The workload of a network is usually a heterogeneous aggregate of services and applications, driven by a large number of users. This complexity makes it challenging to evaluate the performance of network mechanisms and configurations under realistic conditions. We propose a new methodology for transforming anonymized traces of packet headers into application-neutral models of network traffic. These models are suitable for synthetic traffic generation in simulations and testbeds, preserving the end-to-end nature of network traffic. Our approach provides a tool for studying and tuning the realism of synthetic traffic.
Félix Hernández-Campos, F. Donelson