The conjugate gradient method for queueing networks (Q1899336)

The author advocates using the conjugate gradient method to compute the stationary distribution for non-Jackson queueing networks. The corresponding networks are characterised by \(n \times n\) matrices taken from the class \(\mathbf Q\) of matrices \({\mathbf A} = (a_{ij})\) such that \(a_{ij} \leq 0\), \(i \neq j\), and \(\sum a_{ij} = 0,1 \leq j \leq n\), with stationary distribution satisfying \({\mathbf A}{\mathbf p} = 0\). The author argues that the numerical results guarantee the convergence of the algorithm and demonstrates its power by an example.