Database Reference
In-Depth Information
(a) 2nd Query Performance
(b) 2nd Query CPU Load
Fig. 4.
2nd Experiment results for the system's standalone configuration
Similarly to the above case, Figs.
4
(a), (b) show the results of the second
experiment for the standalone configuration, where Fig.
4
(a) shows the progress
in average query time for every 50 requests for the second query, while Fig.
4
(b)
shows the progress in average CPU time per 10 s for the same query. It should
be noted here that based on the results of the 1st experiment, it was decided
that it is pointless in further checking the performance of the old system as it
is expected that this performance will be even worse for a query with a higher
complexity.
The depicted results indicate that the second query is indeed more complex
and leads to even worse results than the first, something that is expected based
on the nature of the second query and its characteristics. Concerning the average
query time, the behaviour of Virtuoso is similar (while the query time increase
pace is even smaller) to the first experiment case but the query times anticipated
are much worse ranging between 274 and 275 s (i.e., they are 9 times greater)
which is quite unacceptable for users especially in real world environments. We
should highlight at this point that although Virtuoso has a hard time in servicing
the requests, its behaviour is almost stable, which means that the system exhibits
a good reliability level.
As far as the CPU is concerned, the behaviour is again similar with respect
to the first experiment where we see a sharp increase in the beginning and then
an increase with a very small pace until a specific point (around 99 %). However,
the peak point is higher than that of the first query. This means that the VM
has a harder time than in the case of the first experiment and in both cases it
reaches its limits in terms of CPU usage. These two facts, i.e., sharper increase
and higher CPU peak values, also highlight the complexity difference between
the queries considered. Based on these results, there is a great need of scaling the
Search WWH ::
Custom Search