JPH0644139A - Disk cache system and its page rewrite control method - Google Patents

Abstract

PURPOSE:To reduce cost in the protection of data when an accident occurs by distributing an access destination to a first cache memory in accordance with a read request and to a second cache memory in accordance with a write request. CONSTITUTION:The first cache memory consisting of a comparatively inexpensive volatile-semiconductor memory 7, which stores read data in a central processing unit 1 and the second cache memory consisting of an expensive non-volatile memory 8 storing write data are used. A memory control part 6 distributes the access destination to the first cache memory in accordance with the read request from the central processing unit 1 and the access destination to the second cache memory in accordance with the write request. Thus, the use quantity of the expensive non-volatile semiconductor memory 8 is supressed to a necessary minimum and an unmatched state on account of unmatch between data of the cache memory and data of a disk 4 owing to the accident of a power cut or the like is avoided.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a cache control technique in a computer system, and more particularly to a disk cache system and its page replacement suitable for avoiding the inconsistent state of the data in the cache memory and the disk at low cost. It relates to a control method.

[0002]

2. Description of the Related Art In a computer system, a semiconductor memory called a cache memory is provided between an external storage such as a magnetic disk and a main storage, and a CPU (Cent) is provided.
A memory control system (hereinafter, referred to as a disk cache system) that fills a gap in access time between the two by a real processing unit (central processing unit) has been developed. In the disk cache system, generally, the cache memory is provided in the magnetic disk controller.

Since the access time of a magnetic disk capable of storing a large capacity is slower than the access time of a main memory, a cache memory is provided to fill such a gap in the access time and store a large amount of data. It enables high-speed access. That is, only the frequently accessed data is stored in the cache memory,
Execution performance can be improved.

Regarding such a technique for replacing the data stored in the cache memory, an LRU (Least)
There is a Recent Used algorithm. This LRU algorithm uses a general property (locality of a program) that a recently accessed page is more likely to be accessed in the near future than a storage area of previously accessed data, that is, a page. It removes (pages out) the page that was accessed in the farthest past and creates a free space in the cache memory. Regarding such a cache memory,
For example, it is described on pages 1705 and 1789 of "Electronic Information and Communication Handbook" edited by the Institute of Electronics, Information and Communication Engineers (1988, published by Ohmsha).

However, in the conventional disk cache system, a volatile semiconductor memory is generally used as the cache memory. For example, when the power is cut off,
When the power supply is cut off, the memory contents (data)
Disappears. For this reason, if there is data in the cache memory that has not yet been written to the magnetic disk and it disappears, the data in the cache memory and the data in the disk do not match, resulting in an inconsistent state.

There is a cache memory which uses a non-volatile semiconductor memory so as to solve such problems and protect data even in the event of power failure. However, the non-volatile semiconductor memory is more expensive than the volatile one, and in order to back up a large amount of data, a large-capacity semiconductor memory must be provided, which increases the cost of the product.

[0007]

The problem to be solved is that in the conventional technique, the cache memory uses only one of volatile and non-volatile semiconductor memories. This is a point that the data in the cache memory cannot be efficiently protected at low cost when an accident such as the above occurs. An object of the present invention is to solve these problems of the prior art and to efficiently erase the data in the cache memory due to power failure etc. at low cost and improve the reliability and performance of the computer. A disk cache system and a page replacement control method therefor are provided.

[0008]

In order to achieve the above object, the disk cache system of the present invention comprises: (1) Transferring and storing frequently accessed data from an external storage device into a provided cache memory; In a disk cache system for controlling access to data of a central processing unit via the cache memory, a first cache memory composed of a volatile semiconductor memory for storing data for reading by the central processing unit, and a central processing unit In response to a read request from the central processing unit, a second cache memory formed of a non-volatile semiconductor memory for storing data for writing of the data is written to the first cache memory as an access destination of the central processing unit. Also, in response to a write request from the central processing unit, the access destination of this central processing unit Wherein the the second cache memory provided with sorting Ru memory controller. Also,
(2) In the page replacement control method for a disk cache system according to the present invention, in the disk cache system according to (1) above, the memory control unit determines a predetermined second cache memory of the second cache memory requested by the central processing unit. Replace the page, which is the data storage area of the size, with L
When the page replacement request is made based on the RU and occurs for the unwritten back page having unwritten back data that does not match the data of the external storage device continuously for a preset number of times, this unreturned page It is characterized in that all the unwritten pages including the written back page are written back to the external storage device and then the requested page is replaced.

[0009]

In the present invention, a relatively inexpensive volatile semiconductor memory is used as the cache memory for reading data (first cache memory), and the cache memory for writing data (second cache memory) is used. memory)
For this, an expensive non-volatile semiconductor memory is used. That is, even if the data for reading disappears due to power failure,
Since the write data does not affect the inconsistent state of the disk, a nonvolatile semiconductor memory is used to protect the data in the cache memory. As a result, the usage amount of the expensive non-volatile semiconductor memory can be suppressed to the necessary minimum, and the inconsistent state due to the inconsistency between the data in the cache memory and the data in the disk due to an accident such as power failure can be avoided. Also,
When the replacement of the unwritten back page having unwritten back data that does not match the data of the external storage device with respect to the second cache memory continues for a predetermined number of times, the external storage device of all unwritten back pages Write back to the group at once. As a result, it is possible to reduce the performance deterioration due to the cache memory being divided into two.

[0010]

Embodiments of the present invention will now be described in detail with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of a configuration relating to the present invention of a disk cache system of the present invention. In FIG. 1, reference numeral 1 is a CPU that performs data processing such as arithmetic processing, and sends control information to a peripheral device (not shown), and 2 is a main memory that stores data to be processed by the CPU 1 and that is directly accessed by the CPU 1. Reference numeral 3 is a cache control device according to the present invention, which controls storage of frequently accessed data, and 4 is a magnetic disk device as an external storage device for storing a large amount of data.

The cache control device 3 controls the operation of the host interface 5 for controlling data exchange with the CPU 1 and the main memory 2 and the cache control device 3 as a whole, and also relates to the present invention and read data. A memory control unit 6 that controls distribution of write data to the cache memory and page replacement control based on the LRU algorithm, and a volatile semiconductor as the first cache memory of the present invention that stores read data. A memory 7, a nonvolatile semiconductor memory 8 as a second cache memory of the present invention for storing data for writing, and a disk interface 9 for controlling data exchange with the magnetic disk device 4 and the like.
It is composed of and.

With this configuration, in the disk cache system of this embodiment, the data once read and written by the magnetic disk device 4 by the CPU 1 is transferred to the cache memory of the cache control device 3, that is, the volatile semiconductor memory 7. The data is stored in the non-volatile semiconductor memory 8 to minimize data input / output with the magnetic disk device 4. That is, when the CPU 1 accesses again, by accessing the volatile semiconductor memory 7 and the non-volatile semiconductor memory 8 instead of the magnetic disk device 4, it is possible to access the magnetic disk device 4 having a long access time. Unnecessary. As a result, data can be exchanged at high speed.

In this embodiment, in particular, the cache memory is divided into the volatile semiconductor memory 7 for storing the read data and the non-volatile semiconductor memory 8 for storing the write data. As described above, by using the nonvolatile semiconductor memory 8 for writing data, it is possible to avoid the disappearance of the data in the cache memory even in the event of a power failure, for example. It is possible to prevent the occurrence of the disc inconsistent state.

The memory controller 6 controls the page replacement of the cache memory by using the LRU algorithm. That is, when replacing a page which is a storage area unit of read data, the memory control unit 6 replaces the page stored in the volatile semiconductor memory 7 for the longest time by the LRU. At this time, when another process is using the page stored for the longest time, the page is signed with "busy". Replace pages that are stored long in.

When replacing a page for writing, the memory controller 6 replaces the page stored for the longest time from the nonvolatile semiconductor memory 8 by the LRU. At this time, if this page is an unwritten back page that has not been written back to the disk of the magnetic disk device 4, it must be written back to the disk before replacement. If all the pages in the semiconductor memory 8 are unwritten back pages, it is necessary to write back the unwritten back pages to the disk every time the writing page is replaced.

Generally, if data is written back to the disk every time a page is replaced, input / output with the magnetic disk device 4 frequently occurs and the processing performance of the system is deteriorated. Therefore, when writing back unwritten pages to the disc, it is effective to write back as many unwritten pages as possible. But on the other hand,
In order to effectively use the cache memory function, it is necessary to keep the unwritten pages in the cache memory as long as possible.

In order to satisfy these two contradictory requirements, the memory control unit 6 of the present embodiment makes it possible to replace the unwritten pages with a page replacement request a predetermined number of times in succession. As shown in FIG. 2 below, all unwritten back pages are collectively written back to the disc.

FIG. 2 is a flow chart showing an embodiment of the page replacement control operation according to the present invention of the disk cache system in FIG. If there is a page replacement request (step 201), for reading (Rea)
It is determined whether or not the page (for d) is replaced (step 202). For reading, the oldest page of the volatile semiconductor memory 7 in FIG. 1 is selected (step 2).
03). If the selected page is used by another process (busy in the figure) (step 204), the oldest page not used by another process is selected (step 205) and the page is replaced (step 205). Step 20
6).

In step 202, if the page is to be replaced not for reading but for writing, the oldest page of the nonvolatile semiconductor memory 8 in FIG. 1 is selected (step 207). If this page is not the unwritten back page (step 208), this page is replaced (step 209). If this page is an unwritten back page, it is confirmed whether replacement of the unwritten back page has been requested twice in succession (step 210). If it is not continuous, this unwritten page is written back to the disk of the magnetic disk device of FIG. 1 (step 211) and the page is replaced (step 209).
Further, in step 210, if the replacement of the unwritten back page is requested twice consecutively, all the remaining unwritten back pages of the nonvolatile semiconductor memory 8 in FIG. The data is written back to the disk of the disk device (step 212), and then the requested page is replaced (step 209).

Thus, for example, when a page replacement request of the non-volatile semiconductor memory 8 is issued twice in succession and the page to be replaced is an unwritten return page, "the writing is continued thereafter. Based on the hypothesis that "there will be continuous", it is preferable to process all the unwritten back pages at once rather than performing the writing back process of the unwritten back pages every time. The reason is that the processing performance related to write-back is not deteriorated, and that if the writing continues after that, there is no reference to the cache memory, so even if all unwritten-back pages are processed, the performance does not deteriorate. Originally, all unwritten back pages in the nonvolatile semiconductor memory 8 in FIG. 1 are written back to the disk of the magnetic disk device 4 in FIG.

As a result, the unwritten pages are wiped out of the semiconductor memory 8 of FIG. 2, and thereafter, in the continuous writing process, the input / output process by the disk interface 9 in FIG. 1 is not performed for each process. It is possible to suppress deterioration of performance.

As described above with reference to FIGS. 1 and 2, in the disk cache system of this embodiment, the cache memory is provided separately for reading and writing. In particular, by using a non-volatile semiconductor memory for writing, it is possible to avoid a disc inconsistent state due to the disappearance of data due to an accident such as power failure. Moreover, the cost can be suppressed by using an inexpensive volatile semiconductor memory for reading. Further, at the time of replacing the page for writing, the write-back process of the unwritten page to the disk can be efficiently performed, the input / output operation of the disk can be suppressed as much as possible, and the input / output performance can be improved.

The present invention is not limited to the embodiment described with reference to FIGS. 1 and 2. For example,
In step 210 of FIG. 2, it is confirmed whether or not the replacement for the unwritten page is requested twice in succession, but the number of times may be three or four, and is limited to two. Not a thing.

[0024]

According to the present invention, the data in the cache memory can be efficiently protected at low cost in the event of an accident such as power failure, and the reliability of the disk cache system can be improved. , It is possible to improve the performance.

[0025]

[Brief description of drawings]

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

FIG. 2 is a flowchart showing an example of a page replacement control operation according to the present invention of the disk cache system in FIG.

[Explanation of symbols]

 1 CPU 2 Main Memory 3 Cache Control Device 4 Magnetic Disk Device 5 Host Interface 6 Memory Control Section 7 Volatile Semiconductor Memory 8 Nonvolatile Semiconductor Memory 9 Disk Interface

Claims (2)

[Claims] 1. A disk cache system in which data frequently accessed from an external storage device is transferred and stored in a cache memory provided, and access control of data of a central processing unit is performed via the cache memory. A first cache memory composed of a volatile semiconductor memory for storing the read data of the central processing unit, and a second cache memory composed of a non-volatile semiconductor memory for storing the write data of the central processing unit. In response to the read request from the central processing unit, the access destination of the central processing unit is set to the first cache memory, and the access to the central processing unit is performed in response to the write request from the central processing unit. A memory control means for allocating the destination to the second cache memory is provided. Squash system. 2. The disk cache system according to claim 1, wherein the memory control means is a page which is a data storage area of a predetermined size of the second cache memory requested by the central processing unit. Replacement is performed based on the LRU, and the replacement request for the page is continuously generated a preset number of times for an unwritten back page having unwritten back data that does not match the data in the external storage device. The page of the disk cache system is characterized in that the requested page is replaced after all the unwritten pages including the unwritten page are written back to the external storage device. Replacement control method.