JPH0566998A - Disk cache system - Google Patents

Abstract

PURPOSE:To eliminate an influence caused by a cache fault to access to a non-store-in processing object by limiting a track address range to inhibit the access when the fault is generated at a cache unit containing unupdated data. CONSTITUTION:This is an external storage sub system loading a cache mechanism, the store-in processing object range is stored in a memory 22 shared by a memory director 21 arranged separately from a cache memory and in the case of the cache fault, the access inhibit range is limited according to this memory. Further, the access inhibit range is limited by storing the presence/ absence of unupdated data. Then, the memory is erased according to a command from a host system 10.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a disk cache system, and more particularly to a disk cache system equipped with a store-in function of writing data only to the disk cache and ending the processing when a write instruction is issued from a host.

[0002]

2. Description of the Related Art Motivation to equip a disk cache system with a store-in function is increasing for the purpose of providing a high-speed access environment for writing.

The store-in function is a mechanism that, in response to a data write command from the host system, reports the end when writing to the cache memory and then writes from the cache memory to the magnetic disk device. ..

In the process of the store-in function, in principle, unupdated data whose writing to the cache memory has been completed and whose writing to the magnetic disk device has not been completed is temporarily stored in the cache memory. ..

A large-scale disk cache system has access paths to a plurality of magnetic disk devices,
Generally, a storage director is arranged on each access path to share one cache unit, and the power supply of the cache unit is installed independently of the storage director.

In the conventional store-in function, management information such as a magnetic disk device address and a track address where unupdated data is to be recorded is stored as a part of directory information for managing the data arrangement in the cache memory. ..

[0007]

When a cache unit including unupdated data cannot retrieve the information in the cache unit due to a power failure, a fatal failure in the cache unit, or the like, Not only the unupdated data but also the management information such as the magnetic disk device address and the track address including the old data are lost.

Therefore, in order to prevent the host system from erroneously processing old data on the magnetic disk device as correct data, it is necessary to prohibit access to all magnetic disk devices in the cache system.

On the other hand, the track range of the store-in control target in the host system coincides with the file layout range for the purpose of emphasizing high-speed writing performance.

An object of the present invention is to provide a disk cache system that solves the above problems.

[0011]

In order to achieve the above object, a disk cache system according to the present invention is an external storage subsystem connected to a host processor and including a magnetic disk device and a cache memory, Store the store-in control target range in the memory shared by the storage director installed separately from the cache memory, and when the cache memory becomes inaccessible due to a failure, prohibit the access to the store-in control target range based on the storage. It has a function to do.

The presence or absence of unupdated data is stored for each store-in control target range, and when the cache memory becomes inaccessible due to a failure, the store-in control target range including the un-updated data is stored based on the storage. It has a function of prohibiting the access of.

Further, it has a function of erasing the memory according to a command from the host system and permitting subsequent access to the relevant range.

[0014]

In the external storage subsystem equipped with the cache mechanism as shown in FIG. 1, the store-in processing target range is stored in the memory 22 shared by the storage director 21 arranged separately from the cache memory, and when the cache fails, the storage is performed. Restrict the access prohibited range. Furthermore, the presence or absence of unupdated data is stored to limit the access prohibited range. The memory is erased by a command from the host system.

[0015]

The present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram of a system including a disk cache system embodying the present invention.

In FIG. 1, the application program 11 running on the host system 10 has a unique file name and a relative storage position of the desired data in the file when accessing the data in the disk file. An access request is sent to the file management unit 12, and the file management unit 12 obtains the address, track address, and record address of the magnetic disk device 24 in which the corresponding data is stored. In this way, the physical recording position of data is determined by the file management unit 12
Is managed by.

In general, when the application program 11 starts to use a disk file and when it finishes using the disk file, a file open / close declaration is made to the file management unit 12, respectively.

The application program 11 notifies the file management unit 12 when the file is opened whether or not the file is to be the store-in processing target.

In addition to the store-in processing attribute, the cache processing attribute includes a store-through processing attribute for the purpose of improving only read performance. The reading of the track range in which the processing attribute is specified is performed at high speed using the cache, but the writing is performed to the magnetic disk device 24, and at the end of the write command, the disk device 24 which is a non-volatile storage medium. Writing to is guaranteed to be complete.

The application program 11 notifies the file management unit 12 whether or not the file is to be the store-through processing target when the file is opened / closed, like the file to be stored-in processing.

The file management unit 12 sends a cache processing attribute code 3 to the disk cache system 20 via the input / output control unit 13 when a file targeted for store-in processing or a file targeted for store-through processing is opened.
2. The extent information (see FIG. 2) including the magnetic disk device address 31, the start track address 33 and the end track address 34 of the store-in process is notified.

The disk cache system 20 is composed of a plurality of storage directors 21, a cache unit 23, and a plurality of magnetic disk devices 24. Each storage director 21 is equipped with a shared memory 22 for storing exclusive control management information of resources shared by the storage director 21. The shared memory 22 is a disk that is promptly restored when another storage director 21 is blocked due to a failure or the like, or when the blocked storage director 21 recovers its function asynchronously with the operation of the storage director 21 in operation. The functions of the entire cache system 20 are mounted on each storage director 21 so that the functions can be restored, and while both storage directors 21 are functioning, the update data from one storage director 21 is shared by both storage directors 21.
Written in.

Since exclusive control management information is stored in the shared memory 22, the storage director 21 that has failed to access the shared memory 22 due to a failure cannot help but be blocked.

If the storage director 21 that receives the extent information notified from the file management unit 12 at the time of opening the store-in processing target file or the store-through processing target file is the extent information having the store-in processing attribute. If the extent information having the store-through processing attribute is stored in both the cache unit 23 and the shared memory 22, it is stored only in the cache unit 23.

In the cache unit 23, a track map is created from the extent information in order to quickly determine whether or not a read / write command for a certain track is a cache processing target. The track map consists of 1-bit information for each track and is "1".
Indicates that the cache processing is targeted, and "0" indicates that the cache processing is not targeted.

In the extent information storage area 30 on the shared memory 22, in addition to the extent information of the store-in processing attribute notified from the file management unit 12, an unupdated track counter 35 is stored for each extent information. The storage director 21 uses the unupdated track counter 3
In 5, the number of unupdated tracks in the applicable track range is stored at any time. No unupdated data exists in the track range where the content of the unupdated track counter 35 is zero.

The storage director 21 is the host system 1
When a read / write command is received from 0, it is determined by referring to the track map whether or not the track for which this command is issued is a cache processing target.

All information references to cache operations for executing read / write commands while cache unit 23 is available are all in cache unit 23.
It is done using the information in. This is because the shared memory 22 stores exclusive control management information of shared resources, and therefore frequent reference to the shared memory causes a delay in exclusive control management processing, which reduces throughput of the cache system as a whole. Because they are connected.

When a failure occurs in the cache unit 23 and the storage director 21 cannot refer to the information in the cache unit 23, the storage director 21 that has received the read / write command receives the shared memory 22.
If the track for which this command is issued is within the store-in processing target range by referring to the extent information loading area 30 in the above, and if it is within the store-in processing target range, the presence or absence of un-updated data by the non-updated track counter To determine.

If the data is out of the store-in processing target range, or if it is within the store-in processing target range and does not include unupdated data, the data on the magnetic disk device 24 is the latest data, so read / write Writing can be normally processed using the information on the magnetic disk device 24.

If it is within the range of the store-in processing and further includes unupdated data, the magnetic disk device 24 within the track range indicated by this extent information.
The data on one of the tracks above is the one before the update. In this case, the storage director 21 reports an abnormal end to the read / write command and prohibits access to the host system 10.

The file management unit 12 performs a blocking procedure for the file that has received the abnormal termination report, and the application program 11 that has used this file aborts.

Thereafter, the blocked file is once erased in the process of performing the file recovery by the same means as the file recovery process after the file blockage caused by the magnetic disk failure, for example. The file management unit 12 issues a command for deleting extent information corresponding to this file to the disk cache system 20 at the time of file deletion, and the storage director 21 receiving this command causes the cache unit 23 and the extents in the shared memory 22 to be deleted. The extent information is erased from the information storage area 30. By deleting the extent information, the prohibition state of the read / write command to the corresponding track range is released.

[0035]

As described above, according to the present invention, when a failure occurs such that the cache unit becomes uncontrollable while containing unupdated data, a file or store with the store-in processing attribute specified is stored. It is possible to place only the file with the in-process attribute specified and containing the unupdated data at the time of the cache failure in the access prohibited state. That is, the data access to the non-cache processing target file or the store through processing target file is not affected by the cache failure.

The file access prohibition state is
Since the cancellation is performed in synchronization with the file erasing operation in the file management unit, a special operator operation for canceling the file access prohibition state is unnecessary.

[Brief description of drawings]

FIG. 1 is a block configuration diagram of a system including a disk cache system embodying the present invention and a disk cache system.

FIG. 2 is a diagram showing information arrangement of an extent information storage area in a shared memory.

[Explanation of symbols]

 10 host system 11 application program 12 file management unit 13 input / output control unit 20 disk cache system 21 storage director 22 shared memory 23 cache unit 24 magnetic disk device 30 extent information storage area 31 magnetic disk device address 32 cache processing attribute code 33 start track Address 34 End track address 35 Unupdated track counter

Claims (3)

[Claims] 1. An external storage subsystem connected to a host processor, comprising a magnetic disk device and a cache memory, wherein a store-in control target range is stored in a storage director shared memory provided separately from the cache memory. However, when the cache memory becomes inaccessible due to a failure, the disk cache system has a function of prohibiting access to the store-in control target range based on the storage. 2. The disk cache system according to claim 1, further stores the presence or absence of unupdated data for each store-in control target range, and stores the data when the cache memory becomes inaccessible due to a failure. A disk cache system having a function of prohibiting access to a store-in control target range including unupdated data based on storage. 3. The disk cache system according to claim 1 or 2, further comprising a function of deleting the storage according to a command from the host system and permitting subsequent access to the relevant range. And a disk cache system.