In the past decade, many quantities characterizing high-speed telecommunication network performance have been reported to have heavy-tailed distributions, namely, with tails decreasing hyperbolically rather than exponentially. Since mixture distributions can approximate many heavy-tailed distributions with high precision, the paper uses mixture distributions to model Internet traffic and applies the EM algorithm to fit the models. Making use of the fact that, at each iteration of the EM algorithm, the parameter increment has a positive projection on the gradient of the likelihood function, the paper proposes a recursive EM algorithm to fit the models, and the Bayesian information criterion is applied to select the best model. To illustrate the efficiency of the proposed algorithm, numerical results and experimental results on real traffic are provided.