Several Oxford PhD students are actively engaged in SIRIUS research projects, including Alessandro Ronca, who is working on Incremental Reasoning for Continuous Queries, and Anthony Potter, who is working on Materialisation- based Query Answering. Alessandro’s work is still at a relatively early stage (he only started his PhD in October 2015, and since then has had to attend a number of taught courses), but Anthony’s work is rapidly maturing, and since returning from a summer internship at Oracle (in their Redwood Shores campus in California) he has produced several important results.
First, he has devised a novel data partitioning algorithm that tries to maximise the “semantic locality” of data by using an approach based on a graph partitioning, where the algorithm tries to create a balanced partitioning while minimising the number of cut edges in the graph.
Second, he has developed a novel query processing technique that exploits the semantic locality of the partition so as to maximise local computation and minimise network communi- cation. The query algorithm runs in a completely distributed manner, with any node able to distribute client queries across the cluster, with no “master” node responsible for accumulating results, and with a novel mechanism for detecting when all answers have been computed. The algorithm also uses a novel approach to message passing and memory management to ensure low and predictable memory usage during query processing, and to avoid non-termination due to full buffers or other memory related issues.
Finally, all of the above ideas have been implemented in a prototype system based on RDFox, and is being evaluated against several other state of the art systems. At the same time, Anthony is writing a paper describing all of the above work, and we plan to submit this to the International Semantic Web Conference.