Abstract:

To reduce the computational overhead of large-scale network simulation while keeping the simulation fidelity, a network simulation method of focusing on the attention area is presented in which the network topology for simulation is partitioned into two parts: attention area and non-attention area. To keep fidelity, the attention area is simulated by discrete event simulation. And to reduce the computational overhead, the non-attention area is simulated by topology - flow combined abstraction simulation. The simulation fidelity of topology-flow combined abstraction simulation is analyzed. Experiments based on NS2 show that, when the focusing coefficient is 4, compared to the traditional network simulation, this method can decrease the computational overhead of simulation by over 85% at most, also, the average error of simulation results of attention area is only about 2%.

Key words: network simulation, attention area, abstract simulation, computational overhead of simulation, simulation fidelity