The properties of self-similar traffic have strong effects on queueing performance. It has been suggested that many existing computer and communication systems need to be reevaluated in the presence of this new type of traffic. To this end, this paper proposes an analytical performance model for adaptive wormhole routing in torus interconnection networks under self-similar traffic. The bursty characteristic of traffic over multiple time scales is captured by the superposition of a number of two-state Markov-modulated poisson processes (MMPPs). Simulation experiments demonstrate that the proposed model exhibits a good degree of accuracy for various system sizes and under different operating conditions. The analytical model is then used to investigate the impact of the important parameter of traffic self-similarity on network performance.