Information Technology Reference
In-Depth Information
5 . 3
P e r f o r m a n c e E v a l u a t i o n
To evaluate the performance of the proposed algorithms, the scope of the simu-
lator discussed in Section
4
was extended to measure the average response time per
object retrieval. To measure the effectiveness of the algorithms across a more unbi-
ased testbed, the degree of connectivity among the objects in the DAG was randomly
varied, and 100 different DAGs were generated. In every DAG, the out-degrees of
the nodes were determined within the range between 0 and 3. The weight of each
node was categorized as strong, normal, and weak.
5.3.1
Number of Air Channels
Figure 20
shows the effect of varying the number of air channels on the average
response time per object. As anticipated, increasing the number of channels resulted
in a better response time for both the LOF and CCP. However, this improvement
flattened as the number of channels increased above a certain threshold, because
additional parallelism provided by the additional number of channels did not match
the number of free nodes available to be allocated, simultaneously. In addition, the
CCP method outperformed the LOF method—the CCP heuristics attempt to smooth
the distribution of the objects among the air channels while clustering the related
objects.
F
IG
. 20. Average response time vs. number of channels.
