We suggest a model for data losses in a single node of a packet-switched network (like the Internet) which reduces to one-dimensional discrete random walks with unusual boundary conditions. The model shows critical behavior with an abrupt transition from exponentially small to finite losses as the data arrival rate increases. The critical point is characterized by strong fluctuations of the loss rate. Although we consider the packet arrival being a Markovian process, the loss rate exhibits non-Markovian power-law correlations in time at the critical point. |