Query Planning in DHT Based RDF Stores

Implementing scalable RDF triple stores that can store many triples and process many queries concurrently is challenging. Several projects have investigated the use of distributed hash tables for this task but query planning has received little attention in this context so far. Given the distributed nature of DHTs, latencies of messages and limited network bandwidth are crucial factors to consider. Also due to a lack of global knowledge in DHTs, query planning is different from centralized databases. This paper discusses a set of heuristics and evaluates their performance on the Lehigh University Benchmark with emphasis on the network traffic. The results show the importance of query planning in DHT based RDF triple stores.