HNDP: A Novel Network Distance Prediction Mechanism

Network distance is an important parameter in optimizing performance of network applications. Although there are a number of network distance prediction mechanisms, they all take no consideration of Internet structure, which has great influence on Internet distance characteristics. By analyzing the hierarchical structure feature of Internet, a hierarchical network distance predication mechanism called HNDP is proposed. HNDP divides Internet into many independent prediction regions, and predicts distance information between network nodes by accumulating distances in different predication regions, which can avoid the problem that short distance and long one cannot be accurately predicated simultaneously. To optimize the influence of landmark selection on HNDP prediction accuracy, a shortest distance cover landmark selection model is proposed, and then a tabu search algorithm called TS_Landmark is given to solve this model in HNDP. Finally, the simulation results under ns-2 show that TS_Landmark can select landmarks effectively, and HNDP provides more accurate results than traditional single layer ones.