Distributed Query Optimization[go to overview]
This thesis focuses on the query optimization of graph databases that are distributed over several compute nodes. Most of these distributed graph databases optimize the query centrally. The resulting optimized query plan is then executed on all compute nodes. Since the different compute nodes store different data items, the centrally optimized query plan may be inefficient on some compute nodes. To overcome this limitation, this thesis proposes and investigates a distributed optimization approach that optimizes an individual query plan for each compute node while considering the stored data. Thereby the presented approach aims to improve the query performance by speeding up the processing of the query plan on all slave nodes. A side effect of the distributed optimization approach is the duplication of certain intermediate results during query processing. The evaluation shows that the actual performance gain of the approach cannot be determined, given the additional workload caused by processing these duplicate intermediate results.
16.01.20 - 10:15