CumulusRDF: Linked Data Management on Nested Key-Value Stores

Abstract

Publishers of Linked Data require scalable storage and retrieval infras- tructure due to the size of datasets and potentially high rate of lookups on popular sites. In this paper we investigate the feasibility of using a distributed nested key- value store as an underlying storage component for a Linked Data server which provides functionality for serving Linked Data via HTTP lookups and in addi- tion offers single triple pattern lookups. We devise two storage schemes for our CumulusRDF system implemented on Apache Cassandra, an open-source nested key-value store. We compare the schemes on a subset of DBpedia and both syn- thetic workloads and workloads obtained from DBpedia's access logs. Results on a cluster of up to 8 machines indicate that CumulusRDF is competitive to state-of-the-art distributed RDF stores.