Dynamic load balancing for simulations of biological aging

The efficient usage of parallel computers and workstation clusters for biologically motivated simulations depends first of all on a dynamic redistribution of the workload. For the development of a parallel algorithm for the Penna model of aging we have used a dynamic load balancing library, called PLB. It turns out that PLB manages a nearly balanced load situation during runtime taking only a low communication overhead. We compare different architectures like parallel computers and nondedicated heterogeneous networks, and give some results for large populations.