Fault tolerant multi-stage data replication with relaxed coherency guarantees

• US 7,653,668 B1
• Filed: 11/23/2005
• Issued: 01/26/2010
• Est. Priority Date: 11/23/2005
• Status: Expired due to Fees

• Current Assignee: Veritas Technologies, LLC (Whitehouse Group Ltd.)
• Original Assignee: Symantec Operating Corporation (NortonLifeLock Inc.)
• Inventors: Ranade, Dilip Madhusudan, Banerjee, Anindya, Kabra, Navin, Nanda, BijayaLaxmi, Shelat, Radha
• Primary Examiner(s): Vital; Pierre M
• Assistant Examiner(s): Raab; Christopher J
• Application Number:
  US11/286,619
• Time in Patent Office: 1,525 Days
• Field of Search: 707/1, 707/10, 707/100, 707201-204, 711/162, 709/201
• US Class Current: 707/610
• CPC Class Codes: G06F 11/2094 Redundant storage or storage space G06F11/2056 takes precedence
• CPC Class Codes: G06F 16/184 implemented as replicated file system

Abstract

A system for fault tolerant multi-stage data replication with relaxed coherency comprises one or more processors and memory coupled to the processors, where the memory stores program instructions executable by the processors to implement a data replication service. In response to an update request directed at a data object, in a first stage of replication, the data replication service is configured to update respective primary replicas at a plurality of primary replica sites for the data object, and to store metadata descriptors at each of the plurality of primary replica sites. The data replication service is configured to determine whether a notification of the update is to be sent to any other replica sites. In response to identifying such other replica sites, an update notification including contents of a metadata descriptor may be sent to the other replica sites in one or more additional replication stages.

Claims

1. A system, comprising:

one or more processors; and

memory storing program instructions executable by the one or more processors to implement a data replication service configured to;

in response to an update request directed at a data object that includes a plurality of data blocks,

update a respective replica of the data object at each primary replica site of a plurality of primary replica sites of the data object, wherein the respective replica includes a respective copy of the data blocks of the data object, wherein updating the respective replica comprises changing data of two or more of the data blocks in the respective copy of the data blocks, wherein the data replication service is configured to update the replicas at the plurality of primary replica sites synchronously with respect to the update request;

store metadata specifying which data blocks were changed by the update at each primary replica site of the plurality of primary replica sites;

in response to determining that a notification of the update is to be sent to a particular additional replica site, send the metadata specifying which data blocks were changed by the update in a notification to the particular additional replica site, wherein the notification does not include the data of the data blocks changed by the update; and

in response to receiving the notification at the particular additional replica site, use the metadata to determine which data blocks were changed by the update, and mark each of the data blocks changed by the update as invalid in an additional replica of the data object at the particular additional replica site, wherein the data replication service is configured to mark the data blocks changed by the update as invalid in the additional replica asynchronously with respect to the update request.

2. The system as recited in claim 1, wherein the data replication service is configured to update the respective replicas at the plurality of primary replica sites within a single transaction, and send the notification to the particular additional replica site after the single transaction is completed.

3. The system as recited in claim 1, wherein the update request is received from a client, wherein the data replication service is configured to update the respective replicas and to store the metadata before sending an indication of an update completion to the client, wherein the notification of the update is sent to the particular additional replica site after the indication of the update completion is sent to the client.

4. The system as recited in claim 1, wherein the data replication service is configured to:

designate a set of two or more nodes of a distributed network of nodes as primary replica sites for the data object, wherein the nodes of the set are linked via a local area network (LAN), wherein the plurality of primary replica sites comprises a quorum of the set.

5. The system as recited in claim 1, wherein the particular additional replica site is linked to the particular primary replica site via a wide area network (WAN).

6. The system as recited in claim 1, wherein the particular additional replica site is a first additional replica site, wherein the notification is a first notification, wherein the data replication service is further configured to:

in response to determining that a notification of the update is to be sent to a second additional replica site, send the data of the data blocks changed by the update in a second notification to the second additional replica site.

7. The system as recited in claim 6, wherein, in response to receiving the second notification including the data of the data blocks changed by the update at the second additional replica site, the data replication service is further configured to store the data of the data blocks changed by the update in a locally cached version of the data object at the second additional replica site.

8. The system as recited in claim 1, wherein the data replication service is further configured to include a version number of the data object in the notification sent to the particular additional replica site.

9. The system as recited in claim 1, wherein, in response to receiving the notification at the particular additional replica site, the data replication service is configured to:

send a request to a particular primary replica site to retrieve the data of the data blocks changed by the update.

10. The system as recited in claim 1, wherein the data replication service is configured to use a configurable parameter to determine whether the notification should include the data of the data blocks changed by the update.

11. The system as recited in claim 1, wherein the data replication service is further configured to:

maintain a respective latest coherent timestamp associated with the data object at each replica site of the plurality of primary replica sites and the additional replica site, wherein the latest coherent timestamp at a given replica site indicates a latest point in time at which an indication of an update to the data object was received at the site, and

determine, using the respective latest coherent timestamp at a respective replica site, whether a local replica of the data object at the respective replica site is to be synchronized with another replica at another replica site.

12. The system as recited in claim 1, wherein the additional replica of the data object at the particular additional replica site is a sparse replica of the data object, wherein the sparse replica does not include at least one data block of the plurality of data blocks of the data object.

25. The system of claim 1,

wherein the notification is sent to the particular additional replica site asynchronously with respect to the update request.

13. A tangible, computer readable medium comprising program instructions, wherein the instructions are computer-executable to implement a data replication service configured to:

in response to an update request directed at a data object that includes a plurality of data blocks,

update a respective replica of the data object at each primary replica site of a plurality of primary replica sites of the data object, wherein the respective replica includes a respective copy of the data blocks of the data object, wherein updating the respective replica comprises changing data of two or more of the data blocks in the respective copy of the data blocks wherein the data replication service is configured to update the replicas at the plurality of primary replica sites synchronously with respect to the update request;

store metadata specifying which data blocks were changed by the update at each primary replica site of the plurality of primary replica sites;

in response to determining that a notification of the update is to be sent to a particular additional replica site, send the metadata specifying which data blocks were changed by the update in a notification to the particular additional replica site, wherein the notification does not include the data of the data blocks changed by the update; and

in response to receiving the notification at the particular additional replica site, use the metadata to determine which data blocks were changed by the update, and mark each of the data blocks changed by the update as invalid in an additional replica of the data object at the particular additional replica site, wherein the data replication service is configured to mark the data blocks changed by the update as invalid in the additional replica asynchronously with respect to the update request.

14. The computer readable medium as recited in claim 13, wherein the data replication service is configured to update the respective replicas at the plurality of primary replica sites within a single transaction, and send the notification to the particular additional replica site after the single transaction is completed.

15. The computer readable medium as recited in claim 13, wherein the update request is received from a client, wherein the data replication service is configured to update the respective replicas and to store the metadata before sending an indication of an update completion to the client, wherein the notification of the update is sent to the particular additional replica site after the indication of the update completion is sent to the client.

16. The computer readable medium as recited in claim 13, wherein the data replication service is further configured to:

determine, based on a value of a configurable parameter, whether the notification sent to the particular additional replica site should include the data of the data blocks changed by the update.

17. The computer readable medium as recited in claim 13, wherein the data replication service is further configured to include a version number of the data object in the notification sent to the particular additional replica site.

18. The computer readable medium as recited in claim 13, wherein, in response to receiving the notification at the particular additional replica site, the data replication service is configured to:

send a request to a particular primary replica site to retrieve the data of the data blocks changed by the update.

19. A method comprising:

updating a respective replica of a data object at each primary replica site of a plurality of primary replica sites of the data object in response to receiving an update request directed at the data object, wherein the respective replica includes a respective copy of a plurality of data blocks of the data object, wherein updating the respective replica comprises changing data of two or more of the data blocks in the respective copy of the data blocks, wherein the replicas at the plurality of primary replica sites are updated synchronously with respect to the update request;

storing metadata specifying which data blocks were changed by the update at each primary replica site of the plurality of primary replica sites;

in response to determining that a notification of the update is to be sent to a particular additional replica site, sending the metadata specifying which data blocks were changed by the update in a notification to the particular additional replica site, wherein the notification does not include the data of the data blocks changed by the update; and

in response to receiving the notification at the particular additional replica site, using the metadata to determine which data blocks were changed by the update, and marking each of the data blocks changed by the update as invalid in an additional replica of the data object at the particular additional replica site, wherein the data blocks changed by the update are marked as invalid in the additional replica asynchronously with respect to the update request.

20. The method as recited in claim 19, wherein the respective replicas are updated at the plurality of primary replica sites within a single transaction, and wherein the notification is sent to the particular additional replica site after the single transaction is completed.

21. The method as recited in claim 19, wherein the update request is received from a client, further comprising:

sending an indication of an update completion to the client, wherein the update completion indication is sent to the client after updating the respective replicas and storing the metadata and before sending the notification of the update to the particular additional replica site.

22. The method as recited in claim 19, further comprising:

determining, based on a value of a configurable parameter, whether to include the data of the data blocks changed by the update in the notification sent to the particular additional replica site.

23. The method as recited in claim 19, wherein the notification sent to the particular additional replica site also includes a version number of the data object.

24. The method as recited in claim 19, further comprising:

in response to receiving the notification at the particular additional replica site, sending a request to a particular primary replica site to retrieve the data of the data blocks changed by the update.

26. A system, comprising:

a plurality of primary replica nodes; and

an additional replica node;

wherein, in response to a request to update two or more data blocks of a plurality of data blocks of a data object, each respective primary replica node of the plurality of primary replica nodes is configured to update a respective replica of the data object stored at the respective primary replica node by changing data of the two or more data blocks in the respective replica of the data object synchronously with respect to the request;

wherein one or more primary replica nodes of the plurality of primary replica nodes are configured to store metadata specifying which data blocks were changed in response to the request;

wherein the plurality of primary replica nodes includes a particular primary replica node configured to send the metadata specifying which data blocks were changed in a notification to the additional replica node, wherein the notification does not include the data of the data blocks that were changed; and

wherein, in response to receiving the notification, the additional replica node is configured to use the metadata to determine which data blocks were changed, and mark each of the changed data blocks as invalid in an additional replica of the data object stored at the additional replica node, wherein the additional replica node is configured to mark the changed data blocks as invalid in the additional replica asynchronously with respect to the request.