Course contentsShow
← Computer Science
Lesson 1428 of 2,87232. File SystemsFree lesson

Comparing Disk Scheduling Algorithms

Analyzing throughput, fairness, and variance across different scheduling strategies with examples.

Comparing Disk Scheduling Algorithms

You'll learn how to evaluate disk scheduling algorithms across throughput, fairness, and variance dimensions.

Now that you've seen FCFS, SSTF, SCAN, C-SCAN, LOOK, and C-LOOK, it's time to compare them systematically. Each algorithm makes trade-offs between throughput (total requests serviced per unit time), fairness (whether all requests get reasonable service), and variance (consistency of wait times).

Throughput

FCFS has the worst throughput—it doesn't optimize seek distance at all. SSTF maximizes throughput by always picking the nearest request, minimizing total seek time. SCAN and C-SCAN offer excellent throughput while adding predictability. LOOK variants perform similarly but avoid unnecessary edge travel.

Example: For requests [98, 183, 37, 122, 14, 124, 65, 67] starting at head position 53:

  • FCFS totals ~640 cylinders moved
  • SSTF might total ~236 cylinders (best throughput)
  • SCAN/C-SCAN typically fall between these extremes

Fairness

FCFS is perfectly fair—every request waits its turn. SSTF can cause starvation—requests far from the current head may wait indefinitely as nearby requests keep arriving. SCAN/C-SCAN and LOOK/C-LOOK guarantee bounded wait times since they sweep predictably across the disk.

Variance

FCFS has high variance—wait time depends entirely on arrival order. SSTF has unpredictable variance (some requests wait forever, others barely wait). SCAN-family algorithms have low variance because wait times are proportional to distance from the current sweep position.

Practical Choice

Modern systems often prefer LOOK or C-LOOK: they balance high throughput with fairness guarantees and predictable performance, avoiding SSTF's starvation risk while outperforming FCFS significantly.

Key Takeaway: SSTF maximizes throughput but risks starvation; SCAN/LOOK variants balance throughput, fairness, and predictable variance, making them practical choices for real systems.