Granular control over the authority of replicated information via fencing and unfencing

• US 20050015413A1
• Filed: 12/10/2003
• Published: 01/20/2005
• Est. Priority Date: 07/10/2003
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Bjorner, Nikolaj S., Teodosiu, Dan
• Granted Patent: US 7,660,833 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/1844 Management specifically adapted to replicated file systems
• CPC Class Codes: Y10S 707/99954 Version management

Abstract

A method and system for controlling which content gets precedence and is replicated. A replica set is comprised of a set of resources. Each resource is associated with resource data and resource meta-data. For files-based systems, resource data includes file contents and attributes, while resource meta-data includes additional attributes that are relevant for negotiating synchronization during replication. An extra field called a “fence value” is added to the meta-data associated with each resource. During synchronization, first fence values are compared. The resource with the highest fence value includes the content that is controlling and replicated. If fence values are equal (and greater than a particular value), the controlling resource is determined based on other meta-data.

Claims

1. A computer-readable medium having computer-executable instructions, comprising:

retrieving meta-data of a resource that is replicated on a plurality of machines, the resource having meta-data and content that resides on each machine on which the resource is replicated, the meta-data including one or more values that are updated whenever the content of the resource is changed via any local update and a fence value that is independent of any local changes to the content;

comparing a first fence value of the content on a first machine of the plurality of machines with a second fence value of the content on a second machine of the plurality of machines; and

if the first fence value is of higher precedence than the second fence value, updating the second machine.

2. The computer-readable medium of claim 1, wherein the meta-data is stored in a store separate from the content.

3. The computer-readable medium of claim 1, wherein updating the second machine comprises determining one or more differences between the content on the first and second machines and transmitting the one or more differences.

4. The computer-readable medium of claim 1, wherein updating the second machine comprises transmitting the meta-data only.

5. The computer-readable medium of claim 4, wherein the content on the first and second machines is the same.

6. The computer-readable medium of claim 1, wherein the content comprises file data and file attributes.

7. The computcr-rcadable medium of claim 1, wherein each meta-data on each machine comprises a digest that summarizes the resource.

9. The computer-readable medium of claim 1, wherein updating the second machine comprises updating the second fence value to equal the first fence value.

10. The computer-readable medium of claim 1, further comprising if the fence values are equivalent, comparing other data in the meta-data to determine whether content should be updated.

11. The computer-readable medium of claim 1, wherein each fence value is assigned to a portion or portions of its respective resource.

12. The computer-readable medium of claim 1, wherein content with a certain fence value is not propagated to other machines.

13. The computer-readable medium of claim 1, wherein content with a certain fence value is invisible to other machines.

14. The computer-readable medium of claim 1, further comprising setting one of the fence values to a maximum of (itself +1) and (a clock time) of the machine on which the resource including the one fence value resides.

15. The computer-readable medium of claim 1, further comprising keeping each fence value the same as the content associated with the fence value changes.

8. The computer-readable medium of claim 8, further comprising comparing the digests of the meta-data on the machines and bypassing updating if the digests are equivalent.

16. A computer-readabte medium having computer-executable instructions, comprising:

determining whether a first resource residing on a first machine should be used to update a second resource residing on a second machine, each resource associated with meta-data and content, each meta-data including one or more fields that are updated whenever the content of the associated resource is changed and a fence value, each fence value indicating whether its associated resource should be used to update a resource on another machine at a higher precedence than from other meta-data;

if the fence value of the second resource indicates that the second resource should not be propagated from the second machine, preventing propagation from the second machine; and

if the fence value of the first resource is of a higher precedence than the fence value of the second resource, updating the second resource from the first resource.

17. The computer-readable medium of claim 16, further comprising if the fence values of the first and second resources are equivalent, determining which machine will update the other based on meta-data other than the fence values.

18. The computer-readable medium of claim 17, wherein the other meta-data comprises a logical clock indicating the last time the corresponding content was updated.

19. The computer-readable medium of claim 18, wherein a fence value indicates that its corresponding resource may be propagated to other machines until another resource with a higher fence value is located on another machine.

20. The computer-readable medium of claim 16, wherein the meta-data associated with the first resource is stored in a separate data structure from its corresponding resource.

21. The computer-readable medium of claim 20, wherein the data structure is corrupted or deleted, further comprising rebuilding the data structure and decrementing the fence value associated with the first resource.

22. The computer-readable medium of claim 21, further comprising rebuilding the data structure a plurality of times and decrementing the fence value associated with the first resource each time the data structure is rebuilt.

23. The computer-readable medium of claim 22, further comprising if the fence value of the second resource is of a higher precedence than the fence value of the first resource, updating the first resource from the second resource.

24. A method for replicating data, comprising:

loading data onto a first machine;

marking the data with a fence value that indicates that the data should not be transmitted from the first machine to update data on any other machine;

synchronizing at least a portion of the data with data on a second machine.

25. The method of claim 24, further comprising after synchronizing the at least a portion of the data, marking the remaining portion of the data with a fence value that indicates that the data can be transmitted from the first machine to update data on any other machine.

26. The method of claim 24, further comprising after synchronizing the at least a portion of the data, deleting the remaining portion of the data.

27. The method of claim 24, wherein the data is loaded from a backup.

28. The method of claim 24, wherein the data is loaded from a copy of data sent from another machine.

29. The method of claim 24, further comprising marking corresponding data on a second machine that indicates that the corresponding data should be synchronized during synchronization.

30. The method of claim 24, further comprising changing a portion of the data and marking the changed portion with a fence value that indicates that the changed portion should be synchronized during synchronization.

31. A system for replicating data, comprising:

a first machine having a first set of resources;

a second machine having a second set of resource, wherein each resource on each machine is associated with meta-data and content, each meta-data including one or more fields that are updated whenever the content of the associated resource is changed and a fence value, each fence value indicating whether its associated resource should be used to update a resource on another machine independently from other meta-data, wherein the first and second machines are configured to;

communicate information regarding the resources contained by both, and

update each resource that is out-of-date according to the following precedence;

if a fence value of a resource on one of the machines is of higher precedence than the fence value of a corresponding resource on the other machine, updating the other machine with the resource on the one machine;

otherwise

updating the resource on the machines based on data other than the fence values.

32. The system of claim 31, wherein the first set of resources is loaded from a backup and the fence values thereof are set to cause the first set of resources to have precedence over any other set of resources, such that any other set of resources on any other machine that corresponds to the set of resources are updated from the first set of resources.

33. The system of claim 31, wherein the fence values of the first set of resources are marked to have precedence over corresponding resources on other machines, such that the corresponding resources are updated from the first set of resources.