A
Framework for Delivering Quality of Service Guarantees in Communication
Networks
Description: This project focuses on
the development of a general framework for performing real-time
resource allocation for network management and control functions
in the context of wired and wireless networks. The framework
that is proposed for this research is a hybrid modeling
paradigm based on Stochastic Fluid Models (SFM), which have
recently been shown to be especially useful for analyzing
various kinds of high-speed networks. Using this modeling
framework a new approach for network management is proposed
which is based on Infinitesimal Perturbation Analysis (IPA).
This approach is computationally efficient; It is based
on estimating the gradient of the performance measure of
interest (e.g., packet loss rate) with respect to control
parameters (e.g., buffer thresholds) and subsequently employ
them in standard stochastic approximation algorithms to
determine the optimal parameter setting.