Disk & I/O Systems - Fundamentals

Quick Reference (TL;DR)

Disk Geometry: Tracks (concentric circles), Sectors (arcs on track), Cylinders (same track across platters). Seek time = move head, Latency = rotation, Transfer = read/write data.

Disk Scheduling: FCFS, SSTF (shortest seek), SCAN (elevator), C-SCAN (circular), LOOK (SCAN without going to end). Goal: Minimize seek time.

I/O Techniques: Buffering (temporary storage), Caching (frequently used), Spooling (queue for slow devices), DMA (Direct Memory Access, CPU-free transfer).

1. Clear Definition

Disk I/O Systems manage how data is stored on and retrieved from storage devices (hard drives, SSDs). The OS schedules disk requests to optimize performance and provides buffering/caching for efficiency.

2. Core Concepts

Disk Geometry

Physical Structure (traditional HDD):

  • Platters: Circular disks (multiple stacked)
  • Tracks: Concentric circles on platter
  • Sectors: Arcs on track (typically 512 bytes or 4 KB)
  • Cylinders: Same track across all platters
  • Read/Write Head: Moves to access tracks

SSD: No moving parts, different structure (flash memory cells).

Seek Time vs Latency vs Transfer Time

Seek Time:

  • Time to move head to correct track
  • HDD: ~3-15 ms (mechanical movement)
  • SSD: ~0.1 ms (no moving parts)

Rotational Latency:

  • Time for disk to rotate to correct sector
  • HDD: Average = half rotation = ~4-8 ms (at 7200 RPM)
  • SSD: N/A (no rotation)

Transfer Time:

  • Time to read/write data
  • Depends on: Data size, rotation speed, data density
  • Typically: ~0.1-1 ms per sector

Total Access Time = Seek + Latency + Transfer

Example (HDD):

Seek: 8 ms
Latency: 4 ms (average)
Transfer: 0.5 ms (1 sector)
Total: 12.5 ms

Disk Scheduling Algorithms

Goal: Minimize seek time (largest component of access time).

1. FCFS (First-Come-First-Served):

  • Process requests in arrival order
  • Simple, but poor performance
  • No optimization

2. SSTF (Shortest Seek Time First):

  • Always serve request closest to current head position
  • Good performance
  • Problem: Starvation (requests far away may never be served)

3. SCAN (Elevator Algorithm):

  • Head moves in one direction, serves all requests
  • Reaches end, reverses direction
  • Fair (no starvation)
  • Good performance

4. C-SCAN (Circular SCAN):

  • Like SCAN, but returns to start immediately (doesn't serve on return)
  • More uniform wait times
  • Better for systems with many requests

5. LOOK:

  • Like SCAN, but reverses at last request (doesn't go to end)
  • More efficient (less unnecessary movement)

Comparison:

AlgorithmPerformanceFairnessStarvation
FCFSPoorGoodNo
SSTFGoodPoorYes
SCANGoodGoodNo
C-SCANGoodGoodNo
LOOKGoodGoodNo

I/O Techniques

Buffering:

  • Temporary storage for data
  • Smooths out speed differences
  • Example: Keyboard input buffered before processing

Caching:

  • Store frequently accessed data in fast memory
  • Reduces disk I/O
  • Example: File system cache (RAM)

Spooling (Simultaneous Peripheral Operations On-Line):

  • Queue for slow devices (printers)
  • Process can continue while I/O happens
  • Example: Print spooler

DMA (Direct Memory Access):

  • Hardware transfers data directly to/from memory
  • CPU doesn't participate (freed for other work)
  • Much faster than programmed I/O (CPU does transfer)

DMA Process:

  1. CPU sets up DMA transfer (source, destination, size)
  2. DMA controller performs transfer
  3. DMA interrupts CPU when done
  4. CPU continues other work during transfer

Benefits:

  • CPU free during transfer
  • Faster (hardware optimized)
  • Better system performance

3. Use Cases

  • File systems
  • Database systems
  • Virtual memory (swap)
  • Logging systems

4. Advantages & Disadvantages

Disk Scheduling Advantages: ✅ Reduces seek time
✅ Improves throughput
✅ Better fairness (some algorithms)

Disadvantages: ❌ Complexity
❌ May not optimize for all workloads

DMA Advantages: ✅ Frees CPU
✅ Faster transfers
✅ Better performance

Disadvantages: ❌ Hardware complexity
❌ Cache coherence issues

5. Best Practices

  1. Choose appropriate algorithm: Based on workload
  2. Use DMA: When available
  3. Cache frequently used data: Reduce disk I/O
  4. Monitor disk performance: Identify bottlenecks

6. Common Pitfalls

⚠️ Mistake: Not understanding seek time importance

⚠️ Mistake: Ignoring starvation in SSTF

⚠️ Mistake: Not using DMA when available

7. Interview Tips

Common Questions:

  1. "Compare disk scheduling algorithms."
  2. "What is DMA?"
  3. "Explain seek time vs latency."

Key Points:

  • Seek time is largest component
  • SCAN/LOOK are good (fair, efficient)
  • DMA frees CPU
  • Buffering/caching improve performance
  • File Systems (Topic 14): How files stored on disk
  • Virtual Memory (Topic 12): Swap to disk

9. Visual Aids

Disk Scheduling (SCAN)

Tracks: 0 ────────────────────────── 199
Head:    →→→→→→→→→→→→→→→→→→→→→→→→→→→
Requests: ●     ●  ●        ●     ●
          Serve all, then reverse

DMA Transfer

CPU: Set up DMA
  ↓
DMA Controller: Transfer data
  │ (CPU free to do other work)
  ↓
DMA: Interrupt CPU when done
  ↓
CPU: Handle completion

10. Quick Reference Summary

Disk Access Time: Seek + Latency + Transfer

Scheduling: SCAN/LOOK are good (fair, efficient)

DMA: Hardware transfer, frees CPU

Buffering/Caching: Improve performance