Querying and Reasoning of Graph-Like Data in Centralised and Distributed Systems
Department of Computer Science, University of Oxford
Oracle Fellowship Recipient
Oracle Principal Investigator
Sungpack Hong, Director, Research And Advanced Development
Zhe (Alan) Wu
Processing graph-like data has recently received a lot of attention in both academia and industry. RDF and property graphs are prominent examples of graph-like data models. A number of systems developed in academic institutions and by major software vendors support a range of tasks over these models, such as evaluation of queries and reasoning with respect to background knowledge (i.e., the derivation of facts that logically follow from the knowledge). Many of such systems are centralised—that is, they store and process all data on one node. A common approach to increasing the capacity of such systems involves distributing the data in a cluster of shared-nothing nodes and using suitable distributed processing techniques; in such cases, it is critical to minimise the communication between the nodes. The goal of this project is to further develop and evaluate techniques for distributed querying and reasoning over graph-like data. In particular, this project builds on the research conducted at the University of Oxford in parallelisation of reasoning in centralised systems and on distributed query evaluation.
The goals of the project are to (i) apply the reasoning techniques developed for centralised, multicore RDF systems and showcased in the RDFox system to reasoning with property graphs; (ii) support parallelisation of query answering using the same principles as for reasoning; (iii) develop distributed query evaluation techniques necessary to support languages for graph-like models such as PGQL and SPARQL; and (iv) combine the distributed query evaluation and centralised reasoning algorithms to support efficient reasoning with graph-like data in a distributed setting.