Sumathi Gopal – MS Rutgers – Doing Phd at WINLAB, Rutgers

I am a native of Bangalore, India.

I graduated from BMS College of Engineering with a degree in Electronics and Communication, in 1998.

I did my Masters from Electrical Engineering Dept, Rutgers University. During this time, I was a Graduate Assistant at CAIP, Rutgers University, NJ, USA. My advisors were Dr. James Flanagan, Dr. Attila Medl and Dr. Richard Mammone.

I then worked as a Research Associate atC&C Research Labs, NEC USA Inc. located in Princeton, NJ. followed by a brief employment at Corporate Research, Thomson Inc., also located in Princeton NJ.

I am currently doing my PhD in wireless networking at WINLAB, Rutgers University. I am supported by a fellowship from Corporate Research, Thomson Inc.

Office Location : CoRE 513/ CoRE 529
Phone # : 732 445 0603/732 445 0614/609 987 7327
Email : sumathi@winlab.rutgers.edu

Hmm.. so who is ..? Hello! I am a PhD student under the able guidance of Dr.Dipankar Raychaudhuri (Director, WINLAB). I completed my Bachelor Of Engineering (‘Electronics and Communication’) at Bangalore University (B.M.S. College of Engineering), India, and my Masters of Science(Digital Signal Processing) at the Electrical Engineering Dept, Rutgers University.

What am I doing here ? My two years of work experience in C&C Research Labs, NEC USA Inc., and Corporate Research, Thomson Multimedia, introduced and enticed me to the world of Computer Networking. WINLAB is a great place to be, to pursue my research interests which include wireless networking, media delivery and systems analysis. My current research pertains to developing new network architectures and protocols to enable efficient media delivery over wireless networks.

Good News! The past decade has seen tremendous advancement in wireless technology with the advent and widespread deployment of wireless WANs (cellular phone technology) and LANs (802.11 protocols, HiperLAN). The wireless LANs in vogue today have facilitated high bandwidth availability. “A picture is worth a thousand words”… this is an understatement when the ‘picture’ under consideration is a video or a multimedia presentation! With the simultaneous advent of Third
Generation Cellular Networks that are IP-based (and internetworking a near possibility), the next generation ‘killer applications’ are expected to be multimedia based.

Well, who said it would be easy!! There are several challenges for media delivery over wireless networks. Media applications are characterized by stringent QoS (Quality of Service) requirements such as strict upper bounds on end-to-end bulk delay and delay-jitter. Wireless channels are characterised by bursty errors, dynamic variation – even blackouts – depending on the incumbent fade. 802.11b does not have provision for QoS. Thus despite availability of high bandwidth, 802.11b networks in the current state cannot support quality multimedia data.

Hmm.. that’s challenging! We propose to approach the problem from two perspectives:

1. Application level channel adaptivity by dynamically adapting encoding/transcoding rates to available channel capacity (in collaboration with Corporate Research, Thomson Inc, Princeton, NJ).

As the first step, we are working on an appropriate channel model at the packet level. There is little prior research in this regard for the wireless channel under consideration – 802.11b. Yet, the following publications provide valuable insight:

<A> Nguyen G.T., Katz R. H., “A Trace-based Approach for Modeling Wireless Channel Behavior”, Proceedings of the Winter Simulation Conference, Coronado, CA. December 8-11, 1996

<B> A. Konrad, A.D. Joseph., R. Ludwig, B.Y. Zhao, “A Markov-Based Channel Model Algorithm for Wireless Networks”, Report No. UCP/CSD-01-1142, Computer Science Division, University of California, Berkeley, 2001.

The traditional method of representing the wireless channel with a two-state markov model is flawed as it is based on the faulty assumption that wireless channel errors are random, while in actual, they occur in bursts. The above two papers suggest alternative ways to model channel errors, based on a trace-based approach.

2. An adaptive transport protocol for efficient media delivery over wireless networks.

The purpose of this transport protocol is to dynamically adapt to requirements specified by the application above considering the channel condition below. We believe that some level of inter-layer optimization will drastically improve performance of media delivery over wireless networks.

This approach results from understanding characteristics of media applications and wireless networks, and limitations of existing transport protocols lsuch as UDP and TCP. These are two extremes in a spectrum of transport protocol functionality. Media applications require intermediate services, such as fully reliable service on few packets (such as I-frames for MPEG video) and tolerable loss of others (P and B-frames). They can also tolerate frame delays within certain upper bounds. The transport layer is also the lowest layer in the OSI protocol stack, that is equipped to percieve the end-to-end network picture. It thus becomes an obvious choice for network adaptation. But this adaptation needs to be honed to the characteristics of the end-to-end network. On the otherhand, TCP and for that matter even the Real-time Transport Protocol (RTP) are specifically optimized for the wired Internet where packet losses are primarily due to network congestion and buffer overflow.

Following are some publications that elaborate on transport protocols that accept service specifications from an application, and likewise adapts their transmission.

1. Senac P., Exposito E., Diaz M., “Towards a New Generation of Generic Transport Protocols”, Proceedings of Thyrrhenian International Workshop on Digital Communications (IWDC 2001), Taormina, Italy, September 17-20, 2001. http://www.laas.fr/GCAP/docs/publications/IDWCSenacExpositoDiaz.pdf

2. Dairaine L., Diaz M., Rojas-Cardenas L.; Senac P., “An adaptive transport service for multimedia streams”, Multimedia Computing and Systems, 1999. IEEE International Conference on , Volume: 2 , Jul 1999

We are also exploring new and more efficient network architectures required for improved media delivery.

The Research section cites many more publications from which we derive related work.

I welcome comments and feedback, so please feel free to contact me if you find this work interesting and would like to know more…

Sumathi Gopal