Washington University, Campus Box 1045/Bryan 509
One Brookings Drive
Saint Louis, MO 63130-4899
Office: Bryan 405D
314-935-4963 (office)
314-935-7302 (fax)
Email: varghese@askew.wustl.edu
I spent 10 years designing real network protocols (some of which are in products) for DECNET. However, I later did my Ph.D. research in the theory of protocol design. For more details, see my vita . For recent papers, click here . To view a report on our recent work in fast IP address lookup algorithms (appeared on front page of St. Louis Post Dispatch, Sunday. January 4th) please click here .
My goal is to fruitfully combine my experience in distributed systems, both theory and practice. My most recent work in network protocols includes the first O(log(log(n)) algorithm for fair queuing, new techniques for fragmentation and reassembly, and new load balancing protocols. Recent work in distributed algorithms include a result that shows that crash failures can cause almost all asynchronous protocols to fail (assuming that nodes keep no NVRAM across crashes), and techniques for making window protocols self-stabilizing. Web pointers to .ps files can be found below. The work on distributed algorithms is supported by an ONR Young Investigator Award and a grant from NSF; the work in network protocols is supported by an NSF Research Initiation Award.
Past work on network protocols includes techniques for speeding up packet processing based on adding fields to packets, making hop-by-hop flow control both efficient and reliable, efficient fair queuing using deficit round robin, and efficient timer algorithms. Patents for past work include patents for bridge protocols, switch design, the OSI routing update protocol, and rate based flow control. Past work on distributed algorithms includes (together with colleagues at MIT and Univ of Texas) a set of general techniques to make protocols self-stabilizing. A protocol is self-stabilizing if it begins to work correctly regardless of what state the protocol starts in: thus self-stabilization is an abstraction of a strong fault-tolerance property. Some of the techniques we have discovered (for making protocols self-stabilizing) include local checking and correction, local checking and global correction, compiling synchronous protocol, counter flushing, and window washing.
I hope that coupling these two themes (distributed algorithms and network protocols, theory and practice) will provide a rich synergy both in terms of research and coursework. For example, I have taught the network protocol class for senior Wash U undergraduates for three semesters, and have taught them some protocol design techniques. I taught the Distributed Algorithms class in the Spring of 95 and taught a tutorial on applying techniques from distributed algorithms in SIGCOMM 95. I illustrated the theory using many real-world protocol examples. I will teach a tutorial on efficient protocol implementation techniques in SIGCOMM 96.
To look at a course outline for CS 564, please click here. To look at the slides for the SIGCOMM tutorial, please click here. Its in landscape format and you may have to "swap landscape" the file before viewing.
User names registered as of Dec 13th 1994:
sandeep@cs.albany.edu mischu@research.att.com herman@cs.uiowa.edu kutten@watson.ibm.com cowen@brutus.mts.jhu.edu burns@nova.bellcore.com loui@geisel.csl.uiuc.edu jhh@cwi.nl kirousis@cti.gr sriram@cs.uiowa.edu fischer-michael@cs.yale.edu masuzawa@is.aist-nara.ac.jp garay@watson.ibm.com zaks@cs.ust.hk marco@cs.utexas.edu prasad@cs.dartmouth.edu