Fig. 4

Query Scaling. A summary of query performance. Our test infrastructure, per dataset, generates a set of nine queries that cover the three query modalities (a single clause, multiple clauses, and timestamp ranges). For the nine queries, the template is fixed, but the generator chooses constraint values randomly. For each dataset, we generated five of these nine query benchmarks, and averaged the observed query times across the five samples. For this plot we have shown four of the nine query types. Examples of the four queries displayed are as follows: “QUERY user=5” clause(user), “QUERY resource=resD activity=activityE” conjunction,“QUERY timestamprange=[99171676,102561181]” timestamp, and “QUERY user=6 timestamprange=[32226847,82574461] sortby=Ref-ID” all three(sorted)