Distributed Query Optimization[go to overview]
In recent years distributed RDF stores have been developed in order to properly scale RDF stores with the increasing capacity of RDF data, that is available on the Web. Distributed RDF stores exceed the storage and query processing capabilities of regular centralized RDF stores by combining the computational power of several compute nodes.
When processing a query request, the structure of the query plan is crucial for the query performance. By adjusting the order of join operations, the amount of intermediate results that arise during query processing can usually be reduced. Consequently, fewer intermediate results have to be transferred and handled by the compute nodes during query execution. Query planning is fairly well studied in the context of centralized RDF stores, however it still remains a considerable challenge given the setting of distributed RDF stores. This talk presents a distributed query optimization approach that aims at jointly minimizing the query workload on all compute nodes. For this purpose the approach optimizes the query plan independently for each compute node.
07.11.19 - 10:15