Remote Comparison of Database Tables

摘要

Database systems hold mission critical data in all organizations. These data are often replicated for being processed by different applications as well as for disaster recovery. In order to help handle these replications, remote sets of data must be compared to detect unwanted changes due to hardware, system, software, application, communication or human errors. We present an algorithm based on operations and functions already available in relational database systems to reconcile remote tables by identifying inserted, updated or deleted tuples with a small amount of communication. A tree of checksums, which covers the table contents, is computed on each side and merged level by level to identify the differing keys. A prototype implementation is available as a free software. Experiments show our approach to be effective even for tables available on a local network. This algorithm provides a communication efficient and general solution for comparing remote database tables.