# Expected value of next CPU burst using exponential averaging

The burst time is needed for Shortest Job First (SJF) and Shortest Run Time First (SRTF) scheduling. To get the approximate burst time, we use the equation $$\tau_{n + 1} = t_n + (1 - \alpha)\tau_n$$

I want to ask whether $$\tau_{n + 1}$$ is the predicted burst time of the $$n + 1$$ th process or it is the predicted burst time of some process say p which is demanding the CPU for the $$n + 1$$ th time.

• Given that you have observed $n$ processes so far, $\tau_{n+1}$ is the $n+1^{th}$ process's expected burst time. – RandomPerfectHashFunction Oct 13 '19 at 13:20
• @RandomPerfectHashFunction I doubt it. What benefit does predicting the burst time of $n + 1^{th}$ process gives us? – FrackeR011 Oct 13 '19 at 13:30
• But if we can predict the next burst time of every process which we have encountered so for, we can schedule the one with smallest predicted burst time. – FrackeR011 Oct 13 '19 at 13:32
• I think this answer explains it better link. – RandomPerfectHashFunction Oct 13 '19 at 14:05