Abstract:

Database replication is usually used to keep data consistency and enhance disaster tolerance and data accessibility. A majority of replication tools are asynchronous, which always face increment replication problems. In this paper, based on the tool named “Daffodil Replicator”, a “windowbased net increment replication” algorithm is used to balance the resource pressure and the time cost when the program generates, delivers and parses net increment data. To optimize this algorithm, sampling is used to analyze the best parameters for the timecost function.

Key words: data synchronization;replication;net increment replication;distributed