Investigating the sensitivity of performance measures with respect to changes in the model parameters is an important technique in the analysis of computer and communication systems. Queueing network models (QNMs) are a popular approach for the modeling and analysis of such systems. In this paper, sensitivity analysis extensions of the exact mean value analysis (MVA) algorithms for single as well as multiclass QNMs are presented. The derivatives of the performance measures with respect to model inputs are of the same recursive structure as the MVA expressions, allowing the simultaneous computation of performance measures and associated sensitivities. The presented algorithms are also discussed in the context of their computational as well as spatial complexity.