A simple solution technique that allows analysis of computer systems and is easy to implement is the Mean Value Analysis (MVA) technique.
This is an iterative technique, iterating on the number of customers. This technique has an advantage over the convolution technique in that it avoids numerical instabilities.
Mean Value Analysis gives the mean performance. This can be applied to networks with a variety of service disciplines and service time distributions. This technique can be applied to both single class and multiple class models as well. Also, load dependent and load independent models can be solved using this technique.
Mean Value Analysis is a recursive algorithm. Computation of performance with N jobs in the network requires knowledge of performance with N-1 jobs. Since performance with N=0 is trivially known, we always start the analysis with N=0 and compute the performance for N = 1,2,.... successively. For smaller values of N, this procedure is not computationally too expensive. However, for large values of N and more number of job classes, particularly if the performance for smaller values of N is not required, it would be preferable to avoid this recursion. For this purpose, several approximate analysis techniques have been developed.
The most commonly used approximation technique is one proposed by Schweitzer.