Abstract: Aspects of the subject matter described herein relate to resource replication systems. In aspects, a mechanism is described that detects when local resources are stale. Local resources may be stale when the time between a last successful synchronization activity and a current time exceeds a staleness value. If the local resources are determined to be stale, the local member may refrain from certain synchronization activities until the local member is allowed to resume these certain synchronization activities. If the local resources are not stale, additional synchronization activities including synchronizing resources between the member and an upstream or downstream partner may be performed.

Abstract: Aspects of the subject matter described herein relate to replicating initial content in resource replication systems. In aspects, each member participating in a replica group obtains a content set state from a directory service. The content set state indicates whether the member is allowed to synchronize resources with another member participating in the replica group. If the member is designated as a primary member, the member may replicate out resources to its downstream partners. If a member is not designated as a primary member, the member may not replicate out resources to its downstream partners until it has successfully synchronized resources with an upstream partner. Fence values may be set which control which resources win (and are replicated out) during synchronization.

Abstract: A method and system are related to updating objects over limited bandwidth networks. Objects are updated between two or more computing devices using remote differential compression (RDC) techniques such that required data transfers are minimized. In one aspect, efficient large object transfers are achieved by recursively applying the RDC algorithm to its own metadata; a single or multiple recursion step(s) may be used in this case to reduce the amount of metadata sent over the network by the RDC algorithm. Objects and/or signature and chunk length lists can be chunked by locating boundaries at dynamically determined locations. A mathematical function evaluates hash values associated within a horizon window relative to potential chunk boundary.

Abstract: A method and system for synchronizing objects among members of a replica set. An interval vector is maintained on each member. The interval vector includes one or more intervals each of which has an upper and lower bound. During synchronization, the member sends the interval vector to the member with which it is synchronizing. The other member examines the interval vector and finds resources with versions that are not included in any of the intervals of the interval vector. The other member then transmits these resources to the first member which uses them to update its store. The first member also updates its interval vector to account for the received resources.

Abstract: A method and system for synchronizing objects among members of a replica set. Objects on members of the replica set are quickly synchronized via a join algorithm that takes advantage of version vectors. When an object is deleted, the object is tombstoned (i.e., marked as “dead”). Tombstones are also replicated between the connected members of the replica set. After a period of time (e.g., a timeout), the tombstones are deleted. A slow sync algorithm executes periodically to ensure that objects that were not deleted via replicated tombstones are eventually deleted. An époque algorithm provides that members that have been offline for a long period of time will update their objects before they participate in a join synchronization.