Service Time Estimation

operating-systems scheduling

Definition

Service Time Estimation

Service time estimation predicts the next CPU burst length for a process, required for SPN, SRT, and HRRN.

Uses exponential averaging of past bursts:

$$S_{n+1}=\alpha T_n+(1-\alpha )S_n$$

• $T_n$: Actual length of most recent burst
• $S_n$: Previous estimate
• $\alpha$: Weighting factor ($0<\alpha <1$)
  • Close to 1: Responds quickly to recent changes
  • Close to 0: Stable, ignores transient spikes

Properties

Fairness

Overestimating a burst may unfairly delay a short process.

Starvation Prevention

Underestimating may allow CPU-intensive processes to monopolise the processor when shorter bursts are prioritised.