2.3 Basic Conflict Resolution

The SData algorithm uses a simple mechanism based on priorities and timestamps to resolve conflicts. The idea is the following:

• Every node is given a “conflict priority”. Usually, the priorities will be rather static but SData also supports scenarios where the priorities are modified dynamically while the system runs. The node with the smallest conflict priority value wins.

• If the conflict occurs between nodes of equal priorities, the modification timestamp of the resource is used to solve the conflict. The node where the resource was last modified wins.

• In the very improbable case where the modification timestamps are equal, some arbitrary rule is applied (smallest node id wins for example).

This can be easily implemented by adding a priority value to the vector clock pairs. The vector clock becomes a list of triplets (Ni, Ti, Pi) where Ni is the node id, Ti is Ni’s tick when R was last modified at Ni, and Pi is Ni’s conflict priority. Also, the synchronization metadata should contain the last modification timestamp of each resource.

So, the synchronization metadata for resource R is the combination of its extended vector clock (list of triplets) and its last modification timestamp.

The conflict resolution algorithm works as described in the following table (we reduced timestamp to hh:mm to keep table clean):