Method and apparatus for establishing and maintaining the status of membership sets used in mirrored read and write input/output without logging

• US 5,917,998 A
• Filed: 07/26/1996
• Issued: 06/29/1999
• Est. Priority Date: 07/26/1996
• Status: Expired due to Term

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Long, Darrell Don Earl, Cabrera, Luis Felipe
• Primary Examiner(s): Beausoliel, Jr., Robert W.
• Assistant Examiner(s): Iqbal, Nadeem
• Application Number:
  US08/690,214
• Time in Patent Office: 1,068 Days
• Field of Search: 395/182.03, 395/182.04, 395/182.18, 395/185.01, 395/185.02, 395/183.01, 395/183.18, 395/183.2, 395/185.07, 711/114, 711/115, 711/118, 711/133, 711/112, 371/40.15
• US Class Current: 714/6.12
• CPC Class Codes: G06F 11/2058 using more than 2 mirrored copies
• CPC Class Codes: G06F 11/2064 while ensuring consistency
• CPC Class Codes: G06F 11/2082 Data synchronisation

Abstract

A method and apparatus that maintains in stable storage an identification of which copies in a mirrored disk system are currently up to date without requiring the use of a logging mechanism. The invention may include a plurality of mirrored disk drives and a processing unit for controlling the writing and reading of data units to and from mirrored disk drives, wherein the processing unit establishes and maintains cohort sets for each mirrored disk drive. The cohort set represents the status of copies of the data units stored on the mirrored disk drives and contains membership information for redundant copies of data units stored in the plurality of mirrored disk drives. The membership information comprises a data structure of 2n status bits for identifying which disks contain current copies of the data units, where n is the number of disks in the system, and wherein the first n bits are used to indicate an intention to update the membership information for the copies and the second n bits are used indicate a current state of the membership information. The data structure may further include log(n) additional bits for identifying each mirrored disk.

Claims

1. An apparatus for maintaining the status of redundant copies of data on mirrored disk drives in a mirrored storage system, each of the mirrored disk drives including a copy of data, comprising a current copy indicator associated with each of the redundant copies of data wherein the current copy indicator further comprises a first set of identifiers indicating when a copy on one of the mirrored disk drives is to be updated and a second set of identifiers indicating the state of each of the copies.

2. The apparatus of claim 1 wherein the first and second identifier are modified to reflect the current set of available copies when a copy becomes unavailable and a write operation is requested.

3. The apparatus of claim 1 wherein the current copy indicator is updated when unavailable copies are repaired.

4. The apparatus of claim 1 wherein out of date copies are updated using copies identified by the current copy indicator and the current copy indicator is modified to reflect a new status for the redundant copies.

5. A system of mirrored disk drives containing redundant copies of data units, the system comprising;

a plurality of mirrored disk drives; and

a processing unit, coupled to the mirrored disk drives, for controlling the writing and reading of data units to and from mirrored disk drives; and

wherein the processing unit establishes and maintains cohort sets for each mirrored disk drive, the cohort set representing status of copies of the data units stored on the mirrored disk drives, the cohort sets containing membership information for redundant copies of data units stored in the plurality of mirrored disk drives, the membership information comprising a data structure of 2n status bits for identifying which disks contain current copies of the data units, where n is the number of disks in the system, wherein the first n bits indicate an intention to update the membership information for the copies and the second n bits indicate a current state of the membership information.

6. The system of claim 5 wherein the processing unit comprises a memory.

7. The system of claim 5 wherein the cohort set for a drive represents the set of copies that participated in the last write.

8. The system of claim 5 wherein the data structure further comprises log(n) additional bits for identifying each mirrored disk.

9. The system of claim 5 wherein the cohort set represents the known current disks, the cohort set being modified every time a failure is detected and every time a repair is completed.

10. The system of claim 5 wherein each mirrored disk is associated with a bit of the data structure.

11. The system of claim 10 wherein a bit is on when the disk is available and off when the disk is not available.

12. The system of claim 5 wherein the data units comprises a unit of storage.

13. The system of claim 5 wherein the cohort set is established and administered by the system.

14. The system of claim 5 wherein the 2n status bits are stored in the space allocation map for each physical volume.

15. The system of claim 5 wherein the cohort set enables the system to recognize whether a complete set of copies is on line.

16. The system of claim 5 wherein the bits in the data structure are modified in all the disks that remain operational when a failure is detected by setting the bit off for the disk that is down.

17. The system of claim 16 wherein the modified data structure is copied to all of the disks that remain operational.

18. The system of claim 5 wherein updating of the membership information comprises writing an updated cohort set to all the disks that remain operational.

19. The system of claim 5 wherein cascading failures are treated sequentially such that a new round of membership information is initiated if an additional disk becomes unavailable while the membership information is being updated.

20. The system of claim 5 wherein the cohort set indicates that a copy is current if the cohort sets are equal and complete.

21. The system of claim 20 wherein cohort sets are equal if the cohort sets have the same membership and the cohort sets are complete if every copy in a subset comprising the union of the cohort sets is present.

22. The system of claim 5 wherein two phases are used to update the membership information, the first n bits being written to identify an intention to update the membership information for the copies in the first phase, and the second n bits being updated to indicate the current state of the membership information and the first n bits being cleared in the second phase.

23. The system of claim 5 wherein the cohort sets are modified when a write operation occurs following a failure.

24. A method of maintaining the status of mirrored disk drives containing redundant copies of data, the method comprising the steps of:

establishing a cohort set on each of the mirrored disk drive representing a status of the copies stored on the mirrored disk drives, the cohort set containing membership information for redundant copies of data units stored in the plurality of mirrored disk drives, the membership information comprising a data structure of 2n status bits for identifying which disks contain current copies of the data units, where n is the number of disks in the system, wherein the first n bits indicate an intention to update the membership information for the copies and the second n bits indicate a current state of the membership information; and

updating the contents of the cohort sets when a disk drive failure has been detected to reflect unavailability of the copy of data on the failed disk drive.

25. The method of claim 24 further comprising the steps of:

repairing the copy of data on the failed disk drive when the failed disk drive is brought up; and

revising the cohort sets when the repair is complete to reflect availability of the repaired disk drive.

26. The method of claim 24 further comprising the steps of:

writing the first n bits to identify an intention to update the membership information for the copies in a first phase; and

updating the second n bits to indicate the current state of the membership information and clearing the first n bits in a second phase.

27. The method of claim 24 wherein the step of updating the cohort set is performed every time a failure is detected and every time a repair is completed.

28. The method of claim 24 wherein the step of updating the membership information further comprises the step of modifying the membership information of all disks that remain operational when a failure is detected.

29. The method of claim 28 wherein the step of modifying the membership information of all disks that remain operational further comprises the step setting a bit off for a disk that is not operational.

30. The method of claim 29 wherein the modified membership information is copied to all of the disks that remain operational.

31. A metadata file structure stored on each of a plurality of mirrored disk drives containing redundant copies of data units, the metadata file structure comprising a first and second set of n identifiers, wherein n is the number of mirrored disk drives and each identifier represents a mirrored disk drive, the first set of n identifiers indicating disk drives to be updated and the second set of identifiers indicating what disk drives contain current copies.

32. The metadata file structure of claim 31 wherein an identifier is set to an on indication when the disk is available and to an off indication when the disk is not available.

33. The metadata file structure of claim 31 wherein the first and second set of n identifiers represents the set of copies that participated in the last write involving the copy.

34. The metadata file structure of claim 31 wherein the first and second set of n identifiers represent membership information for redundant copies of data units stored in the plurality of mirrored disk drives.

35. The metadata file structure of claim 31 further comprising log(n) additional identifiers for identifying each mirrored disk.

36. The metadata file structure of claim 31 wherein the first and second set of n identifiers represent the known current disks, the first and second set of n identifiers being modified every time a failure is detected and every time a repair is completed.

37. The metadata file structure of claim 31 wherein each mirrored disk is associated with an identifier in each of the first and second set of n identifiers.

38. The metadata file structure of claim 37 wherein an identifier is set to an on indication when the disk is available and to an off indication when the disk is not available.

39. The metadata file structure of claim 31 wherein the data units comprises a unit of storage.

40. The metadata file structure of claim 31 wherein the first and second set of n identifiers are established and administered by the system.

41. The metadata file structure of claim 31 wherein the first and second set of n identifiers are stored in the space allocation map for each physical volume.

42. The metadata file structure of claim 31 wherein the first and second set of n identifiers enable the system to recognize whether a complete set of copies is on line.

43. The metadata file structure of claim 31 wherein the first and second set of n identifiers are modified in all the disks that remain operational when a failure is detected by setting the bit off for the disk that is down.

44. The metadata file structure of claim 43 wherein the modified first and second set of n identifiers are copied to all of the disks that remain operational.

45. The metadata file structure of claim 31 wherein cascading failures are treated sequentially such that the updating of the first and second set of n identifiers is reinitiated if an additional disk becomes unavailable while the first and second set of n identifiers are being updated.

46. The metadata file structure of claim 31 wherein the first and second set of n identifiers indicate that a copy is current if each first and second set of n identifiers stored on the disks are equal and complete.

47. The metadata file structure of claim 46 wherein each first and second set of n identifiers is equal if each of the first and second set of n identifiers have the same information and each of the first and second set of n identifiers are complete if every copy is represented in each of the first and second set of n identifiers.

48. The metadata file structure of claim 31 wherein the first and second set of n identifiers are modified when a write operation occurs following a failure.

49. A program storage device, readable by a computer, tangibly embodying one or more programs of instructions executable by the computer to perform method steps for managing mirrored disk drives containing redundant copies of data, the method comprising the steps of:

establishing a cohort set on each of the mirrored disk drives representing a status of the copies stored on the mirrored disk drives, the cohort set containing membership information for redundant copies of data units stored in the plurality of mirrored disk drives, the membership information comprising a data structure of 2n status bits for identifying which disks contain current copies of the data units, where n is the number of disks in the system, wherein the first n bits indicate an intention to update the membership information for the copies and the second n bits indicate a current state of the membership information; and

updating the contents of the cohort sets when a disk drive failure has been detected to reflect unavailability of the copy of data on the failed disk drive.

50. The program storage device of claim 49 further comprising the steps of:

repairing the copy of data on the failed disk drive when the failed disk drive is brought up; and

revising the cohort sets when the repair is complete to reflect availability of the repaired disk drive.

51. The program storage device of claim 49 further comprising the steps of:

writing the first n bits to identify an intention to update the membership information for the copies in a first phase; and

updating the second n bits to indicate the current state of the membership information and clearing the first n bits in a second phase.

52. The program storage device of claim 49 wherein the step of updating the cohort set is performed every time a failure is detected and every time a repair is completed.

53. The program storage device of claim 49 wherein the step of updating the membership information further comprises the step of modifying the membership information of all disks that remain operational when a failure is detected.

54. The program storage device of claim 53 wherein the step of modifying the membership information of all disks that remain operational further comprises the step setting a bit off for a disk that is not operational.

55. The program storage device of claim 53 wherein the modified membership information is copied to all of the disks that remain operational.

56. A memory for storing data for managing mirrored disk drives containing redundant copies of data, comprising:

a data structure stored in a memory of each of the mirrored disk drives, each of the data structure including a first and second set of n identifiers, wherein n is the number of mirrored disk drives and each identifier represents a mirrored disk drive, the first set of n identifiers indicating disk drives to be updated and the second set of identifiers indicating what disk drives contain current copies.

57. The memory of claim 56 wherein an identifier is set to an on indication when the disk is available and to an off indication when the disk is not available.

58. The memory of claim 56 wherein the first and second set of n identifiers represents the set of copies that participated in the last write involving the copy.

59. The memory of claim 56 wherein the first and second set of n identifiers represent membership information for redundant copies of data units stored in the plurality of mirrored disk drives.

60. The memory of claim 56 further comprising log(n) additional identifiers for identifying each mirrored disk.

61. The memory of claim 56 wherein the first and second set of n identifiers represent the known current disks, the first and second set of n identifiers being modified every time a failure is detected and every time a repair is completed.

62. The memory of claim 56 wherein each mirrored disk is associated with an identifier in each of the first and second set of n identifiers.

63. The memory of claim 62 wherein an identifier is set to an on indication when the disk is available and to an off indication when the disk is not available.

64. The memory of claim 56 wherein the data units comprises a unit of storage.

65. The memory of claim 56 wherein the first and second set of n identifiers are established and administered by the system.

66. The memory of claim 56 wherein the first and second set of n identifiers are stored in the space allocation map for each physical volume.

67. The memory of claim 56 wherein the first and second set of n identifiers enable the system to recognize whether a complete set of copies is on line.

68. The memory of claim 56 wherein the first and second set of n identifiers are modified in all the disks that remain operational when a failure is detected by setting the bit off for the disk that is down.

69. The memory of claim 68 wherein the modified first and ,second set of n identifiers are copied to all of the disks that remain operational.

70. The memory of claim 56 wherein cascading failures are treated sequentially such that the updating of the first and second set of n identifiers is reinitiated if an additional disk becomes unavailable while the first and second set of n identifiers are being updated.

71. The memory of claim 56 wherein the first and second set of n identifiers indicate that a copy is current if each first and second set of n identifiers stored on the disks are equal and complete.

72. The memory of claim 71 wherein each first and second set of n identifiers is equal if each of the first and second set of n identifiers have the same information and each of the first and second set of n identifiers are complete if every copy is represented in each of the first and second set of n identifiers.

73. The memory of claim 66 wherein the first and second set of n identifiers are modified when a write operation occurs following a failure.