I want to experiment with the scheduler and I would like to find a
benchmark that could help me understand how the changes I make affect
the system.

This is a complex issue. For instance, if I make the process quanta
bigger, CPU-bound processes would run better, but IO-bound processes
would run worse and the system would feel less resposive. A benchmark
that simply tests floating point operations would give me the wrong

So, to test a scheduler, one needs a more comprehensive benchmark that
combines CPU-bound and IO-bound processes and measures how much work
gets done. And it has to run them at the same time, not one after the

The "responsive" part may be hard to test. One option is to get the
computer doing something heavy and then type "ls" on a terminal and
measure how long it takes to respond.

Does anyone know of a benchmark that does this sort of thing?