How to utilize computational resources

Computational resources could be costly. As you have to specify the number of processors for your job submission respect to content of your work, you might be confused for what kind of reason you could use one more or less processor. "More processors do a faster job" is not guaranteed in every instance and it would be silly if same number of processors could in fact finish two simulations within a week but you have used all of them to speed-up one of the simulations by only 5%. To achieve a better distribution of resources among the group, we are going to employ some of the methods used by those high-end computer centers.

Do a Benchmark Test

Simply speaking, a benchmark test is to run your job using increasing number of processors to see how it scales. Using double the number of processors does not mean they will finish the same job in half of the time. Your goal is to find the number of processors before which the finishing time scales proportionally to the number.

The figure gives a simple example of how a benchmark test for a simulation code is done. The corresponding report is attached here.