Approximate Web Database Snapshots

The amount of data stored in databases is increasing at a tremendous rate. They are oftentimes stored over the web to be accessed by various clients. One useful and interesting query to a collection of databases is to capture a consistent snapshot of a set of interested attributes across the databases. While traditional snapshot algorithms for a distributed database could be adopted, they are mainly designed for database recovery and are costly to execute. We need efficient algorithms to return good-enough snapshots for querying purposes, in the absence of a consistent but costly snapshot as defined in distributed systems. In this paper, we propose the notion of an approximate consistent snapshot by considering and minimizing the deviation of the collected snapshot covering the interested attributes from a reference absolute snapshot. There are several variations of this kind of approximate snapshots. We propose effective algorithms to capture such snapshots in a collection of databases and then conduct performance evaluation on their efficiency. We believe that this notion of approximate consistency would be useful in practical situations.