David E. Lapsley

An enhanced random early marking algorithm for Internet flow control

We propose an optimization based flow control for the Internet called random early marking (REM). In this paper we propose and evaluate an enhancement that attempts to speed up the convergence of REM in the face of large feedback delays. REM can be regarded as an implementation of an optimization algorithm in a distributed network. The basic idea is to treat the optimization algorithm as a discrete time system and apply linear control techniques to stabilize its transient. We show that the modified algorithm is stable globally and converges exponentially locally. This algorithm translates into an enhanced REM scheme and we illustrate the performance improvement through simulation.

Random early marking: an optimisation approach to Internet congestion control

In this paper we present an optimisation approach to congestion flow control. The initial context of this approach was as a rate-based flow control in ATM networks. We describe techniques that enable us to implement this flow control in an explicit congestion notification-capable TCP/IP network. These techniques require only minimal changes to existing TCP host behaviour and RED active queue management routers. We call the collection of techniques random early marking (REM). We present the results of a simulation study that looks at the dynamic behaviour of REM and compares it to that of TCP-ECN with RED and drop-tail queue management. We also show how REM can be used to provide differential service between different users.

Random Early Marking

Random Early Marking (REM) consists of a link algorithm, that probabilistically marks packets inside the network, and a source algorithm, that adapts source rate to observed marking. The marking probability is exponential in a link congestion measure, so that the end-to-end marking probability is exponential in a path congestion measure. Marking allows a source to estimate its path congestion measure and adjusts its rate in a way that aligns individual optimality with social optimality. We describe the REM algorithm, summarize its key properties, and present some simulation results that demonstrate its stability, fairness and robustness.

An IP implementation of optimization flow control

Flow control allows sources to adjust their bandwidth usage to the level of availability in a network, and hence reduces packet loss, increases network utilization, and prevents/reacts to network congestion. The Internet uses TCP flow control; asynchronous transfer mode networks will use rate based flow control. In this paper we present an optimization approach to rate based flow control. We describe an implementation of this flow control on an IP network, and we present some performance results of this implementation.

Convergence of asynchronous optimization flow control

We proposed earlier an optimization approach to reactive flow control where the objective of the control is to maximize the total source utility over their transmission rates. The source utility functions model their valuation of bandwidth and can be different for different sources. The control mechanism is derived as a gradient projection algorithm to solve the dual problem. In this paper we generalize the algorithm and the convergence result to an asynchronous setting where the computations at and the communications among the links and sources are uncoordinated and based on possibly outdated information.