What is context switching and why is it expensive?

Deep dive

Switching work adds overhead and disrupts locality:

• Save/restore registers and stack pointers.
• TLB and cache effects reduce CPU efficiency.
• Scheduler decisions add latency under load.
• Frequent switches hurt throughput for CPU-bound tasks.

Examples

High thread count scenario:

1000 runnable threads -> frequent switches -> lower cache hit rate -> slower throughput

Common pitfalls

• Creating too many threads for CPU-bound work.
• Using blocking I/O without thread pools.
• Confusing context switches with process forks (different costs).

Interview follow-ups

• How do you measure context switch overhead?
• When do you prefer async I/O over threads?
• How do caches and TLBs affect performance?