Abstract: Distributed query processing, when blank nodes might occur in sources and only the signature of sources are known, may in the worst case require that all possible query partitions be evaluated in order to ensure soundness and completeness of answers. The work presented in this paper attempts to push the boundary as to what query sizes and structures lies within practical feasibility. The presented approach utilizes semantic information obtained by probing sources for occurrences of blank nodes, interleaved in a recursive algorithm for calculating restricted growth strings, in order to detect and abort unfruitful branches in the partition generating process. The approach is evaluated against a well-known SPARQL benchmark, modified for the distributed case, and tentative conclusions regarding the effectiveness are drawn.
0 Replies
Loading