Santashil PalChaudhuri
Rice University
My research interests lie in the area of networked computer systems.
I plan to advance the state-of-the-art of networked computing through
the contribution of original ideas
followed by translation of these ideas to end applications and
products.
I have decided to start my career in the industry to realize
this career objective.
Wireless networks are on the verge of massive proliferation. The emergence of protocols like ZigBee, UWB, WiFi and WiMax have led to significant growth and innovation in the multi-hop sensor and mesh networks. Sensor networks enable pro-active computing by sensing the parameters of the physical world. There exists wide scope for enhancement of productivity and efficiency in such a networked environment. Tuning the network for the characteristics of specific application requirements widens this scope further. Mesh networks have the potential to provide ubiquitous connectivity and wireless peer-to-peer networking. One of the most challenging tasks is to realize this potential in a cost effective manner. In my future research, I want to develop networking protocols in these domains. My background and expertise in wireless, ad hoc, and sensor networks, summarized next, will enable me to help drive the wireless revolution. I have also enjoyed working in inter-disciplinary teams at Rice as part of several NSF funded projects.
Design of an adaptive sensor network architecture for multi-scale
communication and collaboration is the focus of my PhD thesis, as part
of the COMPASS project.
I have designed adaptive protocols for routing,
medium access scheduling and synchronization for the proposed architecture.
As a leading member of the team, I have also worked in the project
named PRAN -- a system for enabling physical
implementations of simulation models.
This was an unique experience
that reinforced in me the confidence
of building real life systems. My research experience
on scalable ad hoc networking
systems and on network edge architectures have provided me with a
broad background in the
design of wired and wireless systems.
Sensor Network Architecture
Sensor networks have emerged as a promising tool for monitoring and
actuating the physical world. It employs self-organizing
networks of battery-powered wireless sensors that can sense, process
and communicate. In sensor networks, energy is a critical resource,
while intended applications exhibit a few specific characteristics.
Consequently,
there is both a need and an opportunity to optimize the network
architecture for the specific applications in order to minimize resource
utilization. The requirements and limitations of sensor networks make
their architecture and protocols both challenging and divergent.
Many applications, such as large-scale collaborative sensing, distributed signal processing, and distributed data assimilation require sensor data to be available at multiple resolutions, or to allow fidelity to be traded-off for energy efficiency. I proposed the design of an adaptive cross-layered Sensor Network Architecture for enabling multi-scale collaboration and communication. This architecture enables scalability, localization and resolution-tuning, while simplifying application design by providing communication abstractions. In collaboration with a colleague, I characterized the unique design requirements for sensor network architecture and proposed SensorStack [6] as a suitable architecture for sensor networks. SensorStack enables cross-layering using a notification service, adaptability of the protocols to application-specific needs and an abstraction for data-centric communication. Next I have designed routing, scheduling and synchronization protocols to take advantage of the cross-layering and adaptivity supported by SensorStack.
I proposed a routing protocol [4],
which is a hierarchical overlay to handle aggregation, dissemination,
and multiple resolution. This self-organizing network hierarchy adapts to align
with the data communication for increased efficiency. To simplify
application design, I provide a set of Network Programming Interfaces
to abstract the details of low-level communication and implement these
interfaces efficiently in the network.
For this multi-scale architecture, I have also proposed a medium access scheduling
protocol [3].
This protocol takes advantage of sensor network application
characteristics such as --
periodic nature of communication, limited communication abstractions,
and fusion function techniques -- to improve energy-efficiency.
The scheduling uses a token-passing approach to provide collision-free
neighborhoods for apriori known traffic, as well as provides
contention-based access period for event-driven traffic.
My next contribution is the clock synchronization
protocol [5] which is again made adaptive
for specific application needs at any time.
I have proposed a scheme to
convert service specifications (maximum clock
synchronization error and confidence probability) to actual protocol
parameters (number of messages and synchronization
interval).
Physical Realization of Protocols
Simulation and physical realization are both valuable tools in
evaluating ad hoc networking protocols, but neither alone is
sufficient.
I have co-designed a new system
named PRAN (meaning ``life'' in Sanskrit)
for implementation of ad hoc routing protocols that merges these two
evaluation tools.
PRAN allows existing simulation models to be used -- without
modification -- to create a physical implementation of the same
protocol. We evaluated the simplicity and portability of this approach
across multiple protocols and operating systems through example
implementations of the DSR and AODV routing protocols in FreeBSD and
Linux using the existing, unmodified ns-2 simulation
models [8]. We illustrated the ability of the
resulting protocol implementations by transmitting real-time video
over a multi-hop mobile ad hoc network; a demonstration at
MobiCom 2004 featured mobile robots remotely operated based on
the video stream over the network from cameras attached to the robots.
We also reported a detailed performance evaluation of PRAN to
establish the feasibility of our proposed architecture [7].
Scalable Ad Hoc Networking
As devices with wireless networking become more pervasive, mobile ad hoc
networks are becoming increasingly important.
A prerequisite for such an environment is the design of
scalable ad hoc networking techniques,
which seamlessly integrate with infrastructure-based networks if
available, and enable conventional Internet services.
This scenario provided me the motivation for designing
and prototyping the multi-tier
wireless networks as part of Ad Hoc City and
Safari projects of our research group.
In Ad Hoc City [2], the backbone network in our architecture is a mobile multi-hop network composed of wireless devices mounted on mobile fleets of city vehicles, each of which connects (possibly via other vehicles) to the nearest wired base-stations. We developed a Cellular DSR (CDSR) protocol for this multi-tiered architecture to enable general purpose wide-area communication. We evaluated our design based on traces of actual movement of a fleet of city buses in Seattle.
In Safari [1], we proposed scalable ad hoc network routing,
leveraging the existing synergy with peer-to-peer networking research.
To this end we developed Masai -- a realization of the architecture --
which employs topology aware, hierarchical addressing for the
mobile hosts through a pro-active, self-organizing, network
hierarchy that recursively groups nodes together.
We developed a hybrid routing protocol
that uses this hierarchy as well as performs on-demand discovery of
routes. Borrowing from peer-to-peer research on lookup, we created a
Distributed Hash Table (DHT) to enable fast lookup of the nodes.
We evaluated this design through analysis and simulations to
demonstrate the scalability of our approach.
Network Edge Architecture
My research internships at Bell Labs and Sun Labs introduced me to the
challenges involved in designing high-speed edge networks. Such
experiences also provided me with the necessary exposure to a typical
environment of industrial R & D.
During my Bell Labs internship, I designed a distributed version of
the OSPF protocol
to handle high load in high speed routers,
taking advantage of multiple processors in the new
generation of routers.
During my Sun internship, I contributed in the design and prototyping of NEon -- an
integrated approach to architecting, operating, and managing network
services. The traditional discrete
approach for implementation of network services like
firewall, packet classification and load balancers in tier-0
suffer from scalability and manageability problems.
In a paradigm shift,
NEon architecture has an integrated approach whereby all the
heterogenous network service rules are aggregated and enforced on the
data packets, by a programmable enforcement device. I also developed a novel
rule aggregation algorithm to generate a consistent set of enforceable
rules from multiple network services, which led to a patent being filed.