Scheduling algorithm (SCAN, LOOK) in LLD - Scalability & System Analysis

Scalability Analysis - Scheduling algorithm (SCAN, LOOK)

Growth Table: Scheduling Algorithm (SCAN, LOOK)

Users/Requests	Disk Requests per Second	Queue Length	Response Time	Resource Usage
100 requests/sec	100	Short queue	Low latency	Single disk, CPU low
10,000 requests/sec	10,000	Medium queue	Moderate latency	Disk busy, CPU moderate
1,000,000 requests/sec	1,000,000	Long queue	High latency	Disk saturated, CPU high
100,000,000 requests/sec	100,000,000	Very long queue	Very high latency	Disk overloaded, CPU maxed

First Bottleneck

The disk I/O is the first bottleneck because SCAN and LOOK algorithms optimize disk head movement but cannot increase physical disk speed. As requests grow, the disk queue lengthens, causing delays.

Scaling Solutions

• Horizontal scaling: Add more disks and distribute requests (e.g., RAID, sharding data across disks).
• Caching: Use memory caches to reduce disk reads for repeated data.
• Upgrade hardware: Use SSDs or faster disks to reduce seek time.
• Load balancing: Distribute requests evenly to avoid hotspots.
• Algorithm tuning: Use LOOK to reduce unnecessary disk head movement compared to SCAN.

Cost Analysis

At 10,000 requests/sec, disk I/O bandwidth and seek time become critical. Each request may require 5-10 ms seek time on HDDs, limiting throughput to ~100-200 requests/sec per disk.

To handle 1,000,000 requests/sec, thousands of disks or SSDs are needed, increasing cost significantly.

CPU usage grows with queue management and scheduling overhead but is usually less critical than disk I/O.

Interview Tip

Start by explaining how SCAN and LOOK reduce disk head movement to improve throughput. Then discuss physical disk limits as bottlenecks. Finally, propose scaling solutions like adding disks, caching, and upgrading hardware.

Self Check

Your disk handles 1000 requests/sec. Traffic grows 10x to 10,000 requests/sec. What do you do first?

Answer: Add more disks and distribute requests to reduce queue length and avoid disk saturation.

Key Result

SCAN and LOOK scheduling improve disk efficiency but physical disk speed limits throughput; scaling requires adding disks, caching, or upgrading hardware.