Analysis and Improvement of Optimizer for Query Processing on Graph Store

RDF systems are widely used to store public knowledge bases and process SPARQL queries. A large number of such systems have been proposed in the recent literature to provide low latency and high throughput for concurrent query processing over large RDF data. We perform an in-depth analysis on three key components (cardinality estimation, cost model, and plan enumeration) of the query optimizer to reveal the main issues and challenges on the accuracy and performance for traditional approaches. This calls for a rethink of how to build an accurate and fast query optimizer for modern RDF systems. We introduce a type-centric approach to enhance the accuracy of cardinality estimation prominently, which naturally embeds the lineage of correlated query conditions (triple patterns) into existing type system of RDF data. The preliminary results show that our approach greatly improves the accuracy of query optimization by several orders of magnitude compared to state-of-the-art approaches and provides a better overall performance by reducing execution time or optimization time.