On the Efficiency of Peer-to-Peer Keyword Searching

Rahul Dhar and Boris Jabes Information Networking Institute Carnegie
Mellon University Pittsburgh, PA

{rdhar, bjabes}@cmu.edu

ABSTRACT

Peer-to-peer systems have several advantages over their centralized
counterparts. Foremost among these are the freedom for nodes to join
and leave the network as well as very loose resource coordination
requirements. One of the key factors concerning scalability in
large-scale peer-to-peer systems is efficient
searching. Traditionally, there is no efficient way of searching for
content in these networks, reducing their utility as viable methods
for content distribution. Most peer-to-peer systems today rely on
broadcast mechanisms that scale very poorly. Patrick Reynolds and Amin
Vahdat have proposed a design for an efficient way to perform keyword
searching in distributed hash table based networks. We briefly detail
their search engine design and describe our implementation. We then
evaluate our system according to a number of benchmarks and compare
them to the simulated findings of Reynolds and Vahdat.