Cross-Layer Based Simulation for QoS Guarantees in Satellite IP Networks

Abstract

Cross-layer protocol design and engineering is a novel approach that aims to improve the performance of IP-based satellite communication systems by exploiting interactions among different protocol layers in the air interface. The cross-layer design is an interdisciplinary research area. The cross-layer protocol design includes the study of the interactions among physical, MAC and transport layers that can permit both to improve the utilization of satellite radio resources and to meet QoS requirements of differentiated traffic classes. The main emphasis of this paper is in evaluating the impact on the TCP goodput due to two different transmission mode selection techniques for the air interface. The first one bases the selection of the transmission mode on the packer error rate required at layer 2 (non-crosslayer method); while, the other technique is based on the TCP goodput performance (crosslayer method). Our simulation results demonstrate that the second scheme provides significant end-to-end goodput performance improvement. I. Introduction ATELLITE communications play a significant role in supporting next-generation IP-based networks. To deploy state-of-the-art satellite technologies supporting media-rich applications, efficient utilization of radio resources and end-to-end Quality of Service (QoS) support are mandatory requirements 1 . Within the research community a wide range of issues is currently being investigated to improve the efficiency and the capacity of satellite communication systems. The need of efficiently utilizing satellite radio resources calls for innovative approaches that are based on a fullscale optimization of the satellite radio interface. A possible solution is represented by the cross-layer design proposal, where interactions among different protocol layers, even non-adjacent ones, are considered to improve the capacity of the air interface. An important basis for the layered approach (ISO/OSI and Internet protocol stacks) was the Von Neumann architecture for computers with a separation between software and hardware. Moreover, another element in favor of layering were the Shannon’s results, proving that the layers of source compression (source coding at application layer) and coding for reliable transmission over a communication channel (channel coding at physical layer) may be separately and independently implemented 2 . This result allows the distinct optimization of these coding parts, thus greatly reducing the theoretical complexity. However, this separation theorem in coding theory is no longer valid for general time-varying channels, like in the wireless and satellite scenario; in these cases, channel and source coding need to be jointly optimized. In general, there exists tight interdependence between layers in satellite networks; hence, a strict modularity and layer independence may lead to a non-optimal performance. The interest is here in protocol architectures where the strict layered approach is overcome by either the joint design of different protocol layers that are optimized for suitable operating conditions or the dynamic adaptation of different (even non-adjacent) layers by allowing the direct communication between protocols at non-adjacent layers or sharing state variables