Wireless networks are in the process of becoming the dominant means of client-side access to the Internet and to many Intranets. Yet many aspects of wireless networks are poorly understood, and poor performance occurs in many situations. A key obstacle to improving understanding and performance is to be able to conduct accurate, controlled, and repeatable experiments.

I have developed a functional wireless network emulator that provides researchers and developers with a powerful means of conducting wireless research and development. This approach greatly improves wireless testing accuracy and shortens wireless application, protocol, and hardware development cycles.

This approach has broad applicability both in academic research and industrial development. If you have an interesting application for this technique, I'd like to hear it.

Traditionally, wireless networking researchers were confronted with a fundamental tradeoff between the realism of wireless testbeds and the repeatability and flexibility of wireless simulation. The wireless emulator that I have developed strikes a compromise between the realism of testbeds and the control of simulation.

Sample experiments and results can be found by clicking on the image to the right.

The following papers outline my approach:

         Glenn Judd and Peter Steenkiste
          Using Emulation to Understand and Improve Wireless Networks and Applications
         Proceedings of NSDI 2005

         Glenn Judd and Peter Steenkiste
          Repeatable and Realistic Wireless Experimentation through Physical Emulation
          Proceedings of HotNets-II

This approach has many benefits: unlike wireless networking simulators, it allows researchers to conduct experiments with real applications on real wireless hardware, and unlike wireless testbeds, these results can be repeated and verified by local and remote researchers. My approach is similar to that of traditional wireless channel emulators, but unlike channel emulators, my emulator supports a very large number of channels, thereby enabling simultaneous emulation of an entire wireless network. Moreover, my emulator can trade off fidelity for scale to achieve either fine-grained emulation of a small number of channels or coarse-grained emulation of a large number of channels.

Using this emulator as a basis for research, I am analyzing current wireless data networks to see where they fail, how they can be enhanced, and how they can support novel mobile computing applications.

For a good example of how even a small emulator can shed light on the behavior of live wireless networks see:

         Daniel Aguayo, John Bicket, Sanjit Biswas, Glenn Judd, Robert Morris
          Link-level Measurements from an 802.11b Mesh Network
          Proceedings of SIGCOMM 2004.

(Roofnet is a great example of a live network with plenty of fun issues to research.)

Chaotic Wireless Networks

The massive unplanned deployment of unmanaged devices in license-free spectrum creates the potential for massive interference and poor performance. I have been analyzing these "Chaotic Wireless Networks" to measure how severe this problem is, and how it might be mitigated:

         Aditya Akella, Glenn Judd, Srinivasan Seshan, and Peter Steenkiste
          Self-management in Chaotic Wireless Deployments (journal version)
          Self-management in Chaotic Wireless Deployments (Mobicom version)
         Proceedings of Mobicom 2005

I am also interested in leveraging the new applications made possible by the widespread deployment of high-speed open-access wireless data networks.

Additional Work

My earlier work at CMU includes a look at 802.11 load balancing and a Contextual Information Service for pervasive computing (see the PerCom 2003 paper.) This work was conducted as part of the Aura Project.