An Ontology-Based P2P Network for Semantic Search - Cloud, Grid and High Performance Computing: Emerging Applications - page 309

Figure 5. Average search path length vs. number of nodes for the various numbers of long range contacts

tic clusters in the upper-tier from 2 4 to 2 11 . Since
the number of nodes in each semantic cluster is
fixed in this experiment, the average search path
length in the lower-tier is a constant. Figure 6
plots search path length vs. number of semantic
clusters in our system in the various settings of
numbers of long range contacts. The result shows
that increasing the number of long range contacts
reduces search path length significantly. Figure

6 also reveals that search path length in the upper-
tier matches the small world phenomenon.
We compare the cost of node joining and leav-
ing between our system and Chord in this ex-
periment. We vary network size from 2 5 to 2 14 . In
reality, the number of semantic clusters may in-
crease when the network size increases. To simu-
late this behavior, we increase the number of
semantic clusters with proportional to cccccccccc
by making the number of semantic clusters equal

Figure 6. Average search path length vs. number of semantic clusters in the various settings of numbers
of long range contacts