Uingizwaji wa page hufanyaje kazi, na nini husababisha thrashing?

Question

IT Interview Practice · Accepted Answer

Wakati RAM halisi inajaa lakini process zinahitaji pages zaidi, OS lazima **iondoe (evict)** page iliyopo ili kupata nafasi — ikiiandika kwenye **swap** (diski) ikiwa ni chafu (dirty). Page ipi ya kuondoa ndiyo uamuzi wa **page replacement**, na kuukosea huharibu utendaji.

## Sera za uingizwaji

Kiwango bora cha kinadharia ni kuondoa page itakayotumika mbali zaidi katika siku zijazo (**OPT ya Bélády**) — haiwezekani, kwa kuwa huwezi kuona siku zijazo. Hivyo tunakaribia **LRU (least recently used)**: ondoa page isiyotumika kwa muda mrefu zaidi. LRU halisi inahitaji timestamp kwa kila ufikiaji (ghali mno), hivyo kernels halisi hutumia **algoriti ya clock (second-chance)**:

```text
Frames arranged in a ring; each has a REFERENCE bit the hardware sets on access.
A "hand" sweeps around the ring:
   ref bit = 1 → clear it, give the page a second chance, move on
   ref bit = 0 → evict THIS page (approximates "least recently used")

[1]→[1]→[0] ← evict here   (the sweep clears 1s until it finds a 0)
         ▲___________|
```

Clock hutoa matokeo kama ya LRU kwa gharama ya O(1). Linux huiboresha zaidi kwa **orodha za LRU active/inactive**.

## Working set na thrashing

**Working set** ya process ni seti ya pages inazotumia kikamilifu ndani ya dirisha la muda. Paging hufanya kazi vizuri *maadamu working sets zilizounganishwa zinatoshea ndani ya RAM*. Zisipotoshea:

```text
working set > RAM → evict a page that's about to be needed → immediate page fault →
evict another needed page → ... → the CPU spends ~all its time paging to disk, ~0% on real work
                            = THRASHING
```

Thrashing ni jabali, si mteremko: vuka kikomo kidogo tu na throughput huanguka kwa sababu karibu kila ufikiaji hu-miss. Dalili: CPU karibu haifanyi kazi lakini disk I/O imejaa, na latency inalipuka.

## Kuirekebisha / kuiepuka

- **Ongeza RAM** au **punguza mzigo** (process chache kwa wakati mmoja) ili working sets zitoshee.
- Scheduler inaweza **kusimamisha na kutoa nje (swap out)** process nzima kupunguza mahitaji (medium-term scheduling).
- **OOM killer** ya Linux ni njia ya mwisho — ua process badala ya ku-thrash milele.
- Boresha **locality** ili working set yenyewe iwe ndogo.

## Kwa nini ni muhimu

Hili hufunga hadithi nzima ya kumbukumbu pamoja na ni swali pendwa la senior kwa sababu thrashing ni njia halisi ya kushindwa katika production: service iliyo sawa kwa watumiaji N huanguka kutoka jabali kwa N+1 pale working set yake inapozidi RAM. Kutambua alama (paging nzito, disk iliyojaa, CPU iliyokwama) na kujua vidhibiti (RAM, load shedding, locality — na kwamba swap hubadilisha uhaba wa kumbukumbu kuwa maafa ya latency) ni tofauti kati ya kuitambua kwa dakika chache na kutazama sanduku lililoganda kwa njia ya kushangaza.

Sera za uingizwaji

Working set na thrashing

Kuirekebisha / kuiepuka

Kwa nini ni muhimu

Maswali yanayohusiana