Quantitative end-to-end arguments: Fundamental

The end-to-end (e2e) arguments are of the most influential design principles for the Internet for about three decades, which basically suggest a design principle that puts the application-level functions at the network rather than inside the network as much as possible to simplify network design and implementation. In many cases moving some functions from a network edge into the network is often considered to improve network performance. So it is necessary to learn whether such function displacement is cost-effective. However no quantitative solution is available to determine performance gains given by such displacement and compare different options if any. It is necessary to study quantitative end-to-end arguments. Reliable e2e transmission control usually follows well the end-to-end arguments such as TCP. But such e2e approach faces critical challenges in certain new networks such as multi-hop wireless networks and all-optical networks, in which moving certain functions from the network edge into the network is necessary for good performances. This paper studies some basic reliable e2e transmission control approaches. It is shown that the possible performance gain given by non-e2e over e2e ones is significant.