International Journal of Computer Science Issues

More than a traditional journal...

An effective strategy for distributing data across multiple disks is crucial to achieving good performance in a parallel object-oriented database management system. During query processing, a large amount of data need to be processed and transferred among the processing nodes in the system. A good data placement strategy should be able to reduce the communication overheads, and, at the same time, to provide the opportunity for exploiting different types of parallelism in query processing, such as intra-operator parallelism, inter-operator parallelism, and inter-query parallelism. However, there exists a conflict between these two requirements. While minimizing interprocessor communication favors the assignment of the whole database to a small number of processors, achieving higher degree of parallelism favors the distributions of the database evenly among a large number of processors. A trade-off must be made to obtain a good policy for mapping the database to the processors.We need good heuristics to solve this and more complicated database allocation problems. In this paper, we propose some heuristics for partitioning an OODB so that the overall execution time can be reduced.