PhD graduated
Team : NPA
Departure date : 12/31/2021

Supervision : Naceur MALOUCH, Imad MOUGHARBEL

Co-supervision : DIAB Wafaa Bou

Performance Evaluation and Improvement of Congestion Control of the Constrained Application Protocol for the Internet-of-Things

The Constrained Application Protocol (CoAP) is a lightweight core protocol designed by the Internet Engineering Task Force (IETF) used for communication between devices in the Internet of Things (IoT). Congestion control is substantial to overcome the limitations of such devices and the restrictions imposed by the connecting networks. The CoAP standard defines a simple congestion control mechanism based mainly on retransmissions after timeouts and a binary exponential back-off procedure. However, this simple protocol with its default parameters is not sufficient and can dramatically affect the efficiency of CoAP. Several advanced mechanisms for CoAP were suggested by the literature. Some considered improving retransmission timeout estimation whereas others focused on augmenting the retransmission procedure. The improvements in the second part are twofold: The first is Backoff-based such as CoCoA+ and pCoCoA, while the second is rate-based such as BDP-CoAP. We will perform critical analysis of these works and highlight their shortcomings and pitfalls, then present and evaluate our approach in different uses cases and network scenarios.
In this research work, we propose new exact mathematical models to analyze the performance of CoAP in lossy IoT networks. This study provides insights about improving CoAP congestion control in such networks and highlights the properties -- including the limitations -- of CoAP. Besides, we show that the simple control mechanism reduces significantly CoAP performance especially in terms of bandwidth utilization since it prevents the protocol from acting efficiently during congestion periods. We then propose new improvements in order to enhance the trade-off between reliability and goodput while keeping the algorithms reasonably simple for constrained devices. First, we optimize further the estimation procedure of the retransmission timeout in order to enhance congestion detection. Timeouts are the only indicator used in CoAP to detect losses, and losses are used as a strong indicator to detect congestion. Second, we replace the backoff algorithm by "real" congestion control algorithms inspired from the well-known Additive Increase Multiplicative Decrease technique and a recent measurement-based congestion control called BBR. Our analysis using both our simulator and Contiki/Cooja environment show that the rate-based approach outperforms the backoff-based approach. Moreover, all the results show that our algorithms achieve a much better tradeoff between goodput, reliability and overhead.

Defence : 12/07/2021

Jury members :

Abdallah Kassem - Professor, Notre Dame University [Rapporteur]
Nathalie Mitton - Research Director, INRIA [Rapporteur]
Noël Crespi - Professor, Institut Polytechnique de Paris
Abdulmotaleb El Saddik - Professor, University of Ottawa
Anne Fladenmuller - Professor, Sorbonne Université
Naceur Malouch - Associate Professor (HDR), Sorbonne Université
Imad Mougharbel - Professor, Lebanese University
Wafaa Bou Diab - Assistant Professor, Lebanese University

Departure date : 12/31/2021

2019-2022 Publications