Maintaining the accuracy of focusing topology area in network simulation

Abstract

In network simulation, higher simulation accuracy generally leads to significantly increased computational overhead. To mitigate this issue, we present a new network simulation method for the focusing topology area (NESFOTA). The idea is to partition the topology into two parts that are complementary to each other, namely the focusing topology area and the non-focusing topology area. The focusing topology area is of interest to the simulation users and it is simulated using the traditional packet-level models to attain satisfactory accuracy. On the other hand, the non-focusing topology area is simulated with a higher level of abstraction to decrease the computational overhead. In particular, a method is proposed for the non-focusing topology area simulation, and theoretical analysis shows that it degrades marginally the simulation accuracy of focusing topology area. Compared to the traditional method, the NESFOTA method reduces the computational overhead by about 10 times at most while achieving nearly the same simulation accuracy of focusing topology area.

Keywords:

• network simulation
• discrete-event simulation
• simulation accuracy
• networks and graphs

Acknowledgments

This work is supported by the National Key Research and Development Program of China (Grant No. 2016YFB0800305) and by the National Natural Science Foundation of China (Grant No. 61672264, 61602214).

Statement of contribution

In the evaluation of large-scale and high-speed computer networks by simulation, higher simulation accuracy generally leads to significantly increased computational overhead. To mitigate this issue, we present in this paper a new network simulation method for the focusing topology area (NESFOTA). The main contributions are as follows:

1. The topology is partitioned into two parts that are complementary to each other, namely the focusing topology area (FTA) and the non-focusing topology area (NFTA). The FTA is of interest to the simulation users and it is simulated using the traditional packet-level models to attain satisfactory accuracy. On the other hand, the NFTA is simulated with a high level abstraction to decrease the computational overhead. Meanwhile, we maintain the topology structure of NFTA and do not collapse or ignore it. In this way, all the routing paths are maintained, for simulation accuracy.

2. We propose a high-level abstraction model (CAQUP) for the NFTA, the main feature of which is to maintain the simulation accuracy of drop rate and forward delay when one packet is transmitted through the NFTA. In this way, the CAQUP model degrades marginally the simulation accuracy of FTA.

3. The experimental results prove that the NESFOTA method reduces the computational overhead by about 10 times at most while achieving nearly the same simulation accuracy of FTA. The NESFOTA method also presents its superiority in offering higher simulation accuracy over the existing hybrid methods with fluid models.