or querying over many independent, distributed SPARQL-capable endpoints Cosmin Basca, Abraham Bernstein {lastname}@ifi.uzh.ch http://www.ifi.uzh.ch/ddis avalanche 3] Query Planning and Execution phase AVALANCHE endpoints Search Engine i.e., http://void.rkbexplorer.com/ 1] Source Discovery phase Plans Queue Plan Generator Finished Plans Queue Results Queue Query Stopper Executor Materializer Executor Executor Executor Materializer Materializer Materializer Results Statistics Requester Query Query Parser 2] Statistics Gathering phase Source Selector 0.001 0.01 0.1 1 10 100 1000 10000 FQ0 FQ1 FQ2 FQ3 FQ4 FQ5 FQ6 FQ7 Time (seconds) Cross Domain first (default) total (default) first (cache) total (cache) first (filter) total (filter) first (bloom) total (bloom) 0.001 0.01 0.1 1 10 100 1000 10000 FQ8 FQ9 FQ10 FQ11 FQ13 FQ14 FQ15 Time (seconds) Life Sciences 0.001 0.01 0.1 1 10 100 1000 10000 FQ16 FQ17 FQ18 FQ19 FQ20 FQ21 FQ22 FQ23 FQ24 FQ25 FQ26 Time (seconds) Linked Data 0.001 0.01 0.1 1 10 100 1000 10000 FQ27 FQ29 FQ30 FQ31 FQ32 Time (seconds) Fedbench Queries SP 2 B Query Search Engine i.e., http://void.rkbexplorer.com/ Avalanche SPARQL endpoint 2) get triple pattern cardinalities 3) execute distributed join 1) get sources 0 5 10 15 20 Time (seconds) distribution D1 first (default) total (default) first (cache) total (cache) first (filter) total (filter) first (bloom) total (bloom) 0 5 10 15 20 Time (seconds) distribution D2 0 5 10 15 20 Time (seconds) distribution D3 0 5 10 15 20 Q1 Q2 Q3 Q4 Q5 Q6 Q7 Q8 Q9 Q10 Q11 Time (seconds) Queries distribution D4