JPH1040147A - System controlling virtual file cache - Google Patents

Abstract

PROBLEM TO BE SOLVED: To unnecessitate an exclusive storage area as a cache area by operating the cache of a file piece by using an identifier applied to the file piece. SOLUTION: When a request for file reference is made to a memory page allocating mechanism 18, a cache controlling mechanism 17 checks whether or not the content of a pertinent file is already present in a memory page. When it is already present in the memory page, priority in a page structural body corresponding to it is incremented, and the memory page is provided. When it is not present in the memory page, the memory page with low priority held in the page structural body is selected by a memory page selecting mechanism 16, the content is canceled, the priority is initialized to the minimum value, a virtual identifier corresponding to a new file piece is applied, and this is provided. A side to which the memory page is provided operates actual I/O only when the content is invalid.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a virtual file cache control system.

[0002]

2. Description of the Related Art Conventionally, a memory management function in a computer system includes a mechanism for attaching a file. This file attach mechanism writes the contents of a file read from the file system to the memory in the computer system when there is a reference request for a file stored in the file system, and then refers to the file again. When a request is made, the file can be referred to without performing actual I / O to the file system.

[0003] This file attachment mechanism does not depend on whether the location where the file actually exists is on a distributed file system or whether certain contents belong to the same file. In addition, this attach mechanism has a feature that a dedicated external storage or memory is not required.

However, since the attach mechanism is a very simple mechanism, when the real memory runs short, that is, when the memory is used for another purpose, the order in which files are discarded is determined by the order in which memory pages are released. Are selected in the order in which they were performed. For this reason, if a large amount of file access is performed at one time, even the memory that holds the file contents that were frequently accessed before will be discarded,
There is a problem that the effect of reuse is weakened.

With respect to such an attach mechanism, a cache control method for speeding up file access in a specific case has been proposed.

For example, Japanese Patent Application Laid-Open No. 7-93205 describes a distributed file system that improves the client cache hit rate and speeds up file access. In this distributed file system, the file server has an access frequency database that records a user name and a file name frequently accessed by the user in association with each other, and the client has a login user table that records the logged-in user name; Has a cache priority control module. The cache priority control module sets a priority for each file copy in the cache area based on the contents of the access frequency database and the login user table. Then, a file that is frequently used is preferentially cached for each user who logs in to the client, thereby improving the cache hit rate and speeding up file access.

Japanese Patent Laid-Open Publication No. Hei 3-177946 describes a magnetic disk buffer control system in which a buffer cache memory has a priority (priority) to improve a cache hit rate. In the magnetic disk buffer control method, in a file system of a computer system, a part of the memory is stored in a buffer cache memory for inputting / outputting block data to / from a magnetic disk, and the bribery takes precedence over frequently accessed block data. By doing so, the time required for input / output processing is reduced. The priority is managed in accordance with the block number of the magnetic disk, the priority of the disk block with high access frequency is increased, and the buffer cache corresponding to the block with low priority is preferentially replaced to speed up the file access.

[0008]

The first problem is that unified cache control over the main file systems cannot be performed.

The reason is that the cache control method is limited by the type of the file system, so that it is necessary to be aware of the type of the file system.

A second problem is that a dedicated memory or a secondary storage which can be used only for the purpose as a cache area is required.

The reason is that the cache control mechanism exists independently of the memory management mechanism of the computer system.

[0012] It is an object of the present invention to provide a virtual file cache control method that can uniformly handle all file systems.

Another object of the present invention is to eliminate the need for a dedicated storage area as a cache area by incorporating a cache control mechanism into a memory management mechanism of a computer system.

[0014]

According to the present invention, a file (1 and 2 in FIG. 1) stored in a file system is divided into memory pages and a file fragment (3 in FIG. 1).
(4) and (5), when an identifier is assigned to each of the file fragments, and the cache operation is performed from a computer system connected to the file system, the file fragments are cached using the identifier. Thus, a cache control method is provided in which the cache can be performed irrespective of the type of the file.

In the present invention, a part of the storage area of the memory of the computer system is used as a cache memory area.

According to the present invention, when the file fragment is stored in the memory page corresponding to a memory page of a storage area of the memory, the identifier given to the file fragment and the file There is provided a cache control method including a holding unit for holding a priority that changes according to the number of times of access to a fragment.

In the present invention, file fragments are discarded based on the priority.

Further, according to the present invention, based on an attach mechanism existing in a memory management mechanism of a computer system, cache priority information is added with a virtual identifier added in units of file fragments as a control target. By selecting a page in accordance with the priority information at the time of reassigning the page, a cache control method that enhances the cache effect can be obtained.

Further, according to the present invention, a cache control method is provided which has a cache priority attribute for each file or file system and controls the cache priority.

[0020]

The cache control unit handles files stored in the file system without being aware of the file type by using the virtual storage technology. That is, since the cache management target is virtual information, the cache mechanism does not recognize the type of the actual state of the file. Also, by performing caching based on priority information indicating the access frequency of the file stored in each memory page, file access can be speeded up.

[0021]

Embodiments of the present invention will be described below with reference to the drawings. The virtual file cache system according to the present embodiment is adopted in a computer system connected to a plurality of different file systems.

A first embodiment of the present invention will be described with reference to the drawings.

FIG. 1 shows the correspondence between a file fragment and a structure for managing the file fragment in the virtual file cache system, and includes a distributed file system, a magnetic disk or a magneto-optical disk of a local machine, or a CD-ROM. File 1 existing on the file system
A file 2 existing on a file system of a different type from the file 1, a file fragment 3 and a file fragment 4 which are part of the file when the file 1 is divided into cache units, and a file 2 in a cache unit. A file fragment 5 which is a part of the file at the time of division, a memory 6 of the computer system, a memory page 7 which holds the contents of the file fragment 3 in a part of the memory, and a file fragment 5
, An array 10 of page structures that exist in one-to-one correspondence with all memory pages in the system and manage the state of memory pages, and a page structure corresponding to memory page 7 2 shows a body 11 and a page structure 13 corresponding to the memory page 9. It is assumed that memory pages other than the memory pages 7 and 9 are used for some purpose, and cannot be used as a cache at present.

FIG. 2 shows a part related to the present invention in the information of the page structure 11 in the virtual file cache system, and the memory page managed by the page structure, which is conventionally provided in the memory management mechanism, Identifier 14 that virtually represents whether the content of
Indicates that the memory page has a priority 15 indicating how much priority is given to caching the memory page. The other page structures also hold information on corresponding memory pages.

FIG. 3 shows a memory page selection mechanism 16 for selecting a memory page whose contents are to be discarded from among memory pages currently holding the contents of a file as a cache in the virtual file cache system. , And a mechanism for selecting the memory page with the lowest priority.

FIG. 4 shows that, in the virtual file cache system, when the contents of a file requested by specifying a virtual identifier already exist in a memory page, the corresponding memory page is selected, and at the same time, the corresponding memory page is selected. The cache control mechanism 17 having a function of increasing the priority in the page structure is shown. If the corresponding memory page does not exist, the search fails.

FIG. 5 shows a memory page allocation mechanism 18 in which a memory page selection mechanism 16 and a cache control mechanism are combined in the virtual file cache system. In this mechanism, when an attempt is made to refer to a specific file by designating a virtual identifier, first, the cache control mechanism 17 is used to check whether the corresponding page exists. If the corresponding content already exists in the memory page, it is returned. If not, the memory page selecting mechanism 16 is next used to select a memory page whose content is to be discarded, and finally, the memory page is deleted. Is returned.
The user of this mechanism should actually read from the file when the memory page whose contents are invalid is returned, store the contents in that memory page, and return the memory page where the contents exist Use it as is.

Hereinafter, the operation of the first embodiment of the present invention will be described with reference to the drawings.

When a file reference request is made to the memory page allocating mechanism 18, first, the cache control mechanism checks whether or not the contents of the corresponding file already exist in the memory page. If it already exists in the memory page, the priority in the corresponding page structure is incremented and the memory page is provided. If it does not exist in the memory page, the memory page selection mechanism 16 selects a low-priority memory page held in the page structure, discards the contents, then initializes the priority to the lowest value, and adds a new file. Give a virtual identifier corresponding to the fragment and provide it. The side provided with the memory page need only perform the real I / O when the content is invalid.

In FIG. 1, file fragment 3 and file fragment 5 already exist in memory pages 7 and 9, respectively, and are in a cached state. Here, page structures 11 and 13 for managing respective memory pages
Suppose that the priorities held in are stored in 5 and 2. This indicates that file fragment 3 is referenced more.

Here, when the file fragment 5 is referred to, the memory page allocating mechanism 18 uses the cache control mechanism 17 to store the already existing memory page 9.
Select At the same time, the cache control mechanism 17 increments the priority in the page structure 13 corresponding to the memory page 9, and the priority is updated to 3. Then, the memory page allocation mechanism 18 returns the contents of the memory page 9 as valid. Since the contents of the nine memory pages are valid, the contents of the file fragment 5 can be referred to without actually performing I / O.

Next, when a reference is made to the file fragment 4, the memory page allocation mechanism 18 checks whether or not the file fragment 4 exists in the memory page by using the cache control mechanism. This fails because it is not in a memory page. Then, the memory page allocating mechanism 18 uses the memory page selecting mechanism 16 to select a memory page whose contents are to be discarded. Since the current priority of each memory page is memory page group 5 and memory page 9 is 3, the memory page selecting mechanism 16 selects the memory page 9. The identifier in the page structure corresponding to the selected memory page 9 is replaced with that of the file fragment 4, and the priority is initialized to 0.
Then, the memory page allocating mechanism 18 returns the memory page 9 as invalid. If the contents are returned as invalid, then the I / O is actually performed to read the contents of the file fragment 4 into the memory page 9 and use it.
FIG. 6 shows a state in which I / O is completed.

In this mechanism, since each memory page is unified by a virtual identifier, the virtual file cache method does not consider what file system exists on the file system. Further, since this method is based on the existing memory management function in the computer system, a new storage area dedicated to the cache is not required when implementing this method.

Next, a second embodiment of the present invention will be described with reference to the drawings.

Referring to FIG. 7, according to the second embodiment of the present invention, priority attributes 19 and 20 indicating, for each file or file system, how much the cache priority is increased when referenced once. Is different from the first embodiment in that When increasing the priority in the page structure, the cache control mechanism 17 determines the increase amount by referring to the priority attribute.

The operation of the second embodiment of the present invention will be described below.

In FIG. 7, priority attribute 1 of file 1
9 is 1 and the priority attribute of file 2 is 3. The file fragment 3 and the file fragment 5 already exist in the memory pages 7 and 9, respectively, and are in a cached state. Here, the priorities held in the page structures 11 and 13 for managing the respective memory pages are 5 and 6.
Suppose That is, file fragment 3 has been referenced five times and file fragment 5 has been referenced twice. In terms of access frequency, file fragment 3 is more common, but file 2 has a higher priority attribute, and file fragment 5 that has been accessed only twice has higher priority. Therefore, if the memory page selection mechanism 16 operates in this state, the contents of the memory page 7 will be discarded.

In the second embodiment, the priority of the cache can be set for each file or file system. Therefore, especially when there is a file that requires high speed, the cache mechanism can be effectively operated.

In the embodiment of the present invention, the case where two files or file systems exist is described, but the number of files and file systems is not limited.

[0040]

The first effect is that a cache effect can be obtained irrespective of the file system on which the file is actually located.

The reason is that the cache control target is
This is because the memory page is an abstraction of the contents of a file fragment.

A second effect is that a dedicated storage area is not required for the file cache.

The reason is that the cache control mechanism is incorporated as a part of the memory management mechanism of the computer system.

[Brief description of the drawings]

FIG. 1 is a diagram showing the correspondence between a file and a cache control management structure according to the present invention.

FIG. 2 is a diagram showing information on identifiers and priorities inside a management structure shown in FIG. 1;

FIG. 3 is a diagram showing a relationship between a memory page selection mechanism and a management structure.

FIG. 4 is a diagram showing a relationship between a cache control mechanism and a management structure.

FIG. 5 is a diagram illustrating a memory page allocation mechanism having a cache control mechanism and a memory page selection mechanism.

FIG. 6 is a diagram showing a first embodiment of the present invention.

FIG. 7 is a diagram showing a second embodiment of the present invention.

[Explanation of symbols]

 1, 2, 3, 4, 5 file fragments 6 memory 7, 9 memory pages 10 array of page structures 11, 13 page structures 14 identifier 15 priority 16 memory page selection mechanism 17 cache control mechanism 18 memory page allocation mechanism 19 , 20 Priority attribute

Claims (4)

[Claims] When a file stored in a file system is divided into memory pages as file fragments, an identifier is assigned to each of the file fragments, and a cache operation is performed from a computer system connected to the file system. A virtual file cache control method, wherein the file fragment is cached using the identifier, so that the cache can be performed regardless of the type of the file. 2. The virtual file cache control method according to claim 1, wherein a part of a storage area of a memory of said computer system is used as a cache memory area. 3. When the file fragment is stored in the memory page corresponding to a memory page in a storage area of the memory, the file fragment changes according to the identifier assigned to the file fragment and the number of accesses to the file fragment. 3. The virtual file cache control system according to claim 1, further comprising a storage unit that stores the priority of the virtual file cache. 4. The virtual file cache control method according to claim 3, wherein file fragments are discarded based on the priority.