How Interlinks Influence Federated over SPARQL Endpoints

Abstract

As the Web of Data grows, the number of available SPARQL endpoints increases. SPARQL endpoints conceptually represent RPC-style, coarse-grained data access mechanisms. Nevertheless, through the potential interlinking of the contained entities, SPARQL endpoints should be able to over distinct advantages over plain Web APIs. To our knowledge, to date, there has been no study conducted that gauges the impact of the link on SPARQL query execution, especially in a federated set-up. In this paper, we investigate how the existence and types of typed links influences the execution characteristics of different SPARQL federation frameworks. In order to measure the query performance, we propose a combined cost model based on a statistic analysis of the query performance metrics that involves parameters such as type of link, the data catalogues and cache, number of links, and number of distinct subjects. As result, we show that number of distinct subject and number of links have significant impact on Federation over SPARQL Endpoints performance whereas type of link does not have significantly influence in federation query performance.