JPH04273548A - Disk controller - Google Patents

Abstract

PURPOSE:To dynamically change a disk device to be performed an access of according to the degree of a load, and to efficiently distribute the load among the plural disk devices. CONSTITUTION:The concentration of the load of the disk device corresponding to a block to be ejected is checked, and when the load is concentrated, the prescribed logical block address of the other disk device assigned on a disk cache memory mechanism 114, is selected as the address to be substituted. Then, the selected storage area address is substituted for the logical block address of the block to be ejected on an address conversion table 112 and the disk cache memory mechanism 114. As the result, the destination of the writing of the data of the block to be ejected can be changed to the other device whose address is substituted.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a disk control device, and more particularly to a load distribution technique for a disk subsystem that controls a plurality of disk devices and has a disk cache for them.

0002

2. Description of the Related Art Generally, a disk subsystem controls a plurality of magnetic disk devices (hard disk drives) and has a disk cache memory therefor.

This disk cache memory is a temporary storage area for speeding up data input/output access, and storage areas of multiple disk devices are allocated to it.
Each allocated storage area address and the data of each storage area address are held as one block.

Access using this disk cache memory is performed as follows.

That is, when there is a read access request to a magnetic disk device, the disk cache memory is first referred to, and if data at the requested address exists there, that data is immediately transferred to the host device.

If the data does not exist, the data is read from the magnetic disk device and stored in the disk cache memory. At this time, if there is no free space on the disk cache memory, reassignment is performed in which less frequently used data is ejected and new data is registered in that space.

[0007] Furthermore, when the request from the host device is for writing, if there is data at the requested address on the disk cache memory, only the data on that cache memory is rewritten for the time being, and the magnetic disk device has no disk space. Written back when cache memory is reallocated. In other words, data ejected from the cache is written to the magnetic disk device.

If the data at the requested address does not exist on the disk cache memory, the above-mentioned reallocation is performed to create an unused area on the cache, and the write data is written to the unused area. It will be done.

If such a disk cache mechanism is employed, desired data input/output can be performed without actually accessing the magnetic disk device, so that data input/output access can be made faster.

However, since access requests to a plurality of magnetic disk drives are not uniform, the load may be concentrated on a specific magnetic disk drive even when a disk cache mechanism is used.

For example, data ejected from a disk cache must be written back to a magnetic disk device, but if this write-back is concentrated on a specific magnetic disk device, the load on that magnetic disk device increases locally and temporally. Ru.

[0012] Conventionally, as a method for distributing the load to multiple magnetic disk devices, it is possible to write files in units of files to different magnetic disk devices, or to divide one file into pieces of a certain length and write them to separate magnetic disk devices. File division arrangement such as writing was used.

[0013] Using this file division arrangement method,
In the long run, it becomes possible to distribute the load among multiple magnetic disk devices. However, this load distribution is probabilistic, and when the system is actually used, there are many cases where the load is concentrated only on a specific magnetic disk device in terms of local time.

[0014] Such concentration of load causes a bottleneck and is a major factor in deteriorating the operating performance of the entire disk subsystem.

[0015]

Conventionally, when the system is actually used, the load is concentrated on a specific magnetic disk device, making it difficult to distribute the load.

The present invention has been made in view of the above points, and it is possible to dynamically change the disk device to be accessed according to the degree of load during actual system use, so that multiple disk devices can be accessed. An object of the present invention is to provide a disk control device that can efficiently distribute the load between disks.

[0017]

[Means and effects for solving the problems] This invention has the following features:
In a disk control device that includes a plurality of disk devices and a disk cache memory to which storage areas of these disk devices are allocated and holds each allocated storage area address and data of each storage area address as one block, An address conversion table in which a correspondence between a storage area address specified by a host device and storage areas of the plurality of disk devices is defined, and a disk specified by a storage area address of a block to be ejected from the disk cache memory. means for detecting the presence or absence of load concentration on a device, and a predetermined storage area address of another disk device allocated on the disk cache memory when load concentration on the disk device corresponding to the block to be ejected is detected; and a storage area address of the block to be ejected on the address conversion table and the disk cache memory, respectively, so that data writing to a disk device with a concentrated load is transferred to another disk device. It is characterized by changing the writing to.

In this disk control device, load concentration on the disk device corresponding to the block to be ejected is checked, and if the load is concentrated, a predetermined storage of another disk device allocated on the disk cache memory is checked. The area address is selected as a target for address replacement. Then, the selected storage area address and the storage area address of the block to be ejected are exchanged on the address conversion table and the disk cache memory, respectively. As a result, it is possible to change the writing destination of the data of the block to be ejected to another disk device whose address has been swapped. Therefore, it becomes possible to dynamically change the disk device to be accessed.
It becomes possible to efficiently distribute the load among multiple disk devices.

[0019] Furthermore, since the addresses in the address translation table have also been swapped, an access request from the host device to the storage area selected for the address swapping will be handled as follows:
The request is changed to an access request to the storage area address of the block to be ejected. Since the data on the disk cache memory has not been replaced, the data at the storage area address of the block to be ejected remains the data at the storage area address selected as the address replacement target. Therefore, the host device can issue an access request without being aware of address replacement.

[0020]

Embodiments Hereinafter, embodiments of the present invention will be described with reference to the drawings.

FIG. 1 shows the configuration of a disk subsystem according to an embodiment of the present invention.

This disk subsystem includes a disk subsystem control mechanism 11 and a plurality of magnetic disk devices (
The disk subsystem control mechanism 11 includes a disk subsystem control device 11.
1, address translation table 112, command queue 1
13 and a disk cache memory mechanism 114 are provided.

The disk subsystem control device 111 includes:
It executes and controls data input/output to and from the disk group 12 in accordance with instructions from the host CPU. During data input/output, the address conversion table 112, command queue 113, and disk cache memory mechanism 11
See 4.

The disk subsystem control device 111 also has a function of updating the contents of the directory and address translation table of the disk cache memory mechanism 114 as necessary in response to a request to eject a block from the disk cache memory mechanism 114. have.

[0025] The address translation table 112 is
This is a table that makes a one-to-one correspondence between the storage area address specified by the PU and all storage areas of multiple disk devices A to C. This table allows the logical address from the host CPU to be assigned to the logical block of disk devices A to C. converted to an address.

Here, for the sake of simplicity, the disk device A
Consider the case where ~C each have only 4 logical blocks.

That is, disk device A is provided with four storage areas specified by logical block addresses A0 to A3, and disk device B is provided with four storage areas specified by logical block addresses B0 to B3. The disk device C has logical block addresses C0 to C.
Four storage areas designated by 3 are provided, and the disk device D is provided with four storage areas designated by logical block addresses D0 to D3.

In this case, in the address conversion table 112, the logical block addresses Ao to A3 of the magnetic disk device A are assigned to the logical addresses a0 to a3 from the host CPU, and the logical addresses b0 to b3 are assigned to the logical addresses a0 to a3 from the host CPU. Logical block addresses Bo to B3 of magnetic disk device B are assigned to
Logical block addresses Co to C3 of the magnetic disk device C are assigned to the logical addresses c0 to c3, and logical block addresses Do to D3 of the magnetic disk device D are assigned to the logical addresses d0 to d3.

The command queue 113 is a queue for write and read requests to the magnetic disk devices A to C, and write and read requests are queued for each magnetic disk device A to C.

The disk cache memory mechanism 114 includes:
A part of the storage area of disk devices A to C is allocated,
The logical block address of each allocated storage area and the data of each logical block address are held as one cache block.

That is, the disk cache memory mechanism 114 is provided with a directory area and a data area, logical block addresses are registered in the directory area, and data of corresponding logical block addresses are registered in the data area.

[0032] Also, the disk cache memory mechanism 11
4 contains the least frequently accessed cache block (
A mechanism (not shown) for detecting the access frequency of a cache block is provided in order to perform reassignment of a cache block that has not been used for the longest time as a target for ejection.

Since frequently accessed cache blocks are difficult to be ejected, the disk cache memory mechanism 11
Cache blocks with a high degree of residence and a low access frequency have a low degree of residence.

In order to display the degree of residence, the logical block addresses of the directory area are shown arranged in descending order of degree of residence. In the figure, the logical block address Co has the highest degree of residence, and the logical block address C2 has the lowest degree of residence.

Next, referring to the flowchart of FIG.
The basic access operation of the disk subsystem control mechanism 11 will be explained. Here, for the sake of simplicity, a case will be described in which only write data is allocated to the disk cache memory mechanism 114.

First, the disk subsystem control device 11
1 receives an access request from the host CPU,
By referring to the address conversion table 112, the logical address of the host CPU is converted to the logical block address of the magnetic disk devices A to D (step S11).

Next, the disk subsystem control device 1
11 refers to the directory of the disk cache memory mechanism 114 and checks whether the logical block address resulting from the conversion is assigned to the disk cache memory mechanism 114 (step S12).

If the logical block address exists in the disk cache memory mechanism 114 (cache hit), the disk subsystem control device 111 accesses the disk cache memory mechanism 114 (step S16).

That is, in the case of a write request, the data at the corresponding logical block address on the disk cache memory mechanism 114 is rewritten, and in the case of a read request, the data at the corresponding logical block address on the disk cache memory mechanism 114 is rewritten. is host C
Transferred to PU.

On the other hand, if the logical block address does not exist in the disk cache memory mechanism 114 (mishit), the following processing is performed.

That is, in the case of a miss in a write request, the disk subsystem control device 111 waits for a free block to become available in the disk cache memory mechanism 114, and allocates write data thereto (steps S13, S14, S15). ).

Here, waiting for a free block means that if there is a free block on the disk cache memory mechanism 114, write data is immediately allocated to it, but if there is no free block, it waits until a free block is generated by block ejection processing. This refers to allocating write data.

If there is a mishit in the read request, the disk subsystem control device 111 queues the read request in the command queue 113 (step S17), reads the corresponding magnetic disk device, and reads the data. After reading, the drive termination process of the magnetic disk device is performed (step S18).

Next, referring to the flowchart of FIG.
The operation in response to a cache block ejection processing request will be explained.

This ejection request processing is executed when a miss occurs in a write request and there is no free block on the disk cache memory mechanism 114 at that time.

Now, as shown in FIG. 1, a large number of input/output requests to magnetic disk device C are queued in the command queue 113, and other magnetic disk devices A, B, D
Consider a case where the number of waiting input/output requests for is small.

Assume that in such a situation, it becomes necessary to eject a cache block from the disk cache memory mechanism 114 and reallocate the block.

In this case, the least resident block, that is, the logical block address C2 (magnetic disk device C
A block to which a second storage area (second storage area) is allocated is selected as an ejection target.

[0049] The disk subsystem control device 111
By referring to the command queue 113, the amount of load on the magnetic disk device C corresponding to the logical block address C2 to be ejected is checked (step S21). This recognizes that the load is concentrated on the magnetic disk device C.

Next, the disk subsystem control device 1
11 searches the directory of the disk cache memory mechanism 114 to find the magnetic disk devices A and B assigned to the disk cache memory mechanism 114.
, D, the logical block address with the highest degree of residence is selected as a conversion target (steps S22, S23).

Here, since the logical block address B2 of the magnetic disk device B has the highest degree of residence, the logical block address B2 becomes the object of conversion of the logical block address C2 to be ejected.

[0052] The disk subsystem control device 111
Swap logical block address C2 and logical block address B2 on address conversion table 112 (
Step S24). As a result, for subsequent access requests, the logical address b2 from the host CPU will be converted to the logical address C2 of the magnetic disk device C, and the logical address c2 will be converted to the logical address b2 of the magnetic disk device B. .

[0053] Also, the disk subsystem control device 11
1 replaces the logical block address C2 and the logical block address B2 on the directory of the disk cache memory mechanism 114 independently of the data (step S25).

As a result, the logical block address of the cache block to be ejected is changed from logical block address C2 to logical block address B2, and the logical address of the cache block to which logical block address B2 was assigned is changed from logical block address B2 to logical block address B2. The address is changed to C2.

In this case, since the contents of the data area of the cache block are not replaced, the logical block address C2 is assigned to the data (D-b2) at the logical block address B2, and the data at the logical block address C2 is Logical block address B2 will be assigned to (D-c2).

In this way, when the logical block address of the cache block to be ejected is changed to the logical block address B2 of the magnetic disk device B, the data (D-c2) of the cache block to be ejected is written back to the magnetic disk drive B. This becomes disk device B.

After this, the disk subsystem control device 1
11 queues the data (D-c2) write-back request to the write-back destination position of the command queue 113, that is, the position corresponding to the magnetic disk device B (step S
26). As a result, the data (D-c2) that should originally have been written to the logical block address C2 of the magnetic disk device C is now written to the logical block address B2 of the magnetic disk device B.

Furthermore, as mentioned above, the data (D-b2) that was originally supposed to be written to the logical block address B2 of the magnetic disk device B is written to the magnetic disk device C.
Since the logical block address C2 has been newly assigned, if the cache block with the logical block address C2 is to be ejected, the data (D-b2
) will be written back to the logical block address C2 of the magnetic disk device C where the load is concentrated.

However, since the data (D-b2) is used frequently, the cache block at the logical block address C2 will not be ejected for at least a certain period of time, and will be written back to the logical block address C2 of the magnetic disk device C. There's no chance of it happening. Therefore, during a period when the load on the magnetic disk device C is concentrated, data in the cache block is not written back to the magnetic disk device C.

[0060] Also, the logical address b2 is sent from the host CPU.
When an access request is issued to the logical block address C, the address translation table 112
2 is converted into an access request to the logical block address C on the disk cache memory mechanism 114.
2 data (D-b2) is accessed.

This data (D-b2) is at logical address b
Since this is the original data specified in 2, the host CP
U can normally read and rewrite desired data.

As described above, in this embodiment, the load concentration on the disk device corresponding to the block to be ejected is checked, and if the load is concentrated, other disks allocated on the disk cache memory mechanism 114 are checked. A predetermined logical block address of the device is selected as an address replacement target.

Then, the selected storage area address and the logical block address of the block to be ejected are exchanged on the address conversion table 112 and the disk cache memory mechanism 114, respectively. As a result, it is possible to change the writing destination of the data of the block to be ejected to another disk device whose address has been swapped.

[0064] Therefore, it becomes possible to dynamically change the disk device to be accessed, and it becomes possible to efficiently distribute the load among a plurality of disk devices.

[0065]

[Effects of the Invention] As detailed above, according to the present invention,
When the system is actually used, it becomes possible to dynamically change the disk device to be accessed depending on the degree of load, and it becomes possible to efficiently distribute the load among a plurality of disk devices.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing a system configuration according to an embodiment of the present invention.

FIG. 2 is a flowchart illustrating a disk access operation in the same embodiment.

FIG. 3 is a flowchart illustrating a cache block ejection operation in the same embodiment.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 11... Disk subsystem control mechanism, 12... Disk group, 111... Disk subsystem control device, 112...
Address translation table, 113...Command queue, 1
14...Disk cache memory mechanism.

Claims (2)

[Claims] Claim 1: A disk cache memory comprising: a plurality of disk devices; and a disk cache memory to which storage areas of these disk devices are allocated and holding each allocated storage area address and data of each storage area address as one block. In a disk control device, an address conversion table defining a correspondence between a storage area address designated by a host device and storage areas of the plurality of disk devices, and a storage area address of a block to be ejected from the disk cache memory. means for detecting the presence or absence of load concentration on a disk device specified by the block; writing data to a disk device in which a load is concentrated, comprising address changing means for respectively replacing a predetermined storage area address and a storage area address of the block to be ejected on the address conversion table and the disk cache memory; A disk control device characterized by changing writing to another disk device. 2. Further comprising means for detecting an access frequency of a storage area address allocated to the disk cache memory, wherein the address changing means detects the access frequency of a storage area address allocated to the disk cache memory, and the address changing means detects an access frequency of a storage area address allocated to the disk cache memory. Claim 1, wherein a storage area address with a high access frequency is selected as a target for replacement with the storage area address of the block to be ejected.
Disk controller as described.