A Resilient Index Graph for Querying Large Biological Scientific Data

The biological scientific linked data is large graph that contains billions of triples representing links between massive microorganisms. Now it is challenged by the growing graph size and the costly queries that require massive traversals. This work designs a resilient index graph to model the query pattern. According to given query pattern, it indexes graph traversals between the starting and objective vertices and represents them as index edges. Through visiting index edges, the query can be completed in one hop without repeatedly traversing the graph. It can bound the query to limited traversals and thus could response in real time. Moreover, the index graph can be constructed using BSP computing model with constant rounds. We developed a prototype system based on Titan, and experimental results showed that the index graph can complete complex biological benchmark queries within 400 milliseconds on average.