Implementation of stateless transport protocols in ns-3

Development of transport protocols has received a great deal of attention of network research community at several past decades. One of the general directions of such an effort was to improve a congestion control mechanism of the TCP (Transmission Control Protocol) which is tightly bounded with other components of the protocol responsible, for example, for robust delivery of data and loss detection. Such a solid architecture complicates innovations in this area and leads to inefficient or misleading functioning in different network settings.

Last decade has shown the emergence of new communication paradigms such as cloud computing, software-defined networks, sensor networks, fog networks etc. Thus, rethinking architecture of the transport protocol can be useful to comply new demands. The standard TCP approach tics the transport connection to its endpoints however approaching network applications in new network settings may demand more flexible and transparent data transfer. For example, in cloud computing architectures, servers can dynamically power on or shutdown and such a behavior must be transparent for clients. This can be difficult or even impossible to achieve if the transport protocol's state is distributed between both sides of the connection. We consider a protocol called the Trickles, which is one of the first efforts to migrate all connection state to one endpoint, allows its counterpart to operate without any per-connection state.

In this paper we describe the architecture of the model of such a stateless protocol and describe a framework which can be used to model such protocols in ns-3. Another contribution of the paper is an approach based on ideas of literate programming to achieve reproducible results of analysis of network protocols.