Satellite Terminal Quality of Service Management with AQM Control

Abstract The standardization of a Return Channel via Satellite (DVB-RCS) and the satellite community efforts in term of interoperability over the last few years are expected to play, in a near future, a decisive role in Next Generation Networks (NGNs) through the integration of Satellite networks as an alternative to terrestrial networks like DSL (Digital Subscriber Line) in low terrestrial infrastructure areas. Furthermore, the advance of distributed multimedia applications like voice over IP and videoconference implies some new requirements to guarantee the Quality of Service (QoS). It concerns a limited transmission delay, a weak jitter, a minimal loss rate and a guaranteed bandwidth. The recent NGN architectures prone the adaptation of the diffserv architecture to satellite systems. In this paper, we consider an architecture compliant with the SatLabs recommendation and propose a mechanism to improve the QoS management in Return link Satellite Terminals (ST). The goal of this work is to regulate a priori the Diffserv AF queue and avoid the overflood. Considering the AF queue sheared between UDP based multimedia applications and TCP based data transfer, the main idea consists in controlling TCP streams to guarantee transmission capacity of UDP packets. The most constraint application can then enjoy low buffers time delay and very few losses. In order to solve this problem, we design a congestion control mechanism based on Active Queue Management (AQM) techniques by using control theory. To this end a fluid model of TCP connexion originally designed for wire-networks is proposed for satellite networks. Then, the design of a robust proportional integral (PI) and a robust dead time based controllers are investigated. To avoid AF queue over flooding, TCP packets are voluntary dropped in the ST according to regulation rules. TCP connection throughput is then controlled and limited to protect UDP streams against unnecessary drops. The different methods are then simulated on matlab and NS2-Simulator and compared to a classical DropTail mechanism.