The Center for Pervasive Computing and Networking at
Rensselaer
A
multidisciplinary group of researchers from the Schools of Engineering and
Science has come together in Rensselaer’s Center for Pervasive Computing and
Networking to collaborate on projects that contribute to the goal of
pervasive computing. This vision foresees a world in the not-distant future
in which computer systems are embedded in everything: from personal digital
assistants to implanted biological devices, to bridge-monitoring systems,
and to
teams of robots sent into a collapsed building to locate survivors. Untethered – wireless – communication is constant and, in many cases, so
automated that human intervention is unneeded. Wireless, broadband community
systems inexpensively bring people together for virtual town meetings, video
doctor-patient conferences, and on-line business transactions. Computers in
automobiles share information on congestion, quickly computing alternate
routes. The promises are immense, but the challenges are formidable.
Grids and Worldwide Computing
As workstations and desktop computers gain power and increasing numbers are
connected to the Internet 24 hours a day, a movement has arisen to create
both formal and ad hoc networks in which users combine their computing
power, utilizing idle time on machines ranging from individual desktops and
PCs to clusters of PCs to supercomputers to form parallel processors capable
of tackling very large problems. To achieve these goals, advances are needed
in many areas, including programming and protocols for parallel processing,
tracking and accessing widely distributed pieces of data, and routing
messages over a constantly changing and sometimes unreliable network.
Although grid computing has made a lot of progress in recent years through
projects such as Globus, the focus in Rensselaer’s Center is on more dynamic
and autonomous environments in which task allocation, migration, and fault
tolerance are supported automatically.
Boleslaw Szymanski, professor of computer science and founding
director of the Center for Pervasive Computing and Networking, is developing
high-level services and protocols that boost the performance of grid
computing, offering high data availability, low bandwidth consumption,
increased fault tolerance, and improved scalability.
Carlos Varela,
assistant professor of computer science, is developing SALSA (Simple Actor
Language, System, and Architecture), a programming language and run-time
platform for building dynamically reconfigurable distributed systems based
on the actor model. He is working with IBM on a "transactor" model for
reliable e-business on the Web, and he is also involved in projects that use
worldwide computing for scientific applications.
Alhussein Abouzeid,
assistant professor of electrical, computer, and systems engineering (ECSE),
is developing performance modeling techniques and scalable methods of
routing messages on the very large-scale, variable topology networks formed
in grid computing.
Bulent Yener, associate professor of computer
science, suggests mastering the complex and constantly changing topology of
grid networks by creating a virtual mesh embedded on top of the arbitrary
network mesh.
David Musser, professor of computer science, who is
known for his work in generic software libraries, is extending those
techniques to create generic libraries and high-level optimization for
embedded computing systems.
Chris Carothers, assistant professor of
computer science, who has developed very fast and efficient parallel
computing methods, is applying those methods to laser eye surgery.
Security Issues in Computers, Networks, and Sensors
Without serious
attention to security issues, the world of pervasive computing could turn
rapidly from dream to nightmare, as on-line criminals and terrorists steal
private information, destructively attack individual computers and entire
networks, and send damaged and dangerous programs to unprotected systems. At
Rensselaer,
Vera Kettnaker, assistant professor of computer science,
is developing a time-sensitive video surveillance system to monitor
high-security rooms to detect intruders or suspicious employee behavior.
Mohammed Zaki, assistant professor of computer science, has developed
ADMIT, a data-mining system that detects clusters of unusual behavior by a
computer user.
Boleslaw Szymanski uses probabilistic state finite
automata, mathematical models of computations augmented with probabilities,
to detect variations from the user’s normal behavior. He also has developed
a system to use bioinformatics techniques, normally used to match DNA
sequences, to match command sequences with a user’s normal behavior, and he
uses a technique known as the Conceptor in an application known as COMMAND
to create concepts of a user’s typical behavior and warn of variations. To
detect network attacks, Dr. Szymanski is adapting his DOORS (Distributed
Object Oriented Repository Simulation) program, a network management tool,
to collect information that can be used to recognize a denial of service
attack.
Dr. Musser is exploring the use of generic code-carrying
proofs as a secure and memory-stingy method of sending programming code.
Dr. Yener is working on security for an ad hoc wireless system, in which
an on-line controller would make intelligent decisions about security levels
as battery power diminishes.
Biplab Sikdar, ECSE assistant professor,
is identifying security gaps and designing protection against specific
attacks in the Border Gateway Protocol, which regulates traffic as it moves
from network to network.
Network Modeling, Simulation, and Management
Rensselaer
researchers have developed very efficient methods to run simulations to
detect problems on computer networks and then to apply traffic management
techniques to solve the problems. Their goals are to reduce congestion,
automate many management tasks, and improve quality of service.
Dr.
Carothers created ROSSNet (Rensselaer’s Optimistic Simulation System), a
very fast simulation method, and works with
Shivkumar Kalyanaraman,
ECSE associate professor, to use these simulations to optimize very complex
systems.
Dr. Kalyanaraman, widely recognized for his work in traffic
management, also uses a “recursive random search algorithm” to continuously
improve network performance by adjusting parameters to changing conditions.
He is developing BANANAS, an Internet architectural framework, that gives
messages more flexibility in the routes they choose, and he is working on
overlay systems that can deliver very reliable broadband services to groups
of users.
Dr. Szymanski and his team developed Genesis (The General
Network Simulation Integration System), which divides a large network or
even the entire Internet into domains and runs a simulation of each over a
given time interval on a separate processor. The processors then exchange
information and run new simulations for the time interval until they
converge on a solution.
Wireless Networks
Unlike cell phone
systems, in which messages travel by way of fixed towers, devices in ad hoc
wireless systems communicate directly with each other. They pass messages
from node to node as needed, even as some devices move around and others
unpredictably come on- or off-line, creating a constant need to find new
routes for messages. Rensselaer researchers are working on all levels of the
technology to make such networks efficient and reliable.
Dr. Kalyanaraman
and
Partha Dutta, also an ECSE assistant professor, are using
microelectronics techniques to create a multihop optical wireless system.
Dr. Kalyanaraman also is using both radio frequency (RF) and optical
techniques to build an inexpensive and easily accessible community network
around the Rensselaer campus. ECSE Professor
Kenneth Vastola has
worked on ad hoc wireless technology that can warn when a node is about to
fail, while
Alhoussein Abouzeid is looking at ways to optimize
computation and communication costs by managing message compression.
Costas Busch, assistant professor of computer science, studies packet
routing problems, and with
Dr. Yener, is considering ways to
synchronize nodes to minimize packet collisions. Rensselaer is also working
on distributed networks of sensors and actuators.
Richard Radke, ECSE
assistant professor, is developing methods through which groups of cameras
can exchange information and work together, while
Wesley Huang,
assistant professor of computer science, and
Jeff Trinkle, professor
and chair of computer science, are developing techniques for distributed
groups of robots to communicate and cooperate. Rensselaer’s new Satellite
Center on the Hudson River is developing an array of monitoring,
communication, and visualization tools to analyze the River.