How often does the algorithm enforce partition budgets?
A 100ms averaging window does not produce information about cpu usage only once every 100ms. Rather, it stores a history of cpu usage, with detail for each of the last 100 millisecond intervals. The window rotates, or slides forward in time, every clock tick. So the window provides precise information about the average cpu consumption every millisecond (or clock period). Between clock ticks, when the AP scheduling algorithm is called, cpu usage of each partition is approximated with the assumption that each partition will likely run continuously at least until the next clock tick. In other words, AP scheduling computes cpu time used, and enforces budgets, many times per millisecond.
