Distributed file system permitting each user to enhance cache hit ratio in file access mode

• US 5,559,984 A
• Filed: 09/28/1994
• Issued: 09/24/1996
• Est. Priority Date: 09/28/1993
• Status: Expired due to Term

• Current Assignee: Hitachi, Ltd.
• Original Assignee: Hitachi, Ltd.
• Inventors: Kuroda, Takaki, Domyo, Seiichi, Nakano, Hirohiko, Kobayashi, Atsushi, Shouji, Naofumi
• Primary Examiner(s): Kulik, Paul V.
• Application Number:
  US08/313,844
• Time in Patent Office: 727 Days
• Field of Search: 395/200.01, 395/200.12, 395/448, 395/486, 395/600
• US Class Current: 711/121
• CPC Class Codes: G06F 12/0866 for peripheral storage systems, e.g. disk cache
• CPC Class Codes: G06F 12/122 of the least frequently used [LFU] type, e.g. with individual count value

Abstract

A distributed file system in which the cache hit ratio of a client is enhanced to speed up a file access for each of users logging into the client. A file server includes an access frequency database in which the names of users are listed in association with the names of files that are frequently accessed by the individual users. Each client includes a log-in user table for entering the name of a user who is logging in, and a cache priority control module. The cache priority control module sets priority levels for the copies of the files stored in a cache area, on the basis of the contents of the access frequency database and the log-in user table. The set priority levels function as criteria when any of the file copies is to be expelled from the cache area. Owing to this construction, the copies of the files of high usage frequencies are preferentially kept in the cache area of the client for each user logging into this client, whereby the cache hit ratio can be enhanced to speed up the file access.

Claims

1. In a distributed file system wherein a file server for administering a plurality of files, and clients for accessing the files administered by the file server are connected through a network;

and each of the clients has a cache memory for administering copies of the files, and a priority table for storing therein priority levels which function as criteria for judging, when a copy of one of the files is to be expelled, if the copies of the files placed in the cache memory of the pertinent client through accesses from the pertinent client to the file server are to be kept in the cache memory;

a method of controlling a cache memory for files, comprising the steps of;

prestoring names of users and names of those ones of said files which are frequently accessed by the users, as an access frequency database in said file server, the user names and the file names being associated for the respective users who log into any of said clients, in the access frequency database;

recognizing the user name of a certain one of said users, in one client when the certain user is logging into said one client;

checking if a particular one of said files having a certain one of said file names is the file which is frequently accessed by the recognized user, in said one client and on the basis of the contents of said access frequency database when said recognized user has loaded the copy of the particular file from said file server into said cache memory;

setting the priority level of said particular file in accordance with the checked result so that the priority levels of the files of high access frequencies may become high, and then storing the set priority level in said priority table; and

determining the file whose priority level is low for the user currently operating said one client, as the file to-be-expelled on the basis of the contents of said priority table when said copy of one of said files is to be expelled from said cache memory of said one client.

2. A method of controlling a cache memory for files as defined in claim 1, wherein in said priority table, the file whose access frequency has been checked to be high for the logged-in user on the basis of said contents of said access frequency database is concatenated to a high-priority list, and any other file is concatenated to a low-priority list.

3. A method of controlling a cache memory for files as defined in claim 2, wherein when said copy of said particular file has been loaded into said cache memory and when it has been accessed, times on these occasions are respectively recorded in association with said particular file in said priority table, and said priority levels of said files are updated on the basis of the recorded times so as to become higher in correspondence with newer ones of said recorded times, in each of said high-priority list and said low-priority list.

4. A method of controlling a cache memory for files as defined in claim 1, wherein when said user of said one client logs in and logs out, said priority levels of the respective files in said priority table are updated for the copies of said files already existent in said cache memory and on the basis of the access frequencies of said user for the respective copies.

5. A method of controlling a cache memory for files as defined in claim 2, wherein at the times when said user of said one client logs in and logs out, said priority levels of the respective files in said priority table are updated for the copies of said files already existent in said cache memory and on the basis of the access frequencies of said user for the respective copies.

6. A method of controlling a cache memory for files as defined in claim 1, wherein when said cache memory stores therein the copies of fixed-length block units obtained by dividing each of said files into a plurality of fixed-length blocks, said priority table stores the priority levels in said fixed-length block units therein.

7. A method of controlling a cache memory for files as defined in claim 1, wherein when the copy of a certain one of said files is not existent in said cache memory at an access to the certain file in said one client, said copy of said certain file is loaded from said file server into said cache memory.

8. A method of controlling a cache memory for files as defined in claim 1, wherein when said certain user has logging into said one client, the files whose access frequencies are high for said certain user are found on the basis of said access frequency database, and the copies of said files of the high access frequencies are loaded from said file server into said cache memory without regard to the presence of accesses to said files.

9. In a distributed file system wherein a file server for administering a plurality of files, and clients for accessing the files administered by the file server are connected through a network;

and each of the clients has a cache memory for administering copies of the files;

a method of controlling a cache memory for files, comprising the steps of;

prestoring names of users and names of those ones of said files which are frequently accessed by the users, as an access frequency database in said file server, the user names and the file names being associated for the respective users who log into any of said clients, in the access frequency database;

recognizing the user name of a certain one of said users, in one client when the certain user is logging into said one client;

checking if a particular one of said files having a certain one of said file names is the file which is frequently accessed by the recognized user, in said one client and on the basis of the contents of said access frequency database; and

thenceforth loading said copy of said particular file whose access frequency is high in accordance with the checked result, from said file server into said cache memory without regard to the presence of the access to said particular file when said user has logged into said one client.

10. In a distributed file system wherein at least one file server for administering a plurality of files, and clients for accessing the files administered by the file server are connected through a network;

a distributed file system wherein;

said file server includes an access frequency database in which names of users to log into any ones of said clients and names of the files to be frequently accessed by the respective users are prestored in association; and

each of said clients includes;

a log-in user table for entering therein the name of the user who is logging in;

a cache memory for administering copies of the files stored in said file server; and

cache management means for setting priority levels for the respective copies of said files stored in said cache memory, and for performing a control in which the copy of low priority level is expelled from said cache memory when a sufficient free space cannot be reserved in said cache memory in the case of loading a copy of another of said files of said file server into said cache memory;

said cache management means setting said priority levels for said respective copies of said files stored in said cache memory, on the basis of said log-in user table and said access frequency database.

11. A distributed file system as defined in claim 10, wherein said cache management means includes a local access frequency database which holds therein a copy of a content part of the first-mentioned access frequency database corresponding to the user who is logged in, and said cache management means determines said priority levels on the basis of said local access frequency database and said log-in user table.

12. A distributed file system as defined in claim 10, wherein said access frequency database stores therein an access frequency file in which the names of said files to be frequently accessed by said users are entered for said respective users.

13. A distributed file system as defined in claim 11, wherein said access frequency database stores therein an access frequency file in which the names of said files to be frequently accessed by said users are entered for said respective users.

14. A distributed file system as defined in claim 10, comprising a plurality of such file servers, each of which includes a part of said access frequency database which stores the names of said users to frequently access the corresponding files administered by the pertinent file server.

15. A distributed file system as defined in claim 11, comprising a plurality of such file servers, each of which includes a part of said access frequency database which stores the names of said users to frequently access the corresponding files administered by the pertinent file server.

16. A distributed file system as defined in claim 10, wherein said cache memory stores therein the copies of fixed-length block units obtained by dividing each of said files into a plurality of fixed-length blocks, and said cache management means includes a priority table in which the priority levels are stored in said fixed-length block units.

17. A distributed file system as defined in claim 10, wherein said cache management means includes a priority table which contains a high-priority list for concatenating thereto the file whose access frequency is set high for the logged-in user on the basis of the contents of said access frequency database, and a low-priority list for concatenating any other file thereto.

18. A distributed file system as defined in claim 17, wherein said cache management means includes means for recording respective times, when the copy of a particular one of said files has been loaded into said cache memory and when said copy has been accessed, in association with the particular file in said priority table, and for updating said priority levels of said files on the basis of the recorded times so as to become higher in correspondence with newer ones of said recorded times, in each of said high-priority list and said low-priority list.

19. A distributed file system as defined in claim 10, wherein said cache management means includes a priority table for storing said priority levels of said files therein, and means for updating said priority levels of the respective files in said priority table for the copies of said files already existent in said cache memory and on the basis of the access frequencies of said user for the respective copies, when said user logs in and logs out.