Most of the Internet traffic is transported by TCP (Transmission Control Protocol) which applies a congestion control mechanism. However, this mechanism is not able to work efficiently in today's constantly changing network environments. In order to solve this issue, a number of new TCP variants have been developed which are able to utilize the network resources more efficiently by modifying the conventional congestion control algorithm. Recently, the widely used TCP versions are able to provide more efficient solutions in given environments but they do not give universal and optimal solutions to the challenges of today's ever-changing, heterogeneous network environments. Currently, it seems there is little hope that the TCP's closed-loop congestion control mechanism will be able to provide such a universal solution in the future.
In order to utilize the network resources optimally and find answers to the interoperability and efficiency issues of TCP, extensive research is now being conducted by which innovative concepts have been presented. One of these promising ideas suggests omitting congestion control completely. In this case, every entity in the network is allowed to send data at its maximum rate, which makes it possible to fully utilize the network resources. The emerging huge, mostly bursty packet loss is then compensated by applying effective erasure codes by which all the data can be restored at the receiver. In the past few years, a new transport protocol has been designed in the HSN Lab (High Speed Networks Laboratory) of the Department of Telecommunications and Media Informatics. The protocol is called DFCP (Digital Fountain based Communication Protocol) and is currently under research and development. I have implemented DFCP in the Linux kernel and carried out an extensive performance evaluation of the protocol. Instead of using a congestion control mechanism, DFCP applies efficient erasure codes in order to be able to cope with the emerging packet loss. The operation and behavior of the protocol have been analyzed on various network topologies by carrying out both testbed measurements and simulations. The preliminary results of the experiments are presented in this thesis and have confirmed that DFCP is able to achieve better performance in several network scenarios compared to different TCP variants.
Firstly, following the introduction, this work provides an insight into the operation of the most extensively used TCP versions today reviewing their congestion control algorithms. In the next chapter, a detailed description is given of the underlying concept of the protocol. Subsequently, the basic operation and the features of DFCP are discussed. In the last part of the thesis, firstly, the environment used for the measurements is described. Secondly, the results of the validation measurements are presented and, in addition, a comparative analysis is carried out during which the performance of DFCP is compared to commonly used TCP variants. The results of the measurements are discussed in detail and finally, the future plans and possible improvements regarding the protocol are summarized.