Methods for the Estimation of Traffic Matrices

The accurate knowledge of origin/destination traffic matrices allows network operators to efficiently
perform network management operations to maximize network performance (minimize network
congestion, average delay, jitter, energy consumption, etc.) and increase network reliability in case
of device failures and other exceptional events. However, traffic matrices cannot be directly
measured, and network operators, such Videotron, can only rely on estimations based on the
elaboration of a limited amount of information, such as partial peak flow, link utilization, SNMP link
counts, multicast group composition and multicast trees. The aim of this project consists in
developing an efficient method for the optimal estimation of traffic matrices, when both multicast
and unicast traffic is considered. In this way Videotron will improve the performance of network
management operations that are crucial for the efficient functioning of the network.

Faculty Supervisor:

Brunilde Sansò

Student:

Partner:

Videotron (Montreal, QC)

Discipline:

Engineering

Sector:

Information and cultural industries

University:

École Polytechnique de Montréal

Program: