Apa sabenere regane context switch, lan kenapa iku penting?

Question

IT Interview Practice · Accepted Answer

**Context switch** iku kernel nyimpen state CPU thread sing lagi mlaku (register, program counter, stack pointer) lan mbalekake state thread liyane, saengga siji core bisa multiplex akeh thread. Iki kelakon ing saben preemption scheduler, blocking syscall, utawa interrupt — lan iku ora gratis.

## Ing ngendi biaya ndhelik

```text
DIRECT cost (visible, ~1–5 µs):
  save/restore registers · run scheduler logic · switch kernel stacks
  if PROCESS switch: also swap page tables (CR3) → FLUSHES the TLB

INDIRECT cost (often much larger):
  cold CPU CACHES → the new thread's data isn't in L1/L2/L3 → cache misses to RAM
  cold TLB        → address translations must be re-walked
  lost pipeline / branch-predictor state
```

Biaya **langsung** iku sawetara mikrodetik. Biaya **ora langsung** — cache lan TLB sing adhem sawise switch — kerep dadi pukulan sing luwih gedhe: thread sing mlebu mlaku alon sedhela nalika ngangeti maneh cache-ne saka memori utama (puluhan nganti atusan siklus saben miss).

## Switch thread vs proses

Ngoper antara **thread ing proses sing padha** luwih murah: dheweke nganggo bareng ruang alamat, mula page table (lan TLB, ing perangkat keras tanpa tagged TLB / PCID) ora perlu owah. Ngoper **proses** kajaba iku uga ngunggah maneh register basis page-table lan sacara tradisional nge-flush TLB — iki sebabe regane luwih akeh.

## Kenapa penting ing praktik

- **Kakehan thread** → scheduler thrash, ngentekake siklus kanggo switch tinimbang kerja; ngliwati titik tartamtu, thread luwih akeh tinimbang core *ngurangi* throughput.
- Iki argumen inti kanggo model **async / event-loop** (Node, nginx, `io_uring`) lan **goroutine / green thread**: nangani akeh koneksi nganggo sithik OS thread, ngindhari switch saben request.
- Iki sebabe **CPU pinning / affinity** mbiyantu service sing sensitif latency — njaga thread ing siji core kanggo ngreksa cache-ne sing anget.
- Cacah context-switch ora sengaja (involuntary) sing dhuwur (katon ing `vmstat`, `pidstat`) iku gejala klasik saka lock contention utawa CPU oversubscription.

## Kenapa iki penting

Iki pitakonan level senior amarga misahake wong sing ngerti *istilahe* saka wong sing mangerteni *kenapa* iki larang — lan jawaban sing tenanan (polusi cache/TLB, ora mung nyimpen register) nemtokake keputusan arsitektur: async I/O tinimbang thread-per-request, ngukur thread pool cocog karo cacah core, lan nganggo affinity. Iki nyambungake scheduling, memori virtual, lan cache CPU dadi siji crita performa.

Ing ngendi biaya ndhelik

Switch thread vs proses

Kenapa penting ing praktik

Kenapa iki penting

Pitakon sing gegandhengan