JPH07248973A - Back-up processing method - Google Patents

Abstract

PURPOSE:To fast back up the cache data in a write-back cache system. CONSTITUTION:When a power supply is cut off, the block data on a block address 0 of a cache 12 are read out and written in a block address (k) of a disk 22 showing the head of an area A2. Then the block data on a block address 1 of the cache 12 are read out and written in a block address (k+1) of the disk 22. These operations are repeated thereafter. When a power supply is applied again, the block data on the address (k) of the disk 22 are read out and written in the address 0. Then the block data on the address (k+1) of the disk 22 are read out and written in the address 1 of the cache 12. These operations are repeated thereafter.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a backup processing method for high-speed backup of cache data in a write-back cache system.

[0002]

2. Description of the Related Art FIG. 3 is a diagram for explaining a conventional cache control method. In the figure, 10 is a computer main unit, 11 is a main memory, 12 is a cache, 13 is an input / output processor, 20 is a disk device with cache, 21 is a read-ahead cache, and 22 is a disk device.

The main unit 10 has a main memory 11, a cache 12, an input / output processor 13, and the like. Although not shown, the main body device 10 has an instruction execution processor. The memory having the main memory 11 and the cache 12 is connected to the input / output processor 13 and the instruction execution processor via the system bus. The input / output processor 13
It executes the disk control firmware and manages the past access history.

The disk device with cache 20 has a read-ahead cache 21 and a disk device 22. The read-ahead cache 21 has a capacity of 240, for example.
It is KB. Although not shown, the cached disk device 20 is connected to the system bus via a SCSI bus or a SCSI controller.

When the instruction execution processor issues a read request to the disk device, the input / output processor 13
Checks whether or not the corresponding data exists in the cache 12, and if there is, the corresponding data is written in the main memory 11. When the data specified by the read request does not exist in the cache 12, the input / output processor 13 refers to the past access history and checks whether it is an irregular access or a sequential access.

In the case of random access, the input / output processor 13 requests the cached disk device 20 to read the data requested by the instruction execution processor, and caches the data sent from the cached disk device 20. 12 and the main memory 11 is written. In the case of sequential access, the input / output processor 13
The data requested by the instruction execution processor and the data for prefetching are requested to the cached disk device 20, the data sent from the cached disk device 20 is stored in the cache 12, and the data requested by the instruction execution processor is requested. Write the data in the main memory 11.

The operation when a write request for writing the data in the main memory to the disk device with cache is issued from the instruction execution processor will be described. As a writing method, write through and write back are known. First, the write-through method will be described.

In the case of write through, the input / output processor 13 reads the data specified by the write request from the instruction execution processor from the main memory 11, writes the read data in the cache 12, and reads the read data from the disk device 22. Write to the specified address of. In the case of write back, the input / output processor 13 stores the data specified by the write request from the instruction execution processor in the main memory 1.
1 is read, and read data is written in the cache 12. In the case of write back, data is not written in the disk device 22.

FIG. 4 is a diagram for explaining the principle of the write-back cache. In the case of write back, when a write request is issued from the software, the I / O processor executes the corresponding firmware,
The main memory data specified by the write request is written to the cache, and when the writing is completed, the end of writing is reported to software.

In the case of write-through, when a write request is issued from the software, the input / output processor executes the corresponding firmware to write the data in the main memory specified by the write request to the cache and also to the cache. When a write command is sent to the attached disk device and the end status is sent from the cached disk device, the write end is reported to the software.

FIG. 5 is a diagram for explaining a cache data backup process in a conventional write-back cache. When the power is turned off, the cache data is backed up in the disk device. In the example shown, the cache has four rows, and each row can store four blocks. 0 rows in cache 0
The number 0 is written in the column block, but this means that the block 0 block data of the cache has a disk block address of 0, and this data has not yet been backed up to the disk device. It is shown that. Similarly, the number 0, 1 is written in the 0th row and 1st column of the cache.
The disk block address of the data in the block of columns is 1, indicating that this data has not yet been backed up to the disk device.

At the time of backup, the block data in the 0th row and 0th column of the cache is written into the block of the disk block address 0 of the disk device, and then the block data in the 0th row and 1st column of the cache is written in the disk device. Write to the block at disk block address 1,
Next, the block data of the 0th row and the 2nd column of the cache is written to the block of the disk block address 10 of the disk device, and then the block data of the 0th row and the 3rd column of the cache device is written to the disk block address 3 of the disk device.
Write in the block. Hereinafter, the same operation is performed.

[0013]

[Problems to be Solved by the Invention] Conventional write back
In the backup process in the cache system,
The cache data is written block by block in the corresponding disk device, but the access at this time is a random access and it is necessary to move the disk head frequently, which is very useful for backup processing. Was spending time. The present invention was created in view of this point, and an object of the present invention is to enable high-speed backup processing of cache data in a write-back cache system.

[0014]

[Means for Solving the Problems] and therefore claim 1
The backup processing method of (1) comprises a disk device and a cache that holds a part of the data stored in the disk device, and when a write request instructing to write the main memory data to the disk device is issued, In a computer system that performs write-back processing that writes the main memory data to the cache and does not write to the disk device, when the power is turned off, the contents of the cache are read in order according to the cache address, and the read data is written to the disk address. According to the method, data is written in a predetermined area of the disk device in order.

A backup processing method according to a second aspect of the present invention is the backup processing method according to the first aspect, in which, when power is cut off, a eviction management table indicating which data in the cache has been evicted to the disk device, and the disk device. The bad sector management table showing the bad sectors and their replacement sectors is also saved in the disk device.

[0016]

DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 is a diagram for explaining one embodiment of the present invention. In the figure, 10 is a computer main unit, 11 is a main memory, 12 is a cache, 13 is an input / output processor, 20
Is a disk device with a cache, 22 is a disk device,
A1 is a data area, A2 is a save area, R1 is a route for the eviction process, R2 is a route for the asynchronous eviction process, R3 is a route for the data substitution process, and R4 is a route for the high-speed backup process. Note that the read-ahead cache is not shown in FIG.

The disk control firmware of the I / O processor 13 is executed. In the area for the disk control firmware, there are also an ejection management table and a bad sector management table.

Information indicating which block of the cache 12 has been ejected to the disk device 22 is entered in the ejection management table. The block data of the cache corresponding to the shaded area of the flush management table has been flushed to the disk device 22. For example, the disk block of the block data of the 0th row and 0th column of the cache
Although the address is 4, this block data has been evicted to the block of disk block address 4 of the disk device. The bad sector management table shows bad sector addresses and alternative sector addresses of the disk device.

In the illustrated example, the block data of the disk block addresses 0, 4, 8, 12, 16, ... Of the disk device 22 are 0 row 0 column, 1 row 0 of the cache.
It is written in columns, 2 rows and 0 columns, and 3 rows and 0 columns. Similarly, the disk block addresses 1, 5 of the disk device 22 are
The block data of 9, 13, 17, ... Are written in the cache at 0 row 1 column, 1 row 1 column, 2 row 1 column, 3 row 1 column. Hereinafter, it is as illustrated.

When a write request to write the block data of the disk block address 0 stored in the main memory 11 is received, the input / output processor 13
Is 0 row 0 column, 1 row 0 column, 2 row 0 column of the cache 12,
It is checked whether or not there is a vacant block at the 3rd row and 0th column, and if there is a vacant block, the designated block data is written to the block. In the illustrated example, since there is no empty block, the block data stored in the 0th row and 0th column of the cache 12 is flushed to the disk device 22 and then specified in the 0th row and 0th column of the cache 12 by a write request. Write block data. In the above description, it is assumed that the block data in the 0th row and 0th column of the cache 12 is the oldest in the column.

The input / output processor 13 finds a vacant time and performs an asynchronous eviction process. In asynchronous ejection processing,
The block data stored in the cache 12 that has not been evicted is ejected to the disk device 22. Data substitution processing is performed when there is a medium failure in a block of a disk to which data is written. In the alternative processing of data, an alternative block that replaces the block of the disk having the medium failure is found, and the block data is written in the alternative block.

High-speed backup processing is performed during power failure (UPS
Operation) or normal power off. Note that UPS is an uninterruptible power supply. The power supply time of the uninterruptible power supply is, for example, 3 minutes. When a power failure occurs, the I / O processor 13 sequentially reads the block data from the cache 12 and reads the read block data.
Data is sequentially written to the save area A2 of the disk device 22.

That is, the 0th row, 0th column, 0 of the cache 12
The cache block address of row 1, column 1, ... Is 0,
1, ..., Row 0 of save area A2 of disk device 12
Column, 0 row, 1 column, ... Disk block address k,
Assume k + 1, .... It should be noted that the block of p rows and q columns of the disk device corresponds to the pth track and the qth sector.

First, the block data of the cache block address 0 of the cache is read and written to the disk block address k of the disk device, and then the block data of the cache block address 1 of the cache is read and the disk is read. Device disk
Write to block address k + 1. Hereinafter, the same operation is repeated.

When the power is turned on, the disk device 22
The data in the cache stored in the save area A2 is sequentially read, and the read data is sequentially written in the cache 12. That is, first, the block data of the disk block address k of the disk device is read and written in the cache block address 0 of the cache, and then the disk data of the disk device is read.
The block data of the block address k + 1 is read and the cache block address 1 of the cache is read.
Written in. Hereinafter, the same operation is repeated.

FIG. 2A shows the processing when the power is turned off. Normally, write-back cache operation is performed. When a power failure occurs, the auxiliary power supply (uninterruptible power supply) is started, then the contents of cache memory are saved to the disk save area prepared beforehand (high-speed backup), and then the eviction management table and bad sector management table are written to the disk. And then turn off the auxiliary power supply. FIG. 2 (b) shows the processing when the power is turned on. When the power is turned on, the contents of the cache memory are restored from the save area of the disk, then the eviction management table and bad sector management table are restored from the disk, and then normal operation is resumed.

[0027]

As is apparent from the above description, according to the present invention, a save area for saving the contents of the cache memory as it is is prepared in advance, and the backup area is backed up when the power is turned off. To do. The access at this time becomes sequential, the seek time and the rotation waiting time of the disk are shortened, and the backup operation faster than the conventional one is realized.

[Brief description of drawings]

FIG. 1 is a diagram illustrating an embodiment of the present invention.

FIG. 2 is a diagram showing a process when power is turned off / on in the present invention.

FIG. 3 is a diagram illustrating a conventional cache control method.

FIG. 4 is a diagram illustrating the principle of write-back cache.

FIG. 5 is a diagram illustrating cache data backup processing.

[Explanation of symbols]

 10 computer main unit 11 main memory 12 cache 13 I / O processor 20 cache disk device 22 disk device

Claims (2)

[Claims] 1. A disk device, comprising: a cache for holding a part of data stored in the disk device; and when a write request instructing to write main memory data to the disk device is issued, In a computer system that performs write-back processing that writes main memory data to the cache and does not write it to the disk device, when the power is turned off, the contents of the cache are read sequentially according to the cache address, and the read data is read according to the disk address. A backup processing method comprising sequentially writing data in a predetermined area of a disk device. 2. The disk also includes a eviction management table indicating which data in the cache has been evicted to the disk device when the power is cut off, and a bad sector management table indicating a defective sector of the disk device and its replacement sector. The device is retracted to the device.
Backup processing method.