System for backing up computer disk volumes with error remapping of flawed memory addresses

• US 5,907,672 A
• Filed: 10/04/1995
• Issued: 05/25/1999
• Est. Priority Date: 10/04/1995
• Status: Expired due to Term

• Current Assignee: Veritas Technologies, LLC (Whitehouse Group Ltd.)
• Original Assignee: Stac, Inc. (Titan Corporation)
• Inventors: Matze, John E. G., Whiting, Douglas L.
• Primary Examiner(s): Swann, Tod R.
• Assistant Examiner(s): Peikari, James
• Application Number:
  US08/539,315
• Time in Patent Office: 1,329 Days
• Field of Search: 369/53, 369/54, 365/15, 395/440, 395/622, 395/182.04, 395/489, 395/182.06
• US Class Current: 714/6.13
• CPC Class Codes: G06F 11/1451 by selection of backup contents

Abstract

A system for backing up data from a computer disk volume at very high speed by saving a logical image copy of the volume to a backup medium such as magnetic tape. This logical image copy can later be restored in its entirety to a disk volume with a different physical geometry and flaw map in a disaster recovery mode, significantly reducing the time required for such disaster recovery compared to other backup techniques. In addition, the logical image copy on the backup medium also allows selective file restore with performance comparable to that achievable using traditional file-by-file backup/restore methods. The backup process can thus run considerably faster than conventional approaches without sacrificing the restore flexibility normally associated with those approaches.

Claims

1. A method for backing up data in a computer system from a primary storage means to a backup storage means on a sector-by-sector basis and restoring data in a computer system from said backup storage means to a restore storage means on a sector-by-sector basis, said method comprising the steps of:

reading a set of logically contiguous sectors from the primary storage means using a software call of the operating system that provides access to the files stored on said primary storage means, said call of said operating system performing any physical level remapping necessary to avoid previously detected physical flaws on said primary storage means,

writing said set of logically contiguous sectors to said backup storage means,

creating a partition on said restore storage means of a size at least as large as the size of said primary storage means,

reading a set of logically contiguous sectors from a location on said backup storage means,

writing said set of logically contiguous sectors to said partition of said restore storage means using a software call to the operating system that provides access to the files stored on said partition of said restore storage means, said call of said operating system performing any physical level remapping necessary to detect and avoid physical flaws on said restore storage means.

2. The method of claim 1, further including the steps of

writing on said backup storage means a sector directory table containing information sufficient to indicate the size of said primary storage means and the location of each logical sector written to said backup storage means,

reading said sector directory table from said backup storage means,

using said sector directory table to determine the sector numbers and locations of said logically contiguous blocks to be read.

3. The method of claim 2 wherein said primary storage means consists one or more disk drive partition(s) in said computer system, and wherein said operating system call to read said logically contiguous sectors performs the mapping necessary to locate said logically contiguous sectors on said disk drive partition(s).

4. The method of claim 2 wherein said partition created on said restore storage means is larger than the size of said original primary storage means.

5. The method of claim 2 wherein said partition created on said restore storage means spans multiple physical disk drives.

6. The method of claim 2 wherein unused sectors that do not contain file data are not read from said primary storage means and are not stored on said backup storage means, and wherein the absence of said unused sectors on the backup storage means is indicated in said sector directory table.

7. The method of any of claims 1-6 wherein deleted sectors that contain data from deleted files are not read from said primary storage means and are not stored on said backup storage means, and wherein the absence of said deleted sectors on the backup storage means is indicated in said sector directory table.

8. The method of any of claims 1-6 wherein an open file log is maintained of all files which are opened for write while the backup method is in process.

9. The method of claim 8 wherein said open file log is written to said backup storage means.

10. The method of any of claims 1-6 wherein said operating system allows multi-tasking, further including the step of:

temporarily suspending execution of any tasks that attempt to write a set of sectors to said primary storage means until said set of sectors has been read from said primary storage means by the backup task in preparation for writing said sectors to said backup storage means.

11. The method of claim 10, further including the steps of:

maintaining a cache of sectors read from said primary storage means to be written to said backup storage means,

detecting an attempted write by a task to a set of sectors of said primary storage means which has not yet been backed up,

operative when said sector cache is full, temporarily suspending executing of said task,

operative when said sector cache is not full, reading said set of sectors and adding said set of sectors to said sector cache and then allowing said task to continue execution without suspension,

checking said sector cache for the presence of any portion of said set of logically contiguous sectors to be read from said primary storage means,

operative when no such portion is found in said sector cache, reading said set of logically contiguous sectors from said primary storage means,

operative when such portion is found in said sector cache, reading said portion(s) of said set of logically contiguous sectors from said sector cache, and reading remaining portions not found in said sector cache from said primary storage means,

whereby no said tasks attempting to write to said primary storage means will be suspended unless said sector cache is full.

12. The method of claim 11 wherein said portions of said set of logically contiguous set of sectors found in said sector cache are removed from said sector cache after said portions are read from said sector cache, whereby portions of said sector cache may be re-used in order to minimize the number of times that tasks are suspended.

13. The method of any of claims 1-6 wherein only sectors that have changed since the last backup are written to the backup storage means.

14. The method of claim 13 wherein detection of changed sectors further includes the following steps:

computing a checksum (or similar type of function) on groups of sectors read from said primary storage means,

comparing said checksum with the corresponding checksum stored from the previous backup,

operative when the two checksums do not match,

writing said group of sectors to said backup storage means,

writing said checksum to said backup storage means,

operative when the two checksums do match,

setting the entry (or entries) in said sector directory table corresponding to said group of sectors to point to the corresponding group of sectors from said previous backup.

15. The method of claim 13 wherein detection of changed sectors further includes the following steps:

activating monitor software to detect all writes to said primary storage means,

maintaining a dirty sector table indicating which groups of sectors on said primary storage means have been modified,

using said dirty sector table to determine which groups of sectors have been changed,

operative when said dirty sector table indicates that said group of sectors to be backed up has been modified,

writing said group of sectors to said backup storage means,

operative when said dirty sector table indicates that said group of sectors to be backed up has not been modified,

setting the entry in said sector directory table corresponding to said group of sectors to point to the corresponding group of sectors from said previous backup,

saving said dirty sector table to an auxiliary storage means when said monitor software is deactivated at system shutdown.

16. The method of claim 15 wherein said auxiliary storage means is the same as said primary storage means.

17. The method of any of claims 15 further including the following steps:

operative when said monitor software is deactivated,

computing a checksum on the contents of said dirty sector table,

saving said checksum on said auxiliary storage means,

operative when said monitor software is activated,

performing a validity check on said contents of said dirty sector table using said checksum,

invalidating said checksum on said auxiliary storage means.

18. The method of claim 17 further including the steps of:

operative when said monitor software is deactivated,

saving an indicator of the time of said deactivation on said auxiliary storage means,

operative when said monitor software is activated,

verifying that the operating system has not been active to allow writes to said primary storage means since the last time a valid dirty sector table was written to said auxiliary storage means,

operative when said verification fails, invalidating the contents of said dirty sector table.

19. The method of any of claims 18 wherein the failure of any checks on the validity of the contents of said dirty sector table results in all sectors being marked as having been modified, whereby a complete backup is performed.

20. The method of any of claims 1-6, further including the steps of:

creating a removable disk which contains all files necessary to boot said computer system into said operating system, including software drivers that allow access to said primary storage means and said backup storage means,

booting said computer system using said removable disk.

43. The methods of claim 31 wherein said backup storage means can also be used to perform a sector-by-sector restore as in claim 1.

21. A method for backing up data in a computer system from a primary storage means to a backup storage means on a sector-by-sector basis and for providing file-by-file access to said data on said backup storage means, said method comprising the steps of:

reading a set of logically contiguous sectors from the primary storage means using a software call of the operating system that provides access to the files stored on said primary storage means, said call of said operating system performing any physical level remapping necessary to avoid previously detected physical flaws on said primary storage means,

writing said set of logically contiguous sectors to said backup storage means,

identifying a control set of logical sectors of said primary storage means, said control set including sectors required to mount said primary storage means for file access by said operating system or to traverse the directory structure of the files on said primary storage means,

re-ordering the sequence of writing said sets of logically contiguous sectors on said backup storage means in order to group sectors of said control set in closer physical proximity to one another on said backup storage means than would occur if said sequence were ordered strictly by logical sector number,

caching said control set of logical sectors from said backup storage means to allow fast random access to said control set,

creating a virtual disk partition of said operating system,

servicing logical sector read requests on said virtual disk partition,

operative when a sector of said read request is part of said control set, reading said sector from said control cache,

operative when a sector of said read request is not part of said control set, reading said sector from said backup storage means,

mounting said virtual disk partition as a disk volume of said operating system,

whereby files on said disk volume may be accessed using normal operating system calls and utilities.

22. The method of claim 21, further including the steps of:

writing on said backup storage means a sector directory table containing information sufficient to indicate the size of said primary storage means and the location of each logical sector written to said backup storage means,

reading said sector directory table from said backup storage means,

using said sector directory table to determine the locations of said sectors when servicing logical sector read requests.

23. The method of claim 22 wherein said primary storage means consists one or more disk drive partition(s) in said computer system, and wherein said operating system call to read said logically contiguous sectors performs the mapping necessary to locate said logically contiguous sectors on said disk drive partition(s).

24. The method of claim 22 wherein unused sectors that do not contain file data are not read from said primary storage means and are not stored on said backup storage means, and wherein the absence of said unused sectors on the backup storage means is indicated in said sector directory table.

25. The method of claim 24 wherein deleted sectors that contain data from deleted files are not read from said primary storage means and are not stored on said backup storage means, and wherein the absence of said deleted sectors on the backup storage means is indicated in said sector directory table.

26. The method of claim 21 wherein an open file log is maintained of all files which are opened for write while the backup method is in process.

27. The method of any of claims 21-26 wherein said open file log is written to said backup storage means.

28. The method of any of claims 21-26 wherein said operating system allows multi-tasking and further including the step of:

temporarily suspending execution of any tasks that attempt to write a set of sectors to said primary storage means until said set of sectors has been read from said primary storage means by the backup task in preparation for writing said sectors to said backup storage means.

29. The method of claim 28, further including the steps of:

maintaining a cache of sectors read from said primary storage means to be written to said backup storage means,

detecting an attempted write by a task to a set of sectors of said primary storage means which has not yet been backed up,

operative when said sector cache is full, temporarily suspending executing of said task,

operative when said sector cache is not full, reading said set of sectors and adding said set of sectors to said sector cache and then allowing said task to continue execution without suspension,

checking said sector cache for the presence of any portion of said set of logically contiguous sectors to be read from said primary storage means,

operative when no such portion is found in said sector cache, reading said set of logically contiguous sectors from said primary storage means,

operative when such portion is found in said sector cache, reading said portion(s) of said set of logically contiguous sectors from said sector cache, and reading remaining portions not found in said sector cache from said primary storage means,

whereby no said tasks attempting to write to said primary storage means will be suspended unless said sector cache is full.

30. The method of claim 29 wherein said portions of said set of logically contiguous set of sectors found in said sector cache are removed from said sector cache after said portions are read from said sector cache, whereby portions of said sector cache may be re-used in order to minimize the number of times that tasks are suspended.

31. The method of any of claims 21-26 wherein only sectors that have changed since the last backup are written to the backup storage means.

32. The method of claim 31 wherein detection of changed sectors further includes the following steps:

computing a checksum (or similar type of function) on groups of sectors read from said primary storage means,

comparing said checksum with the corresponding checksum stored from the previous backup,

operative when the two checksums do not match,

writing said group of sectors to said backup storage means,

writing said checksum to said backup storage means,

operative when the two checksums do match,

setting the entry (or entries) in said sector directory table corresponding to said group of sectors to point to the corresponding group of sectors from said previous backup.

33. The method of claim 31 wherein detection of changed sectors further includes the following steps:

activating monitor software to detect all writes to said primary storage means,

maintaining a dirty sector table indicating which groups of sectors on said primary storage means have been modified,

using said dirty sector table to determine which groups of sectors have been changed,

operative when said dirty sector table indicates that said group of sectors to be backed up has been modified,

writing said group of sectors to said backup storage means,

operative when said dirty sector table indicates that said group of sectors to be backed up has not been modified,

setting the entry in said sector directory table corresponding to said group of sectors to point to the corresponding group of sectors from said previous backup,

saving said dirty sector table to an auxiliary storage means when said monitor software is deactivated at system shutdown.

34. The method of claim 33 wherein said auxiliary storage means is the same as said primary storage means.

35. The method of claim 33 further including the following steps:

operative when said monitor software is deactivated,

computing a checksum on the contents of said dirty sector table,

saving said checksum on said auxiliary storage means,

operative when said monitor software is activated,

performing a validity check on said contents of said dirty sector table using said checksum,

invalidating said checksum on said auxiliary storage means.

36. The method of claim 35 further including the steps of:

operative when said monitor software is deactivated,

saving an indicator of the time of said deactivation on said auxiliary storage means,

operative when said monitor software is activated,

verifying that the operating system has not been active to allow writes to said primary storage means since the last time a valid dirty sector table was written to said auxiliary storage means,

operative when said verification fails, invalidating the contents of said dirty sector table.

37. The method of claim 36 wherein the failure of any checks on the validity of the contents of said dirty sector table results in all sectors being marked as having been modified, whereby a complete backup is performed.

38. The methods of claims 21-26 wherein said control set is identified by using a knowledge of the file and allocation format of said primary storage means under said operating system.

39. The method of any of claims 21-26 wherein said control set is identified without a complete knowledge of said file and allocation format of said primary storage means, using a pseudo-drive technique which includes the following steps:

creating a temporary virtual disk partition of said operating system,

servicing logical sector read requests on said temporary virtual disk partition by performing reads of the corresponding sectors of said primary storage means,

monitoring the set of logical sectors that are read from said temporary virtual disk partition and adding each sector read to said control set,

mounting said temporary virtual disk partition as a temporary disk volume of said operating system.

40. The method of claim 39, further including the step of using operating system calls to traverse the entire directory tree of said temporary disk volume.

41. The method of any of claims 21-26 wherein some knowledge of said file and allocation format is used to eliminate duplicate copies of structures in said primary storage means from said control set, whereby the size of said control set is minimized.

42. The method of any of claims 21-26 wherein writes to said disk volume are allowed by caching said writes to a temporary storage means.