Research on Data Partitioning Strategies for Parallel XML Database Systems

Data placement strategies are one of the important impacts on the performance of parallel database systems This paper focuses on data placements for declustering XML documents with very large size in parallel XML database systems, and two different data partitioning strategies with traditional data placements: path schema based path instance balancing （PSPIB） strategy and node schema based round robin （NSNRR） strategy The key idea of the former is that all path instances under the same path schema in a DOM tree are evenly declustered over all sites and then assigned to different sites The basic idea of the latter is that nodes with different node schemas in a DOM tree are assigned to different sites in a round robin way, while the nodes with same node schemas are assigned to a common site The performance tests, analyses and evaluations are given in this paper