The CoQUOS Approach to Continuous Queries in Unstructured Overlays



  The CoQUOS Approach to Continuous.pdf (Size: 450.59 KB / Downloads: 2)

INTRODUCTION

UNSTRUCTURED peer-to-peer (P2P)-based
content/resource sharing platforms such as
Gnutella [1] and Kazaa [2] have experienced tremendous
growths in the past decade. The popularity of
unstructured P2P networks can be attributed to
the simplicity of their designs and their exibility
towards transient node population. Searching in these
networks is essentially performed by circulating query
messages in an ad-hoc manner and probing individual
peer nodes.


Paper Contributions


The difculties in implementing pub-sub systems on top
of unstructured overlays can be attributed to the inherent
mismatch between the design requirements of pub-sub
systems and the very nature of unstructured P2P systems.
In order to provide strict notication guarantee,
pub-sub systems have to maintain subscription information
in well-organized structures. In contrast, unstructured
overlays, by their very nature, are decentralized,
loosely coupled and, to certain extent, haphazard. This
makes it difcult to build full-edged pub-sub systems
on these platforms.


THE COQUOS SYSTEM

In this section, we present a high-level description of the
CoQUOS system architecture. We begin by introducing
a few concepts that are used in the rest of the paper.


Design Overview
Our goal in designing the CoQUOS system is to design
a highly effective notication service on top of arbitrary
unstructured overlay networks. A common way of quantifying
notication effectiveness is through the overall
notication success rate of the system. However, our
design of the CoQUOS middleware strives for a stronger
notion of notication effectiveness, wherein the goal is
not only to achieve high overall success rate, but also
reasonably high success rates for each individual query.



SELECTING BEACON NODES OF A QUERY

The discussion in the previous section highlights the
crucial role played by the beacon nodes in notifying the
source peer of matching data items. Hence, the choice
of beacon nodes would have a signicant impact on the
notication success rates of a continuous query. So, an
important research question is: how do we select the set of
peers that will serve as the beacon nodes of a query?. In other
words, which set of peers should host a particular continuous
query in order to realize high notication effectiveness?



ENHANCEMENTS AND DISCUSSION
The loosely-coupled and highly dynamic nature of underlying
P2P network poses several additional challenges.
In this section, we discuss two issues that are
of particular importance to the performance of the Co-
QUOS system, namely (a) Churn of the P2P overlay;
and (b) Load distribution among peers. We also present a
brief discussion on the implementational aspects of the
CoQUOS system.