Architecture and Models for Self-Adaptability of Transport Protocols

Quality of service (QoS) oriented self-adaptive transport protocols are a major issue in the conception of future services for emergent networking technologies (wireless, mobiles, ad-hoc... ). Indeed, transport protocols will have to cope with dynamic applicative and network requirements/constraints throughout the lifetime of a connection while still having to provide the best possible end-to-end QoS to the end users. This is due to the multiplication of access network technologies and the deployment of cross- technology handover (i.e. convergence WiFi/GSM) combined with the generalized mobility of end users in the future "ambient" Internet. Transport protocols whose internal architecture may be dynamically configured appear to be a very promising solution for the support of QoS oriented adaptation. In this context, various problems such as the coordination of distributed adapting entities as well as the best way to guide this adaptation still remain under exploration. Regarding this last problem, model based decision process is a promising approach to avoid ad-hoc specific solutions. Following this approach, this paper introduces an analytical model for the decision process aimed at choosing the best module composition to be instantiated in order to optimize the QoS in a dynamic context. This approach is further illustrated by a case study in which the model is used to self-adaptively optimize applicative QoS for the recently introduced IEEE 802.11e wireless networks standard.