Towards Realistic Models of Wireless Workload

Performance predictions from wireless networking laboratory experiments rarely seem to match what is seen once technologies are deployed. We believe that one of the major factors hampering researchers' ability to make more reliable forecasts is the inability to generate realistic workloads. To redress this problem, we take a fundamentally new approach to measuring the realism of wireless traffic models. In this approach, the realism of a model is defined directly in terms of how accurately it reproduces the performance characteristics of actual network usage. This cuts through the Gordian knot of deciding which statistical features of traffic traces are significant. We demonstrate that common experimental traffic models, such as uniform constant bit-rate traffic (CBR), drastically misrepresent performance metrics at all levels of the protocol stack. We also define and explore the space of synthetic traffic models, thereby advancing the understanding of how different modeling techniques affect the accuracy of performance predictions. Our research takes initial steps that will ultimately lead to comprehensive, multi-level models of realistic wireless workloads.