Models of each of the primary resources (or sources of delay) in a compute server have been constructed with the intention of providing a sufficiently realistic distributed system simulation for the evaluation of resource balancing.
The simulation is driven by the processes, which move between compute server resources based on each process's individual requirements, thus creating realistic contention for popular resources.
Remotely executing processes differ from local processes by the service time of file IO operations. Other costs that a remotely executing process may involve are remote displays, message forwarding and in some implementations, all kernel calls must be serviced on the original compute server. In order to preserve simplicity in the simulator, these costs are neglected.