25-08-2017, 09:32 PM
Enhancing and Analyzing Search performance in Unstructured Peer to Peer Networks
1Enhancing and Analyzing.pdf (Size: 404.62 KB / Downloads: 32)
Abstract
Peer-to-peer (P2P) networks establish loosely coupled application-level overlays on top of the Internet to facilitate efficient
sharing of resources. It can be roughly classified as either structured or unstructured networks. Without stringent constraints over
the network topology, unstructured P2P networks can be constructed very efficiently and are therefore considered suitable to the
Internet environment. However, the random search strategies adopted by these networks usually perform poorly with a large
network size. To enhance the search performance in unstructured P2P networks through exploiting users’ common interest
patterns captured within a probability-theoretic framework termed the user interest model (UIM).
Introduction
Peer-to-peer (P2P) networks have become, in a short period
of time, one of the fastest growing and most popular Internet
applications [6]. A class of applications that takes advantage
of resources like storage, CPU cycles, content and even
human presence available at the edges of the Internet.One
fundamental challenge of Peer to Peer networks is to achieve
efficient resources discovery. Those networks can be largely
classified into two categories, namely, structured P2P
networks based on a distributed hash table (DHT)[21] and
unstructured P2P networks based on diverse random search
strategies (e.g., flooding)[3]. Without imposing any stringent
constraints over the network topology, unstructured P2P
networks can be constructed very efficiently and have
therefore attracted far more practical use in the Internet [1],
[2] than the structured networks. Peers in unstructured
networks are often termed blind, since they are usually
incapable of determining the possibility that their neighbour
peers can satisfy any resource queries.
Peer to Peer Network
In recent years, Peer-to-Peer (P2P) technologies have
become increasingly popular. A P2P system can be defined
as a distributed network architecture, whereby participants
share a part of their own hardware resources, such as
processing power, storage capacity, or network
bandwidth. The shared resources are necessary to provide
the service and content offered by the network, such as
file-sharing. The service or content provided by the P2P
network is accessible by other peers directly, without
passing intermediary entities. Peer-to-Peer (P2P) systems
make it possible to harness resources such as the storage,
bandwidth, and computing power of large populations of
networked computers in a cost-effective manner. Actually
P2P is a decentralized and distributed and here all the nodes
are equivalent.
User Common Interest Model (UCIM)
This section deals with one essential problem as to how
user’s interests can be modeled properly. Our UIM aims at
characterizing users’ common interest patterns within a
probability-theoretic framework. It is adapted from a general
probabilistic modeling tool termed Conditional Random
Fields (CRFs)[14].Similar with CRF; UIM defines a loglinear
conditional probability distribution Pr (fj\fi) between
any two files fi and fj
In this paper, Pr (fj\fi) refers to the probability that any user
can be interested in sharing file fj, given the fact that he/she
shares another file fi.Probabilistic inference in UIM is very
efficient, without relying on any independence assumptions
as required by other probabilistic modeling techniques such
as the Hidden Markov Model (HMM)[22].
Learning Methods
The primary concern of this paper is to manage network
topology and to enhance resource discovery performance
with the help of UIM. However, to make our discussion
complete, this section will briefly introduce the process
through which UIM can be learned and updated. The basic
design principles of peer to peer networks, UIMs are better
to be learned locally by every peer. However, in order to
ensure that these locally maintained UIMs will remain
consistent with each other, designated servers will also be
employed to fulfil certain computation intensive learning
tasks.
Our Filtering mechanism
Specifically, network users might only query for those files
directly related to their local interests. In other words, the
peer p0 that is able to satisfy any query from another peer p
normally is close in distance to p. Hence, newly added peers
for updating a routing table are not evenly distributed across
the full distance range. This uneven distribution will
possibly render our routing table updating protocol
ineffective. We call this problem uneven updating problem.
In general, it is hard to appropriately model the probabilistic
distribution over newly added neighbour peers, which might
also change across time. Instead of modifying our protocol
in a filtering mechanism will be presented in this section to
further control the routing table updating process.The
purpose of the filtering mechanism is to enforce so that our
protocol can still remain effective.