Showing papers by "Peter Key published in 2001"

System and method for controlling network demand via congestion pricing

Abstract: A method and system for avoiding network congestion by measuring network load to adjust a rate at each source. The load (e.g., as a price value) is provided to network traffic sources, which adjust their flows based on a per-flow weight (willingness to pay). The load may be determined in-band by routers or out-of-band by an observer node. A load value (e.g., as a price) is increased or decreased based on the actual load or an estimate based on the actual load. The sources obtain the load value and control the output rates per flow based on each flow's associated weight (willingness to pay), e.g., as set by an administrator. Further, two classes of applications are enabled by allowing certain applications to ignore price.

Modeling RED with Idealized TCP Sources

Abstract: We analyze the dynamic behavior of a single RED controlled queue interacting with a large population of idealized TCP sources, i.e., sources obeying the rules of linear increase and multiplicative decrease. The aggregate trac from this population is modeled in terms of the time dependent expected value of the packet arrival rate which reacts to the packet loss taking place in the queue. The queue is described in terms of the time dependent expected values of the instantaneous queue length and of the exponentially averaged queue length, for which we also derive a pair of dierential equations. This provides us with a complete model for the dynamics of the system which we use to explore transient and equilibrium behavior. The accuracy of the model is veried by comparison with simulated results.

Distributed control and resource marking using best-effort routers

Abstract: We present a method for creating differential QoS where control is in the hands of the end system or user, and the network distributes congestion feedback information to users via packet marking at resources. Current proposals for creating differential QoS in the Internet often rely on classifying packets into a number of classes with routers treating different classes appropriately. The router plays a critical role in guaranteeing performance. In contrast, there is a growing body of work that seeks to place more of the control in the hands of the end system or user, with simple functionality in the router. This is the approach outlined in this tutorial article: using insights from economics and control theory we show how cooperation between end systems and the network can be encouraged using a simple packet marking scheme. The network distributes congestion feedback information to users via packet marking at resources, and users react accordingly to obtain differential QoS.

Resource Pricing for Differentiated Services

Abstract: In paper we present an overview of recent work on resource pricing for differentiated Services in the Internet. This approach is based upon encouraging Cooperation between the end-systems and the network by use of the correct feedback signals. These Signals reflect the congestion shadow prices at a resource, and their use means then even ‘selfish’ end-systems, acting in their own best interests, will push the System to a global or social Optimum. In contrast to most current Diffserv proposals, little is required from resources in the network; they just have to mark packets correctly, while the end-system can use complex or simple strategies. All that is needed is for the end-systems to have an incentive to react to the feedback signals, and then we have a distributed resource sharing mechanism. We give examples of typical end-system behaviour, and show how this approach can also implement Distributed Admission Control, where the decision is in the hand of the end-system. We comment on how ECN (Explicit Congestion Notification) could be used as an enabling technology. Lastly we outline how guarantees can be constructed with this framework.

Modelling the performance of distributed admission control for adaptive applications

Abstract: We outline a methodology for the performance analysis of distributed admission control for adaptive applications, or long-lived flows, where each flow can transmit at one of a discrete set of rates, and can switch between these rates. Specific examples are certain voice over IP or conferencing applications. In the Internet, UDP based protocols such as RTP have the potential to allow sources to alter their sending rates using periodic feedback, where again the rate can take values from a limited set.

In-call Probing and End-to-End Congestion Control: Theory and Performance

Abstract: This paper looks at adaptive applications that can switch between a small number of different levels, with switching decisions made solely by the originating end-system. Typical of such applications are real time streaming protocols which can use different coding rates. The end-systems probe the network with their own traffic to determine congestion, and decide at what rate to enter according to the fate of their “probe” packets. During the lifetime of a connection, the procedure is repeated to reassess and possibly readjust the rate. We derive analytic models, based on diffusion limits under a natural scaling, to quantify the benefits of in-call probing. We then use simulation to compare the results in a number of scenarios, and show that this simple theory is remarkably accurate in predicting large-scale behaviour. The results also show that a small amount of in-call probing produces significant benefits to the system.