Characterizing On/Off Pattern of Broadband Internet Users

It is well known that the duration of Internet user's sessions grows with the increase of broadband Internet access. Nevertheless, it is still not clear how much their usage increases, that is, whether the user issues requests during the whole session or there are periods during which users request nothing. This work presents a characterization of on/off pattern of broadband Internet users. We determine the statistical distribution of session on/off time and found that on-time is fits better to the Weibull distribution and offtime to the Lognormal. Furthermore, we model the on/off pattern with a state transition graph called CBMG (Customer Behavior Model Graph).