Project name: Enabling quality of service in IP networks (EQUIP)
Research Group: Probability and Stochastic Networks (PNA2)
Coordinator: Michel Mandjes
Members: Ton Dieker, Michel Mandjes.
Partners: University of Twente (dr.ir. V.F. Nicola and Ms. T. Igonina M.Sc.). 

Future communication networks will support a broad variety of services. Among these, some will have strict Quality of Service (QoS) requirements; for example, real-time applications, such as telephony and interactive video. The QoS delivered by the current Internet, based on IP (the Internet Protocol), can be described as best effort: no guarantees are provided. This is sufficient for `traditional' Internet services, such as web browsing and e-mail, but certainly not adequate to support real-time applications.

To enable the support of QoS in IP networks, resource management mechanisms are proposed. These range from packet scheduling and priority mechanisms (at the packet transmission level) to load balancing and QoS routing (at the connection/flow level). A QoS routing mechanism selects paths that meet individual users' QoS requirements while optimizing network utilization. It requires the evaluation of QoS along network paths (thus, essentially, also at individual network nodes) as well as algorithms for classical and stochastic combinatorial optimization (e.g., shortest-path and max/min cuts).

In this project, our objective is to evaluate the effectiveness of QoS-enabling mechanisms. Analysis of mathematical models is overly complex, specially because we aim to consider realistic traffic streams (e.g., heavy-tailed file sizes or extremely bursty video traffic) and user behavior for mobile and wireless services. Therefore, we must resort to efficient simulation methods.

We start with models for a single node and incrementally consider networks with multiple nodes and more complicated topologies. For these models, advanced simulation methods for QoS evaluation are developed, using queueing theory, large deviations and adaptive optimization techniques. The latter are also used in stochastic and combinatorial path selection algorithms.

Key publications of EQUIP

• A.B. Dieker and M. Mandjes (2003). On Spectral simulation of fractional Brownian motion. Probability in the Engineering and Informational Sciences 17, 417--434.