Achieving Communication Efficiency through Push-Pull Partitioning of Semantic Spaces in Client-Server Architectures

Client-server databases that require query results to be up-to-date despite storing data that changes dynamically suffer from heavy communication costs. Client-side caching can help mitigate these costs, particularly when individual PUSH-PULL decisions are made for the different semantic regions in the data space. In the PUSH regions the server notifies the client about updates, and in the PULL regions the client sends queries to the server. We call the problem of partitioning the data space into PUSH-PULL regions to achieve the minimum possible communication cost for a given workload the problem of data gerrymandering. In this paper we present solutions under different communication cost models for a frequently encountered scenario: with range queries and point updates. Specifically, we give a provably optimal-cost dynamic programming algorithm for gerrymandering on a single range query attribute. We propose a family of heuristics for gerrymandering on multiple range query attributes. We also handle the dynamic case in which the workload evolves over time. We validate our methods through extensive experiments on real and synthetic data sets.