Traffic matrix estimation with enhanced origin destination generator algorithm using simulation of real network

Abstract

The rapid growth of the Internet has made the issue of ensuring reliability and redundancy a big challenge. Studies of these issues using Traffic Engineering and simulation have been extensively done. In Traffic Matrix Estimation (TME), the Origin–Destination Generator algorithm (ODGen) is limited to the number of hops, where the Expectation Maximization (EM) accuracy is 92%. Most studies have not taken into account real traffic parameters and integration of TME models with routing protocols in their simulation models. Also, there is no a comprehensive model consisting of TME, Border Gateway Protocol (BGP) and Hot Potato (HP) routing in the NS-2 network simulator based on real networks. In this research, Integrated Simulated Model (ISM) is introduced consisting of ODGen-HP algorithm and BGP integrated into the NS-2 network simulator. ISM is then used to simulate the infrastructure of a real production network using actual captured traffic data parameters. Validation is then done against the changes in network topology based on packet loss, delay and throughput. Results gave the average error for packet sent by simulated and production networks of 0% and the average error for packet received by simulation and production networks of 3.61%. The network is modelled with a baseline topology where 5 main nodes were connected together, with redundant links for some nodes. The simulations were repeated for link failures, node addition, and node removal. TME used in ISM is based on ODGen, that is optimized with unlimited number of hops, the accuracy of EM increases to 97% and Central Processing Unit complexity is reduced. HP helps in improving the node which experiences a link failure to select shorter distance route to egress router. In the case of a link failure, HP switching time between the links is 0.05 seconds. ISM performance was evaluated by comparing trace file before and after link failure or by adding nodes (up to 32) or removing nodes. The parameters used for comparison are the packets loss, delay and throughput. The ISM error percentage obtained for packets loss is 0.025%, delay 0.013% and throughput 0.003%.