PassiveHardwareTrafficDynamics

From GEANT2-JRA1 Wiki

1 Monitoring of traffic dynamics with hardware support

Monitoring of traffic dynamics with hardware support

This application provides a complete answer to what is used bandwidth. When we indicate used bandwith on a certain network link (or available bandwidth, as complement to installed bandwidth), we normally give this number as average bandwidth over certain time interval. Because networks naturally include peaks and drops in used bandwidth as a result of multiplexing flows from different sources, the shorter interval we use to compute average bandwidth, the higher peaks and drops we usually see. So the question is what is the right interval to compute average bandwidth? It depends on application. Network planning or performance debugging may require different time granularity for bandwidth monitoring. When we go down to packet times, the link is ultimately either 100% loaded (there is a packet at this moment) or 0% loaded (there is not a packet at this moment), which is not very useful metrics.

One way to quantify traffic dynamics is to compute histogram of sizes of back-to-back packet bursts. We can express express frequency of bursts sizes in bytes, packets or number of bursts.

We have developed two versions of BURST application that quantifies traffic dynamics: