Research On Traffic Engineering Technologies In Supanet

This thesis is dedicated to Traffic Engineering techniques in SUPANET, an experimental NGI (Next Generation Internet) developed at Sichuan Network Communication Technology Key Laboratory. SUPANET is characterized by a Single-layer User-data switching Platform Architecture to meet the challenges faced with existing Internet, i.e. high-speed switching, QoS-provisioning, security, mobility, and etc. Traffic engineering (TE) is a technique concerned with network operation optimization both in service provisioning and in communication-resource utilization by distributing traffic loads across network links.This thesis first introduces the architecture and working principle of SUPANET, based on the analysis of the necessary of SUPANET Traffic Engineering and the limit of IP Traffic Engineering, studies the basic framework and main components of SUPANET Traffic Engineering and discusses the relationship between them.There are two main approaches towards TE, i.e. to improve routing protocols dynamically according traffic monitoring results and to use of load-balancing algorithms to lead to better communication resource utilization. Considering that SUPANET provides QoS-insured, connection-oriented Virtual Path Service (VPS). The VP determines the distribution of traffic in the subnet, so this thesis focus on the dynamic routing algorithm based on traffic monitoring results collected in user-data platform in SUPANET and distributed via the Traffic Message Exchange Protocol (TMEP), but will not discuss traffic information collection and exchange process further and assumes that such information is available at individual nodes.The SUPANET adopts a two-step routing policy, i.e. port routing to decide a fiber path for a given source-destination pair and lambda routing to select a lambda best suitable for the required QoS for that data stream. Based on research of typical algorithms, the author proposes Service Class Based Routing algorithm (SCBR) for the port routing. SCBR algorithm has two phases:a quota preprocessing phase, where to pre-allocate link capacities for various traffic classes based on the multi-commodity flow problem; and an online routing phase computed the reasonable path for an individual flow request. The online phase takes the function of the available quota as link's weight. Simulation results indicate that the algorithm performs better than the traditional algorithms in aspects of network load balance and request rejection ratio.