An Optimistic Replication Algorithm to Improve Consistency for Massive Data

摘要

Data replication introduces well-known consistency issues. For massive data, how to improve update propagation and how to minimizing space overhead effectually are important. An optimistic replication algorithm is proposed. In our algorithm, home replica is used to resolve updates conflict and pair-wise communication supports the reconciliation of any two replicas. A new anti-entropy partner selection method based on the distribution of updates and the information of local write-log is presented. The algorithm uses write-log truncation appropriately during updates propagation to remove out-of-date updates in time. The simulation results show that the partner selection mechanism can achieve good scalability and adaptability, the average number of updates in write-log does not exceed the number of replicas, and the variety for running different numbers of updates is not obvious.