Laboratory for Analysis and Architecture of Systems
In recent years, the use of both the Internet and wireless services has experienced an explosive growth. Network operators and service providers anticipate further expansion, fueled by the emergence of all-optical networking as well as the convergence of wireless and Internet access, along with a fundamental trend towards service integration. It is expected that future information and communication systems will accommodate a variety of new applications with a diverse range of Quality-of-Service (QoS) requirements.
These observations have raised the need for the development and analysis of mathematical models to predict and control the QoS of information and communication systems, including wired and wireless networks and large-scale distributed systems. Within this area, our scientific contributions are both theoretical, with the development of new modeling formalisms, and applied, with the development of software tools. The research activities are organized around the following themes:
TCP is the protocol that carries most of the data on the Internet. We develop mathematical models that capture the transient and stationary behaviour of TCP and we propose modifications, in collaboration with standardization bodies like the IETF, in order to improve the performance of TCP.
One of the major achievements in this area is the development of the differential traffic modelling theory for the simulation of large-scale communication networks. It is a unique approach, both for its precision as well as for its application domain. Differential traffic theory has also been extended to the innovative concept of hybrid simulation, a global theoretical framework allowing to combine analytical models and event-driven simulation models. A European and North American patent has been granted. This approach is at the foundation of the start-up QoS Design, a spin-off from our team.
The performance obtained by a set of jobs that share a common resource can substantially be improved by deploying an appropriate scheduling policy. In this line of research, we apply results from queueing and scheduling theory, to determine and develop optimal distributed protocols in order to share the resources of a network among the concurrent flows.
The group has a long experience in optimisation theory, mainly focused on network design and planning problems. Our research mainly deals with optimal routing in IP/MPLS networks subject to QoS constraints, capacity planning and optimal topology design.
We aim at developing stochastic models and software tools to represent accurately the traffic profiles generated by Internet applications, such as VoIP and video codecs or TCP-driven traffic sources (FTP, HTTP, etc.).
Based on measures obtained through a potentially intrusive instrumentation of parallel programs (source code annotations, post-mortem logs analysis), we develop mathematical models of distributed applications in order to predict their processing times. Such models are used for dynamic task mapping on the nodes of a computational grid. Applications are in the field of electromagnetic simulation and distributed