Analysis of network traffic focusing on per-time-block flow statistics

摘要

Recently, a number of studies have elucidated that marginal distributions of network traffic exhibit non-Gaussian nature, and this property is crucial for modeling realistic network traffic. In this work, to study causal mechanisms that cause non-Gaussian nature of marginal distributions, we defined 'per-time-block flow', and investigated flow size, hop counts and application breakdown for each per-time-block flow and their relationships.