MATZAKOS Panagiotis

MATZAKOS Panagiotis
  • MATZAKOS Panagiotis
  • EURECOM - Communication systems
  • Post doctoral student
  • 04 93 00 82 32
  • 305


Scheduling and Congestion Management Policies for QoS Provision in Disruption Tolerant Networks

Disruption Tolerant Networks (DTNs) generally refer to a wide range of mobile networking environments which suffer from intermittent connectivity, due to reasons such as nodes mobility, poor wireless link conditions etc. In this context, the store-carry-and forward approach is adopted, allowing to store data locally at the nodes, with the aim of surviving communication disruptions. However, ensuring end-to-end reliable data delivery (as in the case of TCP-based Internet applications) is very challenging, especially for environments where the mobile nodes come in communication range randomly (opportunistic contacts). Many existing approaches provide distributed resources management techniques with the aim of maximizing the performance of the network in such resource limited environments. Such approaches generally consider application sessions of equal importance. However, it is envisioned that in many use cases of interest (e.g., military, vehicular, social networks) the mobile users will be launching many applications in parallel, each one with different QoS requirements.


The aim of this thesis was to propose efficient ways to add the dimension of multi-QoS class support in the resources management framework. Specifically, we express the QoS requirements with respect to specific levels of delivery ratio or delay per application class. We provide a distributed constrained optimization framework based on delivery predictions, which aims to guarantee the optimal balance between: (i) individual QoS classes satisfaction and (ii) overall performance maximization. To this end, our approach can be considered as a "loose" equivalent of reliable delivery, mapped to opportunistic DTNs. We then provide different alternatives for extending our policies, in order to account for real life mobility conditions which affect the accuracy of delivery predictions. Through extensive evaluation based on synthetic and real mobility traces we validate the optimality of our approach as well as the fact that it outperforms other existing QoS prioritization policies.