Abstract in English:

Results from classic queueing theory are shown to be useful for capacity planning of largescale multi-server systems. In the field of mainframe systems on-line monitoring and dynamic workload management care for the fulfilment of short-term performance goals with respect to service class specific execution velocity, wait time and wait time percentiles. The same performance measures are useful for long-term capacity management. On the basis of wellknown approximate results for G/G/m queueing models we derive analytical formulas for the calculation of (global) steady state performance measures. We illustrate the usage of these formulas and show their application in the field of mainframe capacity planning.