Replication-Aware Query Processing in Large-Scale Distributed Information Systems ∗
Jie Xu, Alexandros Labrinidis · 2006
In this work, we address the problem of replica selection in distributed query processing over the Web, in the presence of user preferences for Quality of Service and Quality of Data. In particular, we propose RAQP, which stands for Replication-Aware Query Processing. RAQP uses an initial statically-optimized logical plan, and then selects the execution site for each operator and also selects which replica to use, thus converting the logical plan to an executable plan. Unlike prior work, we do not perform an exhaustive search for the second phase, which allows RAQP to scale significantly better. Extensive experiments show that our scheme can provide improvements in both query response time and overall quality of QoS and QoD as compared to random site allocation with iterative improvement. 1.