The effects of dynamic quantum allocation and swap-time variability on central processing unit (CPU) behavior are investigated using a model that allows both quantum length and swap-time to be state-dependent random variables. Effective CPU utilization is defined to be the proportion of a CPU busy period that is devoted to program processing, i.e. the time not spent in swapping. Using this measure computational results are derived for four cases: (1) both quantum and swap-time constant, (2) varying quantum and constant swap-time, (3) constant quantum and varying swap-time, and (4) both quantum and swap-time varying. An allocation strategy that reduces the quantum as the number of tasks in queue increases proves superior to both a strategy that increases the quantum and to the common constant quantum strategy. Swap-Time variability is shown to have a more pronounced effect than quantum variability. (Author)