Abstract
Contemporary semantic database engines support parallel computation of multiple queries. However, they do not apply parallelism to processing individual queries.
In this paper, we propose decomposition of basic SPARQL operations to increase parallelism during the query evaluation. For efficient implementation, we use the combination of Bobox runtime and Bobolang streaming language.
Using the frameworks, we implemented the in-memory SPARQL engine. According to our experiments, the engine achieves significant speed up on SMP and NUMA systems.