Method and apparatus for asynchronous file writes in a distributed file system

• US 7,130,855 B2
• Filed: 10/09/2003
• Issued: 10/31/2006
• Est. Priority Date: 07/31/2000
• Status: Expired due to Fees

• Current Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.)
• Original Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.)
• Inventors: Lebel, Diane
• Primary Examiner(s): Gaffin, Jeffrey
• Assistant Examiner(s): ABEL JALIL, NEVEEN
• Application Number:
  US10/682,041
• Publication Number:
  US 20040078396A1
• Time in Patent Office: 1,118 Days
• Field of Search: 711/133, 711/136, 709/203
• US Class Current: 707/613
• CPC Class Codes: G06F 16/178 Techniques for file synchronisation in file systems
• CPC Class Codes: G06F 16/182 Distributed file systems
• CPC Class Codes: G06F 2209/5014 Reservation
• CPC Class Codes: G06F 2209/522 Manager
• CPC Class Codes: G06F 2209/523 Mode
• CPC Class Codes: G06F 9/5016 the resource being the memory
• CPC Class Codes: G06F 9/5022 Mechanisms to release resources
• CPC Class Codes: G06F 9/526 Mutual exclusion algorithms
• CPC Class Codes: Y10S 707/959 Network
• CPC Class Codes: Y10S 707/99952 Coherency, e.g. same view to multiple users
• CPC Class Codes: Y10S 707/99953 Recoverability
• CPC Class Codes: Y10S 707/99957 Garbage collection

Abstract

A method and apparatus for delaying asynchronous writes in a distributed file system, wherein the file system includes a unique identifier (“UID”). The method comprises buffering a page of dirty data with the unique identifier upon writing to the server; changing the unique identifier (“UID”) to create a current unique identifier upon a failure of the server; comparing the buffered unique identifier with the current unique identifier when the page is requested while the page is in a written state; and handling the request responsive to the comparison. In other aspects of the invention, the invention comprises a computer programmed to perform the method and a program storage medium encoded with instructions that, when executed by a computer, perform the method.

Claims

1. A method for delaying asynchronous writes in a distributed file system of a computing system to which a unique identifier is assigned, comprising:

said computing system buffering a page of dirty data with the unique identifier upon writing to a server;

said computing system changing the unique identifier to create a current unique identifier that is assigned to the distributed file system upon a failure of the server;

said computing system comparing the buffered unique identifier with the current unique identifier when the page is requested while the page is in a written state;

said computing system handling the request responsive to the comparison by storing the buffered written page to disk storage if the buffered unique identifier differs from the current unique identifier or performing a file sync operation if the buffered unique identifier matches the current unique identifier; and

said computing system maintaining cache consistency by issuing an exclusive mode token from the server to a client to permit the client to dirty the page, issuing a shared mode token from the server to the client to permit the client to use, but not dirty, the page, and revoking the exclusive mode token before issuing the shared mode token.

2. The method of claim 1, further comprising assigning the unique identifier.

3. The method of claim 2, wherein assigning the unique identifier comprises assigning a sequence number.

4. The method of claim 3, wherein assigning the sequence number comprises assigning a cluster-wide sequence number.

5. The method of claim 3, wherein changing the unique identifier comprises incrementing the sequence number.

6. The method of claim 1, wherein buffering the dirty page comprises storing the unique identifier in a header of the page.

7. The method of claim 6, wherein buffering the written page with the unique identifier comprises returning the page to a least recently used queue.

8. The method of claim 1, wherein buffering the written page with the unique identifier comprises returning the page to a least recently used queue.

9. The method of claim 1, wherein the dirty page transitions to a written state upon the write to the server.

10. The method of claim 9, wherein the written page is handled as though in the dirty state when the page is in the written state.

11. The method of claim 1, wherein buffering the written page with the unique identifier comprises caching the written page with the unique identifier.

12. The method of claim 11, wherein caching the written page with the unique identifier comprises caching the written page with the unique identifier on the client.

13. The method of claim 1, wherein changing the unique identifier comprises incrementing the unique identifier.

14. The method of claim 1, wherein comparing the buffered unique identifier with the current unique identifier comprises determining whether the current unique identifier is a numerically higher value.

15. The method of claim 1, further comprising:

buffering a plurality of dirty pages before writing them to the server; and

writing the buffered dirty pages to the server in a single write operation.

16. The method of claim 1, further comprising:

buffering a plurality of written pages; and

writing the plurality of written pages to disk storage in a single write operation.

17. The method of claim 1, wherein at least one of the exclusive mode token and the shared mode token is embedded in at least one of a read and a write operations in an operating system.

18. The method of claim 1, wherein revoking the exclusive mode token comprises revoking the exclusive mode token when another computing system wants to read the data.

19. The method of claim 1, further comprising a server sharing free space information with a plurality of clients.

20. The method of claim 1, further comprising ensuring that a file modification time of a file is updated before or at file close time but not after, comprising:

setting a flag at the client when any page of the file is dirtied;

clearing the flag when dirty data is sent to the server before the file is closed;

notifying the server from the client to update the modification time if the flag is still set at close time;

notifying the server from the client to forego updating the modification time if dirty data is sent to the server after the file has been closed.

21. A program storage medium encoded with instructions that, when executed by a computer, perform a method for delaying asynchronous writes in a distributed file system to which a unique identifier is assigned, the method comprising:

buffering a page of dirty data with the unique identifier upon writing to a server;

changing the unique identifier upon a failure of the server to create a current unique identifier;

comparing the buffered unique identifier with the current unique identifier when the page is requested while the page is in a written state; and

handling the request responsive to the comparison by storing the buffered written page to disk storage if the buffered unique identifier differs from the current unique identifier or performing a file sync operation if the buffered unique identifier matches the current unique identifier; and

maintaining cache consistency by issuing an exclusive mode token from the server to a client to permit the client to dirty the page, issuing a shared mode token from the server to the client to permit the client to use, but not dirty, the page, and revoking the exclusive mode token before issuing the shared mode token.

22. The program storage medium of claim 21, wherein the method further comprises assigning the unique identifier.

23. The program storage medium of claim 22, wherein assigning the unique identifier in the method comprises assigning a sequence number.

24. The program storage medium of claim 23, wherein assigning the sequence number in the method comprises assigning a cluster-wide sequence number.

25. The program storage medium of claim 23, wherein changing the unique identifier in the method comprises incrementing the sequence number.

26. The program storage medium of claim 21, wherein buffering the dirty page in the method comprises storing the unique identifier in a header of the page.

27. The program storage medium of claim 26, wherein buffering the written page with the unique identifier in the method comprises returning the page to a least recently used queue.

28. The program storage medium of claim 21, wherein buffering the written page with the unique identifier in the method comprises returning the page to a least recently used queue.

29. The program storage medium of claim 21, wherein the dirty page transitions to a written state upon the write to the server in the method.

30. The program storage medium of claim 29, wherein the written page is handled by the method as though in the dirty state when the page is in the written state.

31. The program storage medium of claim 21, wherein buffering the written page with the unique identifier in the method comprises caching the written page with the unique identifier.

32. The program storage medium of claim 31, wherein caching the written page with the unique identifier in the method comprises caching the written page with the unique identifier on the client.

33. The program storage medium of claim 21, wherein changing the unique identifier in the method comprises incrementing the unique identifier.

34. The program storage medium of claim 21, wherein comparing the buffered unique identifier with the current unique identifier in the method comprises determining whether the current unique identifier is a numerically higher value.

35. The program storage medium of claim 21, wherein the method further comprises:

buffering a plurality of dirty pages before writing them to the server; and

writing the buffered dirty pages to the server in a single write operation.

36. The program storage medium of claim 21, wherein the method further comprises:

buffering a plurality of written pages; and

writing the plurality of written pages to disk storage in a single write operation.

37. The program storage medium of claim 21, wherein at least one of the exclusive mode token and the shared mode token is embedded in at least one of a read and a write operations in an operating system.

38. The program storage medium of claim 21, wherein revoking the exclusive mode token in the method comprises revoking the exclusive mode token when another computing system wants to read the data.

39. The program storage medium of claim 21, wherein the method further comprises a server sharing free space information with a plurality of clients.

40. The program storage medium of claim 21, wherein the method further comprises ensuring that a the file modification time a file is updated before or at file close time but not after, comprising:

setting a flag at the client when any page of the file is dirtied;

clearing the flag when dirty data is sent to the server before the file is closed;

notifying the server from the client to update the modification time if the flag is still set at close time; and

notifying the server from the client to forego updating the modification time if dirty data is sent to the server after the file has been closed.

41. A computer programmed to perform a method for delaying asynchronous writes in a distributed file system to which a unique identifier is assigned, the method comprising:

buffering a page of dirty data with the unique identifier upon writing the data to a server;

changing the unique identifier upon a failure of the server to create a current unique identifier;

comparing the buffered unique identifier with the current unique identifier when the page is requested while the page is in a written state; and

handling the request responsive to the comparison by storing the buffered written page to disk storage if the buffered unique identifier differs from the current unique identifier or performing a file sync operation if the buffered unique identifier matches the current unique identifier; and

maintaining cache consistency by issuing an exclusive mode token from the server to a client to permit the client to dirty the page, issuing a shared mode token from the server to the client to permit the client to use, but not dirty, the page, and revoking the exclusive mode token before issuing the shared mode token.

42. The programmed computer of claim 41, wherein the method further comprises assigning the unique identifier.

43. The programmed computer of claim 42, wherein assigning the unique identifier in the method comprises assigning a sequence number.

44. The programmed computer of claim 43, wherein assigning the sequence number in the method comprises assigning a cluster-wide sequence number.

45. The programmed computer of claim 43, wherein changing the unique identifier in the method comprises incrementing the sequence number.

46. The programmed computer of claim 41, wherein buffering the dirty page in the method comprises storing the unique identifier in a header of the page.

47. The programmed computer of claim 46, wherein buffering the written page with the unique identifier in the method comprises returning the page to a least recently used queue.

48. The programmed computer of claim 41, wherein buffering the written page with the unique identifier in the method comprises returning the page to a least recently used queue.

49. The programmed computer of claim 41, wherein the dirty page transitions to a written state upon the write to the server in the method.

50. The programmed computer of claim 49, wherein the written page is handled by the method as though in the dirty state when the page is in the written state.

51. The programmed computer of claim 41, wherein buffering the written page with the unique identifier in the method comprises caching the written page with the unique identifier.

52. The programmed computer of claim 51, wherein caching the written page with the unique identifier in the method comprises caching the written page with the unique identifier on the client.

53. The programmed computer of claim 41, wherein changing the unique identifier in the method comprises incrementing the unique identifier.

54. The programmed computer of claim 41, wherein comparing the buffered unique identifier with the current unique identifier in the method comprises determining whether the current unique identifier is a numerically higher value.

55. The programmed computer of claim 41, wherein the method further comprises:

buffering a plurality of dirty pages before writing them to the server; and

writing the buffered dirty pages to the server in a single write operation.

56. The programmed computer of claim 41, wherein the method further comprises:

buffering a plurality of written pages; and

writing the plurality of written pages to disk storage in a single write operation.

57. The programmed computer of claim 41, wherein at least one of the exclusive mode token and the shared mode token is embedded in at least one of a read and a write operations in an operating system.

58. The programmed computer of claim 41, wherein revoking the exclusive mode token in the method comprises revoking the exclusive mode token when another computing system wants to read the data.

59. The programmed computer of claim 41, wherein the method further comprises a server sharing free space information with a plurality of clients.

60. The programmed computer of claim 41, wherein the method further comprises ensuring that a file modification time a file is updated before or at file close time but not after, comprising:

setting a flag at the client when any page of the file is dirtied;

clearing the flag when dirty data is sent to the server before the file is closed;

notifying the server from the client to update the modification time if the flag is still set at close time; and

notifying the server from the client to forego updating the modification time if dirty data is sent to the server after the file has been closed.