Characterization of Probabilistic Structure of Internet Traffic During COVID-19: A Study Based on MAWI Data

The COVID-19 pandemic greatly affected all aspects of human life including operations of offices, businesses, industries and educational institutions. With an increased shift to online work, changes in Internet traffic characteristics are inevitable. This paper presents a study based on MAWI data characterizing traffic in terms of multimodal and unimodal probability distributions. It is found that a transition occurs from multimodal Gaussian mixture model of byte and packet counts in the pre-COVID period to unimodal Laplace distribution during COVID period. Further, it is observed that the preponderance of small and large packets prior to COVID changes to small size packets during COVID period. These findings are likely to be useful to Internet service providers in managing future traffic. A rationale for the emergence of unimodal Laplace distribution is also provided.