Abstract :
Examines why terrestrial protocols suffer so badly in the satellite environment and offers suggestions as to what might be done to overcome the problems. The paper considers the issues with respect to TCP/IP and looks also at the intense interaction observed between some popular end user applications, the protocol and the satellite environment. The lesson to be learnt is that no element of the system may be viewed in isolation. To do so will inevitably result in a catastrophic reduction in performance. The paper first briefly examines the TCP/IP protocol and looks at the performance of the protocol over a typical VSAT link without any optimisation being made for the VSAT environment. We then look at two possible approaches to improving the situation. In the first we find that significant improvements may be made by tuning some of the protocol parameters to suit the satellite link, a simple step, but one that is often ignored and may be difficult if the end users are not within your control. The second step, more along the lines of a gedanken experiment, involves making slight modifications to the TCP/IP protocol. Since it is unlikely that the Internet Engineering Task Force, who control the evolution of the TCP/IP protocol, would ever consider making changes to the TCP/IP protocol stack solely for the benefit of satellite users, this remains a tantalising impossibility. Finally we look at what might be achieved by means of protocol encapsulation, protocol spoofing, protocol enhancement etc., which give valuable pointers to the possibilities (and problems) of integrating terrestrial/satellite protocols
