On the Variability of Internet Host Interactions
Dongjin Lee, Nevil Brownlee · 2008
This paper proposes a measurement metric to observe and analyze traffic variations at the higher abstraction level, i.e., IP host and its interactions. Each interaction is extracted for three common attributes: number of flows, size in bytes and lifetime. Applying the coefficient of variation metric for the interactions (flow, size, lifetime) of each host, our analysis shows that although many hosts (interactions) do not vary considerably, several hosts have high variations. It also shows some correlations of the variations, e.g., variations of flow counts impacts with the size. Further, two examples are examined so as to demonstrate the usefulness of our proposed metric. We believe that such variation behaviors should be of interest to both researchers and network operators.