Parallelism and Efficiency in Discrete-Event Simulation

Discrete-event models depict systems where a discrete state is repeatedly altered by instantaneous changes in time, the events of the model. Such models have gained popularity in fields such as Computational Systems Biology or Computational Epidemiology due to the high modeling flexibility and the possibility to easily combine stochastic and deterministic dynamics. However, the system size of modern discrete-event models is growing and/or they need to be simulated at long time periods. Thus, efficient simulation algorithms are required, as well as the possibility to harness the compute potential of modern multicore computers. Due to the sequential design of simulators, parallelization of discrete event simulations is not trivial. This thesis discusses event-based modeling and sensitivity analysis and also examines ways to increase the efficiency of discrete-event simulations and to scale models involving deterministic and stochastic spatial dynamics on a large number of processor cores.