Abstract:
Efficiently searching is the vital operation in any Peer-to-Peer (P2P) network. Users
may not know about the exact keywords of their requested files in such distributed
systems which causes searching using partially specified and misspelled keywords as a
challenging problem in P2P networks.
This thesis presents mPlexus, a P2P search mechanism for efficient information retrieval
in presence of misspelled and partially specified keywords. mPlexus uses the
Double Mclaphone algorithm to match phonetically misspelled query keywords with
the advertised keywords. For achieving bandwidth efficiency and similarity matching,
",Plexus incorporates second order Reed-Muller code within the Plexus [221 protocol
having a logarithmic routing efficiency on overlay network size. mPlexus supports large
scale networks and achieves better resilience to peer failure by keeping redundant routing
paths and by systematically maintaining index replicas. The concept presented in thie
thesis paper is supported by necessary theoretical model and experimental evaluation.
This thesis also presents FM-Chord (Fault-tolerant Chord supporting Misspelled keywords),
a modified Chord [66] protocol for queries containing spelling mistakes and to
ensure the availability of advertised indexes.
Finally, this thesis performs comparative study on search completeness, routing efficiency,
topology maintenance, fault tolerance etc. between mPlexus and FM-Chord as
there is no existing search techniques capable of searching with queries containing partially
specified misspelled keywords.
