| UP | AT BOTTOM | ||
|---|---|---|---|
| Adaptive Statistical QoS: Learning Parameters to Maximize End-to-End Network Goodput | |||
| Evans,Scott | GE Research | ||
| Yan,Weizhong | GE Research | ||
| Weerakoon,Ishan | Lockheed Martin | ||
| Rothe,Asavari | GE Research | ||
| Liu,Ping | GE Research | ||
| Goebel,Kai | GE Research | ||
| Egan,Martin | Lockheed Martin | ||
| We present an Adaptive QoS System that seeks to maximize end-to-end success through learning algorithms that take queue depths as input to control Weighted Fair Queue provision. Utilizing an analytical model we generate queue depth and E2E success data for various levels of load and WFQ provision and generate a WFQ provision surface for two classes of real time traffic using Neural Network techniques. We verify the nature of the surface through event driven simulation and discuss future opportunities for adaptive QoS policy management. | |||
| Dr. Scott C. Evans is an Electrical Engineer at General Electric Research in Niskayuna, NY. Scott conducts research in advanced communications, ad hoc networks, and algorithmic information theory for information security and DNA sequence analysis. He has numerous patents in the area of wireless and power line communications, network routing and information security. Scott holds a PHD in Electrical Engineering from Rensselaer Polytechnic Institute, an MS in Electrical Engineering from the University of Connecticut and a BS in Electrical Engineering from Virginia Tech. Before joining General Electric Research, Scott served as nuclear-trained Submarine Officer in the United States Navy. He can be reached at evans@research.ge.com. | |||