DETECTING WHEN TO PREFETCH INODES AND THEN PREFETCHING INODES IN PARALLEL

• US 20080126299A1
• Filed: 01/14/2008
• Published: 05/29/2008
• Est. Priority Date: 07/18/2000
• Status: Active Grant

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Schmuck, Frank B., Wyllie, James C.
• Granted Patent: US 7,707,360 B2
• Time in Patent Office: Days
• US Class Current: 707/2
• CPC Class Codes: G06F 12/0862 with prefetch
• CPC Class Codes: G06F 16/10 File systems; File servers
• CPC Class Codes: G06F 2212/6026 Prefetching based on access pattern detection, e.g. stride based prefetch

Abstract

The decision to prefetch inodes is based upon the detecting of access patterns that would benefit from such a prefetch. Once the decision to prefetch is made, a plurality of inodes are prefetched in parallel. Further, the prefetching of inodes is paced, such that the prefetching substantially matches the speed at which an application requests inodes.

Claims

1. A method of managing the prefetching of data of files, said method comprising:

detecting a pattern of requests for data of multiple files, wherein the pattern is based on one or more user-defined attributes of the multiple files;

automatically initiating prefetch of data of a plurality of files, in response to the detecting; and

prefetching data of the plurality of files, wherein the detecting comprises automatically repeatedly determining whether the pattern of requests exists, and wherein the automatically initiating and the prefetching do not occur when it is determined that the pattern of requests does not exist, and wherein the automatically repeatedly determining occurs each time a directory having said multiple files and said plurality of files is accessed.

2. The method of claim 1, wherein said data comprises meta data.

3. The method of claim 1, wherein said multiple files and said plurality of files are within a single directory.

4. The method of claim 1, wherein said detecting comprises determining whether a cache miss threshold has been exceeded, wherein said detecting indicates said pattern when said cache miss threshold has been exceeded.

5. The method of claim 4, wherein said determining comprises comparing a counter of cache misses that occurred within a preselected time interval to said cache miss threshold to determine whether said cache miss threshold has been exceeded.

6. The method of claim 5, wherein said counter and said cache miss threshold are associated with a directory block of a directory of files, said directory of files comprising said multiple files and said plurality of files, and said directory of files comprising one or more directory blocks.

7. The method of claim 1, wherein said prefetching comprises prefetching data of at least some files of said plurality of files in parallel.

8. The method of claim 1, wherein said prefetching comprises prefetching data at an average rate that substantially matches a speed of requests for data.

9. The method of claim 1, wherein said prefetching comprises:

obtaining data associated with a number of files of said plurality of files;

determining whether a cache hit threshold has been reached; and

obtaining data associated with one or more additional files of said plurality of files, in response to reaching said cache hit threshold.

10. A method of managing the prefetching of data of files, said method comprising:

detecting a pattern of requests for data of multiple files, wherein the pattern is based on one or more user-defined attributes of the multiple files;

automatically initiating prefetch of data of a plurality of files, in response to the detecting, wherein the automatically initiating is absent a disruption of access to data by applications; and

prefetching data of the plurality of files.

11. The method of claim 10, wherein said multiple files and said plurality of files are within a single directory.

12. The method of claim 10, wherein said detecting comprises determining whether a cache miss threshold has been exceeded, wherein said detecting indicates said pattern when said cache miss threshold has been exceeded.

13. The method of claim 10, wherein said prefetching comprises prefetching data of at least some files of said plurality of files in parallel.

14. The method of claim 10, wherein said prefetching comprises prefetching data at an average rate that substantially matches a speed of requests for data.

15. The method of claim 10, wherein said prefetching comprises:

obtaining data associated with a number of files of said plurality of files;

determining whether a cache hit threshold has been reached; and

obtaining data associated with one or more additional files of said plurality of files, in response to reaching said cache hit threshold.

16. The method of claim 10, wherein the detecting comprises automatically repeatedly determining whether the pattern of requests exists, and wherein the automatically initiating and the prefetching do not occur when it is determined that the pattern of requests does not exist.

17. A system of managing the prefetching of data of files, said system comprising:

a first node adapted to detect a pattern of requests for data of multiple files, wherein the pattern is based on one or more user-defined attributes of the multiple files; and

at least one second node adapted to automatically initiate prefetch of data of a plurality of files, in response to the detecting, wherein the automatically initiating is absent a disruption of access to data by applications, and to prefetch data of the plurality of files.

18. A system of managing the prefetching of inodes associated with files of a directory, said directory comprising one or more directory blocks and each directory block having associated therewith zero or more files, said system comprising:

a first node adapted to detect a pattern of requests for multiple inodes associated with multiple files of a directory block of said one or more directory blocks, wherein the pattern is based on directory entries of the multiple files being within said directory block; and

at least one second node adapted to automatically initiate prefetch of a plurality of inodes associated with said directory block, in response to detecting said pattern, wherein the automatically initiating is absent a disruption of access to directory blocks by applications, and to prefetch the plurality of inodes associated with said directory block.

19. At least one program storage device readable by a machine, tangibly embodying at least one program of instructions executable by the machine to perform a method of managing the prefetching of data of files, said method comprising:

detecting a pattern of requests for data of multiple files, wherein the pattern is based on one or more user-defined attributes of the multiple files;

automatically initiating prefetch of data of a plurality of files, in response to the detecting, wherein the automatically initiating is absent a disruption of access to data by applications; and

prefetching data of the plurality of files.

20. The at least one program storage device of claim 19, wherein said prefetching comprises:

obtaining data associated with a number of files of said plurality of files, wherein said obtaining data associated with said number of files comprises issuing a plurality of I/O requests to read data of at least a portion of said number of files in parallel;

determining whether a cache hit threshold has been reached; and

obtaining data associated with one or more additional files of said plurality of files, in response to reaching said cache hit threshold.