Project Description


Background

TCP congestion control had managed to maintain the stability of the Internet in the past decades but it has reached its limitations of the challenges of present networks. However, the congestion control of TCP has been continuously developed from the first version which dates back to 1981 (RFC 793). The basic mechanism was incrementally developed and tuned introducing new additional algorithms, e.g. slow start, congestion avoidance, RTO calculation and delayed ACK in 1989 (RFC 1122), SACK in 1996 (RFC 2018) and NewReno in 2004 (RFC 3782) just to mention a few of them. The challenges of next-generation networks (e.g. high-speed communication or the communication over different medias) generated an urgent need to further develop the congestion control of the Internet.

The research community has responded with many proposals developed in several excellent research groups all over the world. This significant research work, especially in the previous years, resulted in a huge number of new ideas. These ideas have been developed and implemented in new TCP version proposals. These proposals can efficiently address different aspects of the challenging future network environments and applications but it seems that the process towards finding the transport protocol of the future internet has been slowed down. One of the main problems is that in spite of the fact that lots of proposals have been published finding the best one is difficult since these works are mostly independent research studies and do not compare their results to each others'. Or even if they do comparisons, usually, there is no common performance evaluation framework (same set of metrics and topologies etc.) where a comprehensive performance evaluation of the new versions can be made. Such a comprehensive study is very rare, however, without such studies finding the next step in the jungle of the proposed TCP versions is impossible. This is the need which motivated our work.

Goals

The goal of our research work is twofold. First, we have decided to make a comprehensive performance evaluation of the most promising proposals. Some of our results are already known, since when these proposals were suggested they were investigated too. In these cases our results can confirm or question already published results. However, we also aim to answer many questions which remained unanswered in the literature and we think that they are important to understand the mechanism and impacts of these new algorithms. We call this study as horizontal research because the results of this work give an overall picture about all the investigated TCP proposals where the already remained "black holes" are filled. The main question we have in the focus of the analysis is the fairness characteristics of the different TCP versions including both inter- and intra-protocol behavior with different topologies and parameter settings. We put emphasis not only on the equilibrium behavior but also on the transient characteristics with the dynamic behavior and investigate how the transient state can influence on the long-term fairness performance. We have recognized that this problem has been neglected by the research community but we show that the transient parameters can have significant impact on the equilibrium fairness results. We also introduce a new performance metric to describe the transient state behavior. The main massage of this research shows that the dynamic characteristics should not be neglected in a comprehensive fairness analysis. This overall view is vital when making decisions among these proposals.

Second, we have decided to make a root-cause analysis to get a deeper understanding in the case of some of the promising TCP versions. This study not only fills the "black holes", the questions which remained unanswered in some cases but rather goes deeper and investigates questions which have never been asked yet. We call this study vertical research because in this part we suggest novel methods and get new results which help to provide an in-depth understanding and explanations of the investigated algorithm. For this purpose we have introduced a spectral analysis for the first time as far as we know to analyze and understand fairness performance. With this spectral analysis together with flow-level, packet-level and queueing analysis we have got a good understanding of the investigated phenomena.

Our Related Research Projects


  • Project of Inter-University Centre for Telecommunications and Informatics Cooperative Research Center (ETIK) on "High Speed Transport Protocols"

  • ELTE-BUTE-Ericsson NKFP project on "Research and Developments of Tools Supporting Optimal Usage of Heterogen Communication Networks"