Verification of Network Simulators : The good, the bad and the ugly

Abstract

The purpose of a simulation is to try to get an indication on how a given system will behave in a range of different scenarios. For this kind of research to have any value, it is important that the simulator models reality in a reliable way. In this thesis, we have chosen a common test suite through which we evaluate the TCP functionality of three major open- source simulators: ns-2, ns-3 and OMNeT++. As our set of test cases, we have used the TCP evaluation suite, which is meant to become a standardized test suite for evaluating TCP extensions. The suite already exists for ns-2, while we have implemented elements of the suite in both ns- 3, and in OMNeT++. As a consequence of this, we have had to implement the tmix traffic generator as well for OMNeT++, and we have also worked on improving tmix for ns-3. These implementations of the test suite lead to some interesting observations regarding the TCP functionality of each network simulator. Through working with the test suite however, we have revealed several weaknesses in it. We have identified a range of items that need improvement before the test suite may become a useful tool.