There is a gap between what wireless theory professes and what wireless systems practice. For instance, it is well known in the RF community that signal strength is temporally correlated. However, many wireless networks still use MAC protocols that immediately retransmit after a failure. Other examples of a gap between theory and practice exist. Some of these discrepancies exist, in part, due to extending wired protocol design principles to wireless networks. Understanding when the comparative simplicity of wired networks can be applied to the complexity of wireless is an open and challenging problem.
  In this talk, I examine how we might measure and quantify properties of wireless networks that can help understand when and how they behave differently. I present two metrics, Beta and Kappa, that capture how packet losses are temporally or spatially correlated. Measuring Beta and Kappa on several different wireless networks, I present results showing that individual testbeds vary greatly. Examining the implications of Beta and Kappa to higher level protocol performance, I present results showing that they can have significant effects on experimental results. Together, these results suggest that Beta and Kappa can be used to guide protocol selection in a wireless network, and that using single testbed experiments may be insufficient to demonstrate general utility.

  Kannan Srinivasan completed his Masters in Oklahoma State University in 2002. After that, he worked at the National Research Council of Canada for 3 years in the beautiful Island of Cape Breton. Srinivasan came to Stanford in 2005.
  Kannan Srinivasan main research interests are in Wireless Network Measurement and Characterization. He is part of the Stanford Information Networks Group (SING). His advisor is Prof. Philip Levis.