Bridging the Gap Between Reality and Simulations: An Ethernet Case Study

Simulation is a widely used technique in networking research and a practice that has suffered loss of credibility in recent years due to doubts about its reliability. In this paper, we conduct a simple Ethernet experiment on a test-bed, and try to reproduce the results using three popular network simulators: ns-2, QualNet and OPNET Modeler. We find that after careful simulation we are able to approximate test-bed performance in all three simulators, and we document the simulation process for each one. We also identify blocking socket calls as a behavioral attribute of applications that affects performance, but cannot be set in any of the three simulators used.