Processing (Multiple) Spatio-temporal Range Queries in Multicore Settings

Research Moving Objects Databases (MOD) has addressed various aspects of storing and querying trajectories of moving objects: from modelling, through linguistic constructs and formalisms/algebras, to indexing structures and efficient processing of different query-categories have been subjects to a large body of works. Given the architectural trends of multicore CPUs becoming a commonplace, in this work we focus on efficient processing of spatio-temporal range queries in such settings. We postulate that coupling the semantics of the problem domain into the query processing algorithms in a manner that is aware of the multicore features, can yield performance improvements that surpass the gains obtained by relying solely on the compiler-generated threads parallelization. Towards that end, we present and evaluate heuristics for processing variants spatio-temporal range queries in multicore settings by partitioning the load (i.e., data set) and assigning partial tasks to the individual cores. Our experiments demonstrate that 5-fold speed-ups can be achieved, when compared to the (semi) naive approach which relies on the compiler to generate the multicore-compatible code.