 Time slice

A time slice is an interval of CPU time that a scheduler allows a runnable operating-system process or thread before it may preempt it to serve another. In round-robin scheduling, runnable tasks take repeated turns using a scheduling quantum; a task can block or finish before consuming its whole turn.^[ref]

If three continuously runnable tasks share one CPU with quantum $q$ and negligible switching cost, each receives about one third of its time and waits about $2q$ between turns. Shortening $q$ improves responsiveness but increases the frequency of context switches. With fixed switching cost $s$ per full turn, the overhead fraction is $s/(q+s)$. This simple model explains the trade-off without assuming that real switching costs or workloads are constant.