Efficient Parallel Semantic/O-O Database Management.

Abstract : We have developed an optimistic concurrency control method for a massively parallel semantic database machine. Our concurrency control algorithm achieves very fine granularity, ensures serializability and external consistency, and uses local logical clocks which do not require physical clock synchronization. We have also developed a dynamic load balancing algorithm which repartitions data among processors using a fault-tolerant data transfer policy to produce a more evenly balanced load. We have implemented benchmarks on our experimental semantic database system that have shown it to be more than competitive with current commercial products. In addition to these results, we have continued to perform research on semantic databases. Our research into applying SQL to semantic databases has shown the advantages of the semantic binary model even when using standard relational languages.