DETERMINISTIC ANALYSIS OF
		 FLOW AND CONGESTION CONTROL POLICIES
			 IN VIRTUAL CIRCUITS

			     Samar Singh
	   Department of Computer Science and Engineering,
		    Indian Institute of Technology
				   
			  Ashok K. Agrawala
		   Department of Computer Science,
			University of Maryland
				   
			  Srinivasan Keshav
	    Computer Science Division, Department of EECS,
		 University of California, Berkeley,
				   
			       Abstract

Network congestion can be triggered by transient phenomena that are
hard to study using classical stochastic approaches.  We show that a
deterministic analysis of virtual circuits (VC) allows us to precisely
quantify these transients.  We analyze some existing window and rate
based flow control schemes, and propose a new protocol, 2P, that
exploits the deterministic nature of a VC to provide optimal
performance.

Our analysis focuses on two issues: utilization of available bandwidth
and congestion avoidance. We study the transient behavior of the
system when a sudden change occurs, such as loss of a packet or a
reduction in the available bandwidth, and give precise expressions for
the performance of these schemes. We show that 2P performs better than
existing schemes since it adapts to changes in the state of the VC by
keeping track of the effective transmission rate as well as the
round-trip delay. This is done using a novel rate probing technique
based on short bursts.