A study of network for multi hour traffic under splittable and non splittable flow conditions
Abstract
In a multi hour networking environment, there are different traffic matrices operating during different time periods of the day depending on the demand. It is applicable to traffic networks such as IP networks or optical networks. In this work, we attempt to analyze how the different networking objectives relate to or differ from each other under multi hour traffic conditions. In particular, we have used three different objectives for our study, namely minimum cost routing, load balancing optimization (minimize link utilization) and minimize average delay. We consider them under both Splittable and non-Splittable flows. Our study focuses on considering three demand conditions: low load, moderate load, and high load. Besides the objectives as specified, we also consider the following indicators: (1) Number of non-zero paths to observe how the model behaves in terms of satisfying the demand, (2) Number of demands taking more than one path to observe how likely a model tries to spread the demand to more than one paths, (3) A distance measure, which represents the difference in path allotment between Splittable and non-Splittable cases to observe how closely related the models are with respect to satisfying a demand. Through an extensive study with four different network topologies and demand scenarios, we observe that the minimum cost routing model tries to use a single path to satisfy a demand as much as possible, while the minimizing average delay model shows the most tendency to split the demand and use more than one paths. For the load balancing model, once it chooses a set of paths, it retains those paths unless the problem becomes infeasible.
Table of Contents
Introduction -- Materials and methods -- Optimization models -- Results -- Pre analysis -- Conclusions -- Appendix A. Model files for the objectives
Degree
M.S.