Distributed Scalable RDFS Reasoning

A number of reasoning studies on big ontology have been carried out in the recent years. However, most of the existing studies have focused heavily on Hadoop MapReduce. In this paper, we propose a reasoning approach for Resource Description Framework Schema (RDFS) that employs optimized methods based on Spark. Spark is a general distributed in-memory framework for large-scale data processing that is not tied to the two-stage MapReduce paradigm. In our work, we devised an extensive optimization method to cope with the communication bottleneck of data shuffling between machine nodes in a distributed system. From empirical evaluations, the proposed reasoning system produces at most the throughput of 4166KT/sec which is almost 80% faster than the MapReduce based reasoner WebPIE.