Link delay inference based on Simulated Annealing constrained optimization method

With the rapid development of the network, it is necessary to know its internal characteristics, so to design, control and manage the network. At present, network measurement is mainly deployed to acquire internal characteristics of the network and identify network topology. Network Tomography technology is one of the network measurement methods. Inference of network internal link characteristics is a very important to Network Tomography technology. On the basis of traditional methods used to infer link delay characteristics in Network Tomography, we propose a new method based on Simulated Annealing constrained optimization to improve the precision of inference of link delay cumulant generating functions by gathering the constrained information in network. The simulation results show that with measurement error, our constrained optimization method infers link delay more precisely than non-constrained ones do.