7.5.3 Load Balancing

The implementation of load balancing is crucial since the resulting algorithm needs to be stable, which means that any failures must not lead to overloading of some parts of a computer cluster. Therefore, the implementation of the job farming module pays a lot of attention to this fact (host state detection etc.). Particularly, algorithms are designed and implemented in a way which in the worst case lead to less jobs than would theoretically be possible. Network failures, or the overloading of the host which runs SIESTA usually provoke such situations.