Our metrology research employs both passive and active network measurements. We measure both classic wired networks and new wireless environments. The passive measurement work aims at understanding traffic flows in large backbones and corporate networks. We develop techniques to identify and classify these flows with the aims of helping network management and of recognizing and responding to security threats. For instance, we have created algorithms that rapidly and reliably determine which flows are the 'elephants' (the few and large), and which are the “mice” (the many and small), and we have tackled the difficult problem of classifying those flows that fall in the middle. As another example, we have found efficient ways of estimating the traffic matrix for a network, through a combination of modeling and judicious flow sampling. In the security area, we are looking at ways of discerning attack traffic among all the packets that enter a network.
In active measurements, we consider problems relating to the topology of graphs of the internet, as well as geolocalization of hosts in the internet. Our focus in topology measurement has been on how to improve the efficiency of measurements using route traces. We are now bringing intra-domain and inter-domain routing knowledge to bear on the problem. For geolocalization, we have come up with improved methods for situating a host using pings towards the host and towards a set of reference landmarks.
Our wireless work has been informed by measurements of bit-loss rates and packet-loss rates, using such measurements as a basis for developing more effective channel coding techniques. We are also using measurements of traffic on wireless networks as a basis for modeling new traffic generators for network simulation.
- Bruno Baynat
- Timur Friedman
- Renata Teixeira