Information Technology Reference
In-Depth Information
PG5
PG1
PG6
P
P
PG3
P
SRA
P
P
PG2
P
Peer
PG5
PG4
PG: Peer group
SR: Search route
SRB
FIGURE 13.3
Peer groups in PIndex.
also have a minimal amount of random addresses to external peer groups
called Table of External Contacts (TECs) for communication with the rest
of the network. TECs can be self-cleaning and dynamic using the heart-
beat mechanism provided by the index service.
Since each TEC is unique for each node, many routing paths exist which
are dependent on where the search begins. As each search follows a differ-
ent path, three main advantages are apparent. First, intrinsic to an unstruc-
tured P2P algorithm, the reliance on a central server or a hierarchical
structure is eliminated along with their inherent delays and trafi c costs.
Second, search loads are distributed across the network, eliminating the
case where a single node can be overloaded and act as a bottleneck. Finally,
having many search paths also means that if a node fails only a small
number of paths are affected, with the majority of network operating
unhindered. This indirectly creates a virtual overlay for each node as
depicted in Figure 13.3. From Figure 13.3, search route A differs from
search route B because in route A, a node in PG1 has contacts existing in
PG2 and PG6 which then extend to PG3, PG4, and PG5 through PG2 and
PG4. However, in route B a different node in PG1 has contacts existing in
PG6 and PG3, which respectively have contacts to PG5, PG4, and PG2.
Depending on which node in PG1 a search starts, a route will be dynami-
cally determined and taken.
13.3.2
Querying Resources in PIndex
Besides a robust, dynamic, and well-balanced architecture, PIndex needs
a fast yet efi cient search algorithm to query resources. PIndex localizes
Search WWH ::
Custom Search