Information Technology Reference
In-Depth Information
Fig. 3. Query execution time of FedX and its LAWS extension
the LAWS extension, the precision of source selection increases by a percentage
of 75.2, 55.04 and 27.34 respective for DARQ, FedX and SPLENDID.
We also measured the average query execution time in each of the federated
query approaches and also in their LAWS extension. Figures 1, 2 and 3 show the
results. We can see that LAWS improves the query performance for most of the
cases. For twelve of the benchmark queries, LAWS improved the query execution
times of all federated systems tested. The query performance for queries CD7
and LS6 showed the highest improvements. This is due to the large number of
source selection plans that were pruned.
6.2 As an Independent Query Engine
For a widely experimental study of technologies presented in this paper, we
have implemented LAW as a federated query engine, and, again, compare it
with DARQ, SPLENDID and FedX using another Benchmark. For our tests we
adopt the Berlin SPARQL Benchmark (BSBM) [3]. The BSBM executes a mix
of 12 SPARQL queries over generated sets of RDF data; the datasets are scalable
to different sizes based on a scaling factor. The hardware environments used by
BSBM is the same as the one where FedBench is deployed.
Experimental Results. We execute the query mix for datasets generated with
scaling factors of 10 to 60; these datasets have sizes of 4,971 to 26,108 triples,
respectively. For each dataset we run the query mix 6 times where the first run
is for warm up and is not considered for the measures.
Figure 4 depicts the average times to execute the query mix by LAW and
the other three federated query engines: DARQ, SPLENDID and FedX. As can
be seen from the measures, LAW is better than other query engines for all mix
queries with respect to execution times.
Table 3 summarizes the average number of requests sent to the data sources
during query evaluation in the SPARQL federation. These numbers immediately
explain the improvements in query performance of LAW. With our optimization
techniques, LAW is able to minimize the number of selection plans necessary to
process the queries.
Search WWH ::




Custom Search