access result of the disk access request block 106 is judged (S508),
If it is abnormally terminated, the subsequent use of the primary disk device 21 is stopped (S513). If it is normal end, the disk access block 106 is released (S509), 1 is added to the disk access completion number of the access request block 108 of the completion waiting queue 104 (S510), and it is determined whether the disk access completion number is 2 or not. Yes (S511).

The fact that the disk access completion number is not 2 means that the write request processing (that is, the read request processing before that) in the secondary disk device 22 has not been completed, so the processing returns to the processing of S404, and It waits for an access completion interrupt notification from the disk device 22 (S511).

When the access completion interrupt is notified from the secondary disk device 22, the same process as the process when the access completion interrupt notification is issued from the primary disk device is performed before S511. And the number of disk access completion is 2
When it becomes (S511), the access request block is returned to the upper program 30 (S512).

As a result of the control according to the above flow, when the access request from the upper program 30 is a read request, the disk device on the mirror disk side that is occupied is always the one having a smaller load, and the write request Among them, the write request whose access target area overlaps with the read request waiting for the completion of processing is kept waiting until the processing for the read request is completed, so that the contradiction of the read data is avoided. Further, the time required for the read / write request processing at that time is as shown in FIG. 7D, and it can be seen that the time is shortened as compared with each conventional method. It should be noted that the present invention is not limited to the flow of the above-described embodiment, and the procedure can be changed without departing from the scope of the invention.

[0039]

As is apparent from the above description, according to the mirror disk access control method of the present invention, the disk device on the mirror disk side occupied when the access request from the upper program is a read request is relatively Since the one having a small load is allocated to the mirror disk, there is an effect that the bias of the load on the mirror disk is prevented regardless of the presence or absence of the deviation of the access target area to the mirror disk from the upper program.

As a result, the throughput of the mirror disk access from the upper program is significantly improved.
The conventional problems can be solved.

[Brief description of drawings]

FIG. 1 is a schematic configuration diagram for realizing a mirror disk access control method of the present invention.

FIG. 2 is a conceptual explanatory diagram of a mirror disk access control method according to an embodiment of the present invention.

FIG. 3 is an explanatory diagram of contents of an access request block used in this embodiment.

FIG. 4 is a flowchart showing the flow of processing according to this embodiment.

FIG. 5 is a flowchart (continuation) showing the flow of processing according to the present embodiment.

FIG. 6 is a flowchart (continuation) showing the flow of processing according to the present embodiment.

FIG. 7 is a processing time comparison diagram of each conventional method and the method of the present embodiment.

[Explanation of symbols]

 10 Operating System 101 Processing Queue 102 Primary Disk Queue 103 Secondary Disk Queue 104 Completion Waiting Queue 105 to 108 Access Request Block 109 Mirror Disk Management Tables 21 and 22 Disk Devices Constituting Mirror Disk

Claims (3)

[Claims] 1. A mirror disk access control unit for controlling reading or writing of data to a mirror disk according to an access request from a higher-level program, wherein the mirror disk access control unit is a type of reading or writing of the access request. Of the access requests sent from the higher-order program in the operating system having a means for determining the access target area, a means for recognizing the access target area, and a means for detecting the number of processing waiting access requests for each disk device forming the mirror disk. The mirror disk access control method, wherein the read request is assigned to one of the disk devices for which the number of access requests waiting to be processed is relatively small. 2. The mirror disk access control method according to claim 1, wherein the number of access requests waiting to be processed is reduced each time an interrupt is received from the disk device. 3. The mirror disk access control method according to claim 1, wherein when an access request sent from the upper-level program is a write request, an access target area overlaps with a read request waiting for processing completion. Mirrors characterized by deciding whether or not to do so, assigning non-overlapping write requests to both disk devices, and waiting for duplicate write requests until the processing for the read request in the processing end waiting state is completed. Disk access control method.