Self-organized criticality in a computer network model

We study the collective behavior of computer network nodes by using a cellu lar automaton model. The results show that when the load of network is cons tant, the throughputs and buffer contents of nodes are power-law distribute d in both space and time. Also the feature of 1/f noise appears in the powe r spectrum of the change of the number of nodes that bear a fixed part of t he system load. It can be seen as yet another example of self-organized cri ticality. Power-law decay in the distribution of buffer contents implies th at heavy network congestion occurs with small probability. The temporal pow er-law distribution for throughput might be a reasonable explanation for th e observed self-similarity in computer network traffic.