Towards Optimal Snapshot Materialization to Support Large Query Workload for Append-Only Temporal Databases

Abstract

We present several results on optimal snapshot materialization for append-only temporal databases in order to support very large scale query workload. Our data model is temporal relational data stored in an append-only database. When the temporal database receives multiple queries querying at different timestamps along the timeline, it would be prohibitively expensive to recompute the snapshots at each of the timestamps. In this paper, we present a practical solution to support large query load by materializing m snapshots at optimal timestamps. We show that optimal snapshot timestamps can be computed efficiently in linear time. We further show that with varying query load, we can dynamically adjust snapshots to adjust to the changin query load.