On Modeling, Wireless Links, Transport Protocols and Roofnots

The paper by Gurtov and Floyd argues that the design and implementation of transport protocols can be improved by carefully modeling specific characteristics and behaviors of wireless links. It seems an appropriate next paper to cover since the “tussle” and the interplay of wireless links and transport protocols have been introduced in previous discussions. They discuss modeling and the effects on transport protocols of specific wireless link characteristics such error losses, delay and bandwidth variation, packet reordering, on-demand resource allocation, and asymmetry. Queue management and mobility models and its effects are also covered. A nice take away from this paper is that both designers of the transport and wireless protocols must consider each other’s concerns to provide a better and more efficient end user experience.

Bicket and et. al paper examines how wireless mesh architectures with minimal planning and management can be effective in providing end-to-end quality Internet access. It uses as a case study the Roofnet mesh network. The results are generally positive such as inter-node dsl-like throughput and delay characteristics; and the usable multi-hop throughput and coverage. I was first disappointed at some of things the study did not include. In particular, the study does not include the evolution of the network (users entering and leaving) which is I think is very important. While my initial reaction was perhaps “clouded” (influenced)  by Gurtov and Floyd’s discussions on assumptions and modeling, I do realize that the contributions of this work is substantial and that it is within the scope of what they have set out to do. I also appreciate Gurtov and Floyd’s take on modeling must be tempered in certain contexts such as the case of examining and evaluating behavior in test-beds.