JPH08185272A - Magnetic disk device with disk cache - Google Patents

Abstract

PURPOSE: To make possible faster access by prereading data on the basis of prepared preread address information. CONSTITUTION: A read command is received first through a host data bus and when an address register 5 is effective (e bit), the length of the address of this command is recorded in a corresponding entry of an address table 3. For the processing of this command, a cache memory 4 is searched through a table managing mechanism 4 and data is transferred from the cache memory 2 when present on the cache or from the disk device 1 when not. Thus, the address and length of this command are recorded in the address register 5, and the address table 5 is searched by using the address of this command as a key to obtain a next access address and length information. Then when the obtained data are effective, the data transfer from the disk device 1 to the cache memory 2 is performed according to the address and length information to make a preread.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a magnetic disk device having a disk cache.

[0002]

2. Description of the Related Art Magnetic disk devices are frequently used as large-capacity external storage for computers. Peripheral devices have become faster as computers have become more sophisticated, and high-speed access to magnetic disks becomes possible by using a disk cache.

The disk cache stores the data in a memory that can be accessed at high speed on the assumption that the previously used data is reused, and when a request is made again, the data is not stored in the disk but in the memory. By reading the data, it is a method to perform high-speed access.

The drawback of this disk cache is that it is relatively ineffective for random access, although it depends on the algorithm. Furthermore, in order to obtain a high hit rate, it is necessary to increase the memory capacity, resulting in an expensive system.

[0005]

As described above, the conventional magnetic disk device using the disk cache is
The hit rate decreases for data access with a special sequence, and the intended purpose is not achieved. Further, there is a drawback that a large capacity memory is required to increase the hit rate, resulting in an expensive system.

The present invention has been made in view of the above circumstances,
An object of the present invention is to provide a magnetic disk device having a disk cache that realizes further high-speed access by pre-reading data based on pre-reading address information prepared in advance.

[0007]

According to the present invention, a magnetic disk having a disk cache which receives an access command, transfers data from a cache memory if necessary data is present in a cache memory, and transfers necessary data from a disk device if not. In the disk device, an address table that records the address to be accessed next for each data access unit and the address table that is searched immediately after the access command is executed, and the magnetic disk based on the address recorded there And a memory control device for accessing the device in advance and reading the data in advance and transferring the data to the cache memory.

[0008]

According to the present invention, for each block, the address table for recording the address to be accessed next and its length, and the address table immediately after execution of the read command are searched, and the next reference address and length recorded therein are searched. By providing a mechanism for reading data based on the original read operation and a cache memory for storing prefetched data, the address and length are recorded at the position of the block accessed immediately before on the address table at the time of media access. Immediately after the end of media access, disk access is performed based on the next access address and length information recorded in the start address, and data is stored in the cache memory. As a result, the cache hit rate is dramatically improved especially when a certain pattern is provided during read access.

[0009]

An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing the configuration of an apparatus according to an embodiment of the present invention. In FIG. 1, reference numeral 1 is a magnetic disk device. The magnetic disk medium of the magnetic disk device 1 stores control information such as an OS (operating system) in addition to user data.

Reference numeral 2 is a cache memory, which is composed of a high-speed accessible memory for temporarily storing read data. Reference numeral 3 is an address table. A valid bit (e), an address that is likely to be accessed next (next address), length information, and the like are allocated and stored in each entry that constitutes the address table 3. Entries 0 to n exist for the total number of blocks.

Reference numeral 4 is a table management mechanism, which is hardware for updating and searching the address table 3. Reference numeral 5 is an address register. The address register 5 is a storage area for recording and holding the address and length information for each disk access.

FIG. 2 is a flow chart for explaining the operation in the embodiment of the present invention. In the figure, A is a step for updating the address table 3, and the address and length of this command are recorded in the entry corresponding to the address held in the immediately preceding processing step B.

B is a step of holding the start address, and the address and length are stored in the next address table 3
Keep for updates. C is a step of searching the address table 3, and acquires the next access address and length in the table by using the start address as an index.

D is a step of pre-reading, and the data is pre-read from the disk device 1 to the cache memory 2 based on the information obtained in step C. The operation of the embodiment of the present invention will be described below with reference to the drawings.

First, the cache memory 2, the address register 5, and the address table 3 are initialized. Next, a read command is received via the host data bus, and if the address register 5 is valid (e bit), the address and length of this command are recorded in the corresponding entry of the address table 3.

Then, as the processing of this command, the cache memory 2 is searched through the table management mechanism 4,
If the data is on the cache, it is transferred from the cache memory 2, and if not, it is transferred from the disk device 1.

Then, the end of this command is reported to a host device (not shown). A feature of the present invention is to record the address and length of this command in the address register 5, search the address table 3 using the address of this command as a key, and obtain the next access address and length information.

If the acquired data is valid, the data is transferred from the disk device 1 to the cache memory 2 according to the address and the length information, and the prefetch is executed.

Hereinafter, except for the initialization process, the operations after the read command reception process will be repeatedly described. In the above-described embodiment, the read command is mainly described, but when the next access address is written in the address table 3, it is valid even if the access immediately before is a media access command other than read (for example, write). is there. For example, in the case of repeated accesses such as write, read, write, read ..., a method of performing a read access that will be performed after a certain write in advance and storing data in the cache memory 2 is also effective. There is.

In the initialization of the address table 3,
There is also a method of invalidating all the entries, but for example, a method of storing the address table 3 on the disk medium every time and reading out the stored table at the time of initialization can be considered.

The memory size required for the address table is as follows. (1) Entry size = (Number of bytes required for valid bits) x (Number of bytes required to hold address) x (Number of bytes required to hold length) (2) Required memory size = Total number of blocks in disk device x Entry size Of course, it is advantageous in terms of search speed to make the entry size a power of two.

The address table 3 can be multiplexed. For example, A + B + C, A + B '+ C, A +
Considering the case of accessing B + C, A + B ′ + C, etc., the data of B and the data of B ′ are read alternately. Therefore, if B is read next when A is read, the next access will never occur. Will not hit. At this time, if the prefetch information for A is multiplexed and B and B ′ are alternately pointed, the hit rate is improved. It is also conceivable to use the information of not more than one as the key of the read-ahead information, but not the information before two. For example, if you access A + B, read C next, and if you access A + B ', next read D.
Try to read. Further, a method may be considered in which the address table 3 is composed of a ring buffer as the ultimate read-ahead information, and the access pattern in a certain section is searched from the access patterns before that and returned there. Recently, it is often the case that the disk device is duplicated. In this case, if the access to the disk is alternately switched, the overhead for prefetching is reduced.

[0023]

As described above in detail, according to the present invention, when a medium is accessed, its address and length are recorded at the position of the block accessed immediately before on the address table, and immediately after the end of the medium access. It is characterized by performing disk access based on the next access address and length information recorded in the start address and storing the data in the cache memory. If you have a certain pattern, the hit rate of the cache will be dramatically improved, which will further contribute to high-speed processing.

[Brief description of drawings]

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

FIG. 2 is a flowchart showing a processing procedure of the above embodiment.

[Explanation of symbols]

1 ... Disk device, 2 ... Cache memory, 3 ... Address table, 4 ... Table management mechanism, 5 ... Address register.

Claims (4)

[Claims] 1. A magnetic device having a disk cache which receives an access command, transfers data from the cache memory when necessary data is present in the cache memory, and transfers necessary data from a disk device when necessary data is not present in the cache memory. In the disk device, the address table that records the address to be accessed next for each data access unit and the address table that is searched immediately after the access command is executed, and the magnetic field is recorded based on the address recorded in the table. A magnetic disk device having a disk cache, comprising: a memory controller for accessing the disk device in advance to read data and transferring the data to a cache memory. 2. A magnetic disk device having a disk cache according to claim 1, wherein the previously accessed address and length information are recorded in each data access unit. 3. At the time of media access, the address and length are recorded at the position of the block accessed immediately before on the address table, and immediately after the media access is completed, the next access address recorded at the start address is recorded. A magnetic disk device having a disk cache characterized in that disk access is performed based on length information and data is stored in a cache memory. 4. An address table for recording an address to be accessed next is configured by a ring buffer for each data access unit, an access pattern in a certain section is searched from access patterns before that, and the process returns to that. 4. A magnetic disk device having a disk cache according to claim 3.