Database Reference
In-Depth Information
(a) 1st Query Performance
(b) 1st Query CPU Load
Fig. 3. 1st Experiment results for the system's standalone configuration
beginning which drop and become stabilized at some time point. In this way, it
is apparent that the new LMS system is better than the old one and more cost
effective as it uses just one instance instead of two. The query time decrease in
the old system is due to the distribution of the respective data in two instances,
thus requiring less processing per instance to be performed for each user query
request. We should also indicate that the initial high query time observed in
both system versions is due to a main memory effect where Virtuoso realizes
that its currently allocated main memory is not sucient and attempts to gain
more resources, thus further delaying the servicing of the current requests. After
the main memory allocated reaches a particular value, then the query addressing
gets improved which subsequently leads to a good reduction in query time.
On the other hand, the CPU load follows a clear course on all system versions
where it sharply increases in the beginning and then increases in a very small
pace until a specific point (97.5 % and 90 % for the new and old LMS system,
respectively). This means that throughout the time where all requests are made,
the VMs/instances reach quite high numbers of usage. So, if the load was shared
between two or more VMs, then obviously the query service time would decrease
more sharply. Indeed, by lowering the number of concurrent users from 100 to 50,
then the CPU load gets decreased and the respective query time becomes quite
small. Thus, the results of the first query indicate that there is an opportunity
here for obtaining more resources in order to better serve the incoming requests
to the system.
We need to note here that the CPU load on the old system version is lower
than that imposed on the new one due to the rationale provided above for query
response time: less (related) data are stored in each VM which leads to a smaller
amount of processing.
Search WWH ::




Custom Search