Method for taking snapshots of data

• US 7,363,444 B2
• Filed: 01/10/2005
• Issued: 04/22/2008
• Est. Priority Date: 01/10/2005
• Status: Expired due to Fees

• Current Assignee: Hewlett Packard Enterprise Development LP (Hewlett-Packard Enterprise Company)
• Original Assignee: Hewlett-Packard Development Company, L.P. (HP Inc.)
• Inventors: Ji, Minwen
• Primary Examiner(s): Kim; Matthew
• Assistant Examiner(s): DUDEK JR, EDWARD J
• Application Number:
  US11/033,343
• Publication Number:
  US 20060155946A1
• Time in Patent Office: 1,198 Days
• Field of Search: 711/161
• US Class Current: 711/161
• CPC Class Codes: G06F 3/0617 in relation to availability
• CPC Class Codes: G06F 3/0619 in relation to data integrity, e.g. data losses, bit errors
• CPC Class Codes: G06F 3/065 Replication mechanisms
• CPC Class Codes: G06F 3/0689 Disk arrays, e.g. RAID, JBOD

Abstract

A method for taking snapshots of data. In an embodiment, a first map data structure is obtained that records locations for a plurality of data blocks. A second map data structure is formed that is initially empty of locations, the second map data structure representing the snapshot after the snapshot is started. Writes on data that arrive before the snapshot starts are applied to the first map, and writes that arrive after the snapshot starts are applied to the second map.

Claims

1. A method of taking a snapshot of data comprising:

obtaining a first map data structure that specifies a relationship of logical addresses to physical addresses for a plurality of data blocks of a volume, the first map data structure initially being identified as a current map for the volume;

forming a second map data structure that is initially empty, the second map data structure initially being identified as a tentative map data structure for the volume;

committing the snapshot by identifying the second map data structure as the current map for the volume and identifying the first map data structure as a snapshot for the volume;

taking one or more additional snapshots by repeating said steps of forming and committing for each additional snapshot, each snapshot being completed upon being committed and the snapshots being linked such that each snapshot is linked to a next most-recent snapshot; and

performing writes to data blocks of the volume while taking the one or more additional snapshots by writing a data block to a new physical location and updating the map identified as the current map at the time of the write with a physical address of the data block being written, wherein each committed snapshot only contains values corresponding to the physical addresses of the data blocks of the volume that were written to while the map data structure for the snapshot was identified as the current map.

2. The method according to claim 1, further comprising overwriting the data in place without updating the second data structure when a data block is updated again in response to a write.

3. The method according to claim 1, wherein the volume is distributed among a plurality of storage devices and wherein said steps of forming and committing are performed at sufficient ones of the plurality of the storage devices that the snapshot can be reconstructed.

4. The method according to claim 3, wherein the volume is distributed and replicated among the plurality of storage devices.

34. The method according to claim 1, further comprising reading a data block that was last updated prior to a completed snapshot by following a plurality of the links to locate the most-recent version of the data block.

5. A method of taking a snapshot of a data distributed among a plurality of data storage devices, the plurality of data storage devices redundantly storing the data and each having a map data structure that specifies a relationship of logical addresses to physical addresses for a plurality of blocks of the data, the method comprising:

forming new map data structures by at least a quorum but not necessarily all of the storage devices, each new map data structure initially being empty of physical addresses and initially being identified as a tentative map data structure; and

committing the snapshot by at least a quorum but not necessarily all of the storage devices by identifying the map data structures as snapshot maps for the data and identifying the new map data structures as current maps for the data.

6. The method according to claim 5, further comprising inserting a new physical address into the new data structures when a data block is updated by a new version of the data, the new version of the data being stored at the new physical address.

7. The method according to claim 6, further comprising overwriting the new version of the data in place without updating the new data structure when the data block is updated again.

8. The method according to claim 5, further comprising taking one or more additional snapshots by repeating said steps of forming a and committing for each additional snapshot.

9. The method according to claim 5, wherein the data is replicated.

10. The method according to claim 5, wherein the data is erasure coded.

11. The method according to claim 5, the plurality of data storage devices being a group of devices designated for redundantly storing at least a portion of a data volume and wherein one or more additional groups of storage devices are each designated for storing a corresponding portion of the data volume and the quorum comprising a quorum but not necessarily all of each of the groups of the devices.

12. The method according to claim 5, said forming comprising sending a first command and wherein each storage device that successfully forms its new map data structure in response to the first command returns an affirmative response.

13. The method according to claim 12, said committing being performed in response to receiving affirmative responses from the quorum.

14. The method according to claim 13, said committing comprising sending a second command and said committing is performed by the storage devices in response to the second command.

15. A method of taking a snapshot of a data distributed among a plurality of data storage devices, the plurality of data storage devices redundantly storing the data and each having a map data structure That specifies a relationship of logical addresses to physical addresses for a plurality of blocks of the data, the method comprising:

sending a first command to the plurality of storage devices instructing the snapshot of the data to be prepared;

receiving any affirmative responses from the plurality of storage devices, each affirmative response indicating that the corresponding storage device is able to form a new map data structure, the new map data structure initially being empty of physical addresses; and

sending a second command to the plurality of storage devices, the second command instructing the snapshot to be committed if a quorum of affirmative responses is received, but not necessarily a response from each of the plurality of storage devices being received, and the second command instructing the snapshot to be aborted if the quorum of affirmative responses is not received.

16. The method according to claim 15, further comprising inserting a new physical address into the new data structures when a data block is updated by a new version of the data, the new version of the data being stored at the new physical address.

17. The method according to claim 16, further comprising overwriting the new version of the data in place without updating the new data structure when the data block is updated again.

18. The method according to claim 15, further comprising taking one or more additional snapshots by repeating said steps of sending the first command, receiving any responses and sending the second command for each additional snapshot.

19. The method according to claim 15, wherein the data is replicated.

20. The method according to claim 15, wherein the data is erasure coded.

21. The method according to claim 15, the plurality of data storage devices being a group of devices designated for redundantly storing at least a portion of a data volume and wherein one or more additional groups of storage devices are each designated for storing a corresponding portion of the data volume and the quorum comprising a quorum of each of the groups of the devices.

22. The method-according to claim 15, the map data structure for each storage device initially being identified as a current map for the data and the new map data structure for each storage device initially being identified as a tentative map data structure.

23. The method according to claim 22, wherein in response to being instructed to commit the snapshot, each storage device identifies its map data structure as the snapshot map for the data and identifies the new map data structure as its current map for the data.

24. The method according to claim 22, wherein write operations are directed to the tentative map for the data after the first command is received and before the snapshot is committed or aborted and, otherwise, the write operations are directed to the current map for the data.

25. The method according to claim 22, wherein in response to being instructed to abort the snapshot, each storage device merges any write operations in its tentative map for the data into its current map for the data and deletes or ignores its tentative map.

26. The method according to claim 22, wherein a timestamp for the snapshot is associated with the first command and each storage device has information that indicates the last time that a snapshot was started on the volume and the last time that any data corresponding to the current map was updated or prepared for update.

31. The method according to claim 15, the plurality of data storage devices being a group of devices designated for redundantly storing at least a portion of a data volume and wherein one or more additional groups of storage devices are each designated for storing a corresponding portion of the data volume add the snapshot being determined valid or invalid from information obtained from a quorum but not necessarily all of the groups of the storage devices.

32. The method according to claim 15, the plurality of data storage devices being a group of devices designated for redundantly storing at least a portion of a data volume and wherein one or more additional groups of storage devices are each designated for storing a corresponding portion of the data volume and the snapshot being determined valid or invalid from information obtained from only one group of the storage devices.

33. The method according to claim 31, further comprising recovering the snapshot if the snapshot is determined invalid.

35. The method according to claim 15, wherein each of the data storage devices stores a replica of the data and the quorum comprises a majority of the data storage devices.

36. The method according to claim 15, wherein the data is erasure coded where m data blocks and p parity blocks are stored across a set of n of the storage devices, where n=m +p, the quorum is m+p/2 of the storage devices if p is even, and the quorum is m+(p+1)/2 if p is odd.

27. A method of taking a snapshot of data distributed among a plurality of data storage devices, the plurality of data storage devices redundantly storing the data and each having a map data structure that specifies a relationship of logical addresses to physical addresses for a plurality of blocks of the data and the map data structure for each storage device initially being identified as a current map for the data, the method comprising:

sending a first command to the plurality of storage devices instructing the snapshot of the data to be prepared, wherein a timestamp for the snapshot is associated with the first command and each storage device has information that indicates the last time that a snapshot was started on the volume and the last time that any data corresponding to the current map was updated or prepared for update and wherein each storage device receiving the first command determines from the timestamp and the information whether data corresponding to the current map was updated or prepared for update after a time indicated by the timestamp;

receiving any affirmative responses from the plurality of storage devices, each affirmative response indicating that the corresponding storage device is able to form a new map data structure, the new map data structure initially being empty of physical addresses and the new map data structure for each storage device initially being identified as a tentative map data structure; and

sending a second command to the plurality of storage devices, the second command instructing the snapshot to be committed if a quorum of affirmative responses is received, but not necessarily a response from each of the plurality of storage devices being received, and the second command instructing the snapshot to be aborted if the quorum of affirmative responses is not received.

28. The method according to claim 27, wherein if the data corresponding to the current map was updated or prepared for update after a time indicated by the timestamp at a storage device, the storage device does not return an affirmative response to the first command.

29. The method according to claim 28, the value of the timestamp for a retried snapshot after an abort is determined by adding to a present time an adjustment period in response to one or more attempted snapshots being aborted.

30. The method according to claim 29, wherein the adjustment period is at least the difference in time between the timestamp of an aborted snapshot and last time that any data corresponding to the current map was updated or prepared for update.

37. A computer readable media comprising computer code for implementing a method of taking a snapshot of data, the method comprising the steps of:

obtaining a first map data structure that specifies a relationship of logical addresses to physical addresses for a plurality of data blocks of a volume, the first map data structure initially being identified as a current map for the volume;

forming a second map data structure that is initially empty, the second map data structure initially being identified as a tentative map data structure for the volume;

committing the snapshot by identifying the second map data structure as the current map for the volume and identifying the first map data structure as a snapshot for the volume;

taking one or more additional snapshots by repeating said steps of forming and committing for each additional snapshot, each snapshot being completed upon being committed and the snapshots being linked such that each snapshot is linked to a next most-recent snapshot, and

performing writes to data blocks of the volume while taking the one or more additional snapshots by writing a data block to a new physical location and updating the map identified as the current map at the time of the write with a physical address of the data block being written, wherein each committed snapshot only contains values corresponding to the physical addresses of the data blocks of the volume that were written to while the map data structure for the snapshot was identified as the current map.