Text Mining

1
Top-k Queries in Wireless Sensor Networks Amber
Faucett, afaucett_at_islander.tamucc.edu Dr.
Longzhuang Li, Longzhuang.Li_at_tamucc.edu
Abstract
In todays world, wireless sensor networks are
being used to monitor everything from the
environment to industrial areas. Monitoring top-k
queries is an important aspect needed for
wireless sensor networks especially with
environmental monitoring. From top-k queries,
information can be gathered about the highest or
lowest readings within the sensor network and
from which sensors this data is being logged. By
observing the positive effects that this type of
software could bring, we have decided to look
further into finding the most efficient
algorithms to monitor top-k queries. Because
sensor nodes are limited by a small battery life,
battery consumption must be considered when
designing the query algorithm(s) to identify the
top-k readings. By thoroughly researching
currently proposed solutions for top-k queries,
we will compare the strengths and weaknesses and
hopefully find what is currently the best
algorithm. For future plans, we will look into
constructing a simulation of a wireless sensor
network in order to observe and compare the
numerous top-k monitoring algorithms against one
another as well as our own proposal. The end goal
is to create energy efficient algorithm to
monitor top-k queries in wireless sensor networks
that outperforms any currently proposed solution.
Approach
In order to test the algorithms, we will be using
TinyOS which is an open-source operating system
that can be used to test networking devices
specifically those that require very low power
usage. On top of the TinyOS, we will be using
TOSSIM which is a TinyOS simulator that compiles
directly from TinyOS code. TOSSIM provides the
same tools as would be provided in real-world
networking environment. Along with TOSSIM, we
will be looking at using TinyViz which is a GUI
tool of TOSSIM. TinyViz allows you to interact
with a running simulation of a wireless sensor
network. With TinyViz, we will be able to
visually observe the execution of the TinyOS
applications.
Background
With wireless sensor networks, sometimes it is
desired to find the k objects with the
highest/lowest overall values. Algorithms that
attempt to solve this request have become known
as top-k queries. In order for an algorithm to be
considered successful, it must have low latency
and low battery consumption. There are already
several algorithms out that have been proposed
and tested throughout the world. We will review
over two examples to give a better explanation of
what algorithms consist of for top-k query
monitoring. The first algorithm is what is known
as the Tiny AGgregation or TAG approach. In the
TAG approach, the sensors use a routing tree
rooted at the base station to transfer the data
back to the requesting user. As the data flows
back up through the sensors, it is collected and
divided along the way to the base station
according to the specifications of the query. A
visual representation of the TAG approach can be
seen in Fig a. At each sampling instance, each
node sends its current reading to the next node
which then compares the received data to its own.
With the given example, the TAG approach takes a
total of nine messages to find the top-k
result. A second approach to top-k query
monitoring is called FILA, which is a
filter-based monitoring approach. Each node is
given a filter range so that only data that is
above (or below) that range is returned to the
base station. The base station also keeps a copy
of the filters placed on all of the nodes. The
purpose of applying a filter to the sensors is to
reduce the amount of unnecessary traffic within
the network. A visual representation of FILA can
be seen in Fig b. With this example, FILA takes a
total of four messages to return the top-k result
to the base station.
Future Work
Once familiar with TinyOS and TOSSIM, the plan is
to begin plugging in the discussed algorithms and
viewing their capabilities. From here, we will
choose one of the tested algorithms and look at
ways of improving on its weaknesses. Eventually,
we would like to get to the point of building our
own algorithm that will outperform any currently
proposed algorithms for top-k queries.
References
1 D. Zeinalipour-Yazti, Z. Vagena, D.
Gunopulos and V. Kalogeraki, V. Tsotras, M.
Vlachos, N. Koudas, D. Srivastava, Finding the K
Highest- Ranked Answers in a Distributed
Network, Computer Networks, 25 June 2009. 2
M. Wu, J. Xu, X. Tang, and W.C. Lee, Top-k
Monitoring in Wireless Sensor Networks, IEEE
Trans Knowledge and Data Eng., July 2007. 3 P.
Andeou, D. Zeinalipour-Yazti, M. Vassiliadou,
P.K. Chrysanthis, G. Samaras, KSpot Effectively
Monitoring the K Most Important Events in a
Wireless Sensor Network, In Proceedings of the
25th International Conference on Data Engineering
(ICDE), April 2009. 4 S. Madden, M.J. Franklin,
J.M. Hellerstein, and W. Hong, TAG A Tiny
Aggregation Service for Ad Hoc Sensor Networks,
Proc. Usenix Fifth Symp. Operating Systems Design
and Implementations, Dec. 2002.
Acknowledgments
Presentation of this poster was supported in part
by NSF Grant CNS-0837556. I would also like to
acknowledge the National Science Foundation -
Louis Stokes Alliance for Minority Participation
(grant 0703290) for its support. Thanks also to
Dr. Li for his help and guidance through this
research.