Abstract

Part 3: Resource Allocation; International audience; There has long been a need for a robust and reliable system which distributes traffic across multiple paths. In particular such a system must rarely reorder packets, must not require per-flow state, must cope with different paths having different bandwidths and must be self-tuning in a variety of network contexts. PREFLEX, proposed herein, uses estimates of loss rate to balance congestion. This paper describes a method of automatically adjusting how PREFLEX will split traffic in order to balance loss across multiple paths in a variety of network conditions.Equations are derived for the automatic tuning of the time scale and traffic split at a decision point. The algorithms described allow the load balancer to self-tune to network conditions. The calculations are simple and do not place a large burden on a router which would implement the algorithm. The algorithm is evaluated by simulation using ns-3 and is shown to perform well under a variety of circumstances. The resulting adaptive, end-to-end traffic balancing architecture provides the necessary framework to meet the increasing demands of users while simultaneously offering edge networks more fine-grained control at far shorter timescales. Document type: Part of book or chapter of book

Abstract

This paper uses analysis and experiments to study the minimal buffering requirements of congestion controlled multimedia applications. Applications in the Internet must use congestion control protocols, which vary transmission rates according to network conditions. To produce a smooth perceptual quality, multimedia applications use buffering and rate adaptations to compensate these rate oscillations. While several adaptation policies are available, they require different amounts of buffering at end-hosts. We study the relationship between buffering requirements and adaptation policies. In particular, we focus on a widely pursued policy that adapts an application's sending rate exactly to the average available bandwidth to maximize throughput. Under this adaptation policy, at least a minimal amount of buffering is required to smooth the rate oscillation inherent in congestion control, and we view this minimal buffering requirement as a cost of maximizing throughput. We derive the minimal buffering requirement for this policy assuming that applications use an additive-increase-and-multiplicative-decrease (AIMD) algorithm for congestion control. The result shows the relationship between parameters of AIMD algorithms and the delay cost. We show that the buffering requirement is proportional to the parameters of the AIMD algorithm and quadratic to the application's sending rate and round-trip-time. We verify this relationship through experiments. Our results indicate that adaptation policies that maximize throughput are not suitable for interactive applications with high bit rates or long round-trip-times. Document type: Part of book or chapter of book