next up previous contents
Next: 6.1 Parallel Computation Up: Dissertation Previous: 5.5.3 Fault Tolerance

6. Parallel and Distributed Simulation

Computing power is a vital resource for the extensive application of TCAD simulation. Moreover, the efficient utilization of the computational resources is almost as important as their existence. Excess computing power can often not be tapped due to lacking functionality of the software components involved in a simulation. The shortcomings in this respect exist in two senses. For the first, the employed algorithms are not capable of doing their computations in parallel, which means they can not compute on more than one processor at a given time and thus excess processor, e.g. workstations, although available, can not be used. The second reason which inhibits parallel and distributed computation, is the lack of a middle-ware which cares for the problems that arise from parallel computation, namely synchronization, load balancing, or error handling, to name some of them. SIESTA puts strong emphasis on these issues. It's job-farming facilities offer a very efficient front-end for parallel computation on a cluster of heterogeneous workstations.

Rudi Strasser