On Micro VAX farms and shower libraries: Monte Carlo techniques developed for the D0 detector

In order to predict correctly the effects of cracks and dead material in a nearly hermetic calorimeter, hadronic and electromagnetic showers need to be simulated accurately on a particle by particle basis. Tracking all the particles of all showers in the calorimeter leads to very large CPU times (typically 5 hours on a VAX780) for events at √(s) = 2TeV. Parametrizing the energy deposition of electromagnetic particles in showers with energy below 200 MeV results in event times of the order of 1 hour on a VAX780. This is still unacceptably large. The D0 collaboration then employed a farm of 16 MicroVax II's to get acceptable throughputs. The calorimeter hit patterns of each individual track was output, to be summed up by a later job. These individual hit patterns were entered into a random access shower library file, which was then used for subsequent Monte Carlo simulations. This shower library technique results in further speed-ups of a factor of 60 without degrading the quality of simulation significantly