Characteristics of successive loss process

We discuss two characteristics of successive packet loss. Firstly, we show a method of providing a search range of the order for a given Markov chain from investigations of successive packet loss. The process of packet loss is generally modeled as low ordered Markov chains. However, we have to set a proper range to evaluate which orders is better as an optimal order of a given Markov chain. Our method can reduce exceeding the calculation or states of Markov chain and we can find an optimal order in the search range applying our experimental data. Next, we also find one model for the process of successive packet loss. This model consisted of three timed-transitional states of an m-th order Markov chain (m=0, 1, 2). Accordingly, modeling the process for successive packet loss, two long-lived durations such as "quiescent" and "burst" on time scales of hours exist. We propose another new model that categorizes the part of the "burst" more precisely as two parts of low ordered Markov chains.