Parallel query processing strategies for object-oriented temporal databases

An object-oriented (OO) temporal database management system can better meet the data management requirements of many applications because it not only provides powerful facilities for modeling and processing the structural and behavioral properties of complex objects but also supports the management of temporal data. However, due to the generality and high functionality of an OO system, the append-only nature of temporal databases and the irregular evolutions of object instances, processing efficiency is very difficult to achieve using serial algorithms and conventional computing systems. The paper presents two multi-wavefront parallel query processing strategies, five primitive time-alignment operations and their implementations on a parallel computer nCUBE2. The multi-wavefront algorithms allow asynchronous processing and propagation of temporal object instance identifiers and data in order to identify temporal object instances which satisfy a complex multi-class query. They offer both intraquery parallelism and interquery parallelism in query processing. The time-alignment operations allow data with different evolution time intervals to be aligned to determine the common intervals in which the combined factual information is valid. Some performance evaluation results are also reported