Traffic matrix estimation approach based on partial direct measurements in large-scale IP backbone networks

With the rapid development of the IP network, it has been a complex and hybrid network. Specially, the scale of the IP backbone network is much larger than before. For maintaining the normal running of our backbone networks, lots of network management operations are applied to our backbone networks. The network management operators should know the status of our backbone network before carrying out some network management operations. For instance, the end-to-end traffic information is necessary for traffic engineering and network planning. Limited by current traffic monitoring techniques, the traffic information of a large-scale backbone network is hard to be collected. Motivated by that, we propose a network traffic estimation approach to the large-scale IP backbone network. We take advantage of link loads, routing matrix and partial direct measurements of network traffic to construct an optimization model in our method. Then we validate the properties of the proposed approach by the real network traffic dataset from the Abilene backbone network.