SaGe: Preemptive Query Execution for High Data Availability on the Web

Abstract: Semantic Web applications require querying available RDF Data with high performance and reliability. However, ensuring both data availability and performant SPARQL query execution in the context of public SPARQL servers are challenging problems. Queries could have arbitrary execution time and unknown arrival rates. In this paper, we propose SaGe, a preemptive server-side SPARQL query engine. SaGe relies on a preemptable physical query execution plan and preemptable physical operators. SaGe stops query execution after a given slice of time, saves the state of the plan and sends the saved plan back to the client with retrieved results. Later, the client can continue the query execution by resubmitting the saved plan to the server. By ensuring a fair query execution, SaGe maintains server availability and provides high query throughput. Experimental results demonstrate that SaGe outperforms the state of the art SPARQL query engines in terms of query throughput, query timeout and answer completeness.