A Framework for Workload-Aware Views Materialisation of Semantic Databases

Abstract

Views materialisation is well known in the context of relational databases. However, unlike relational databases, the semantic graph model lacks restrictive structure. Instead, the semantic data rely on an evolving schema. This resulted in a challenge for views materialisation while allowing for open repositories of data to emerge. Open repositories combine knowledge from many different areas. Therefore, it can be assumed that various data structures within a repository may exhibit different daily access patterns, i.e. That the user interests change during the day. This research verifies this assumption and proposes a new views selection model. By analysing how access patterns of individual views contribute to the overall system workload, the proposed model aims at selection of candidates offering the highest reduction of the peak workload. The proposed selection method has been integrated as a part of a new optimisation framework, which operates as a proxy for a SPARQL-enabled database. The approach has a potential to accelerate the adaptation of views materialisation for SPARQL. The proposed approach is evaluated both experimentally and using qualitative analysis.