Summary of resource-oriented scaling models:

1. In problem constrained scaling, the problem sized is kept fixed, regardless of the number of processors on the machine.
2. In time constrained scaling, the wall-clock execution time needed to complete the program is kept fixed. The problem is scaled so that the new problem's execution time on a large machine is the same as as the old problem's execution time on a small machine.
3. In memory constrained scaling, the amount of main memory user per processor is kept fixed. The problem is scaled so that the new problem uses exactly k time as much main memory as the old problem.

With problem-constrained scaling, speedup = time by 1 processor/time by P processors

With time-constrained scaling, speedup = work done by P processors/work done by 1 processor

With memory-constrained scaling, speedup = work per unit time on P processors/work per unit time on 1 processor