A matsayin tech lead, ta yaya za ka shirya wa ƙaruwar masu amfani sau goma?

Question

IT Interview Practice · Accepted Answer

A matsayina na tech lead, zan fara da mayar da **“users sun ninka sau goma” zuwa yanayin amfani da ake tsammani, manufofin service da shirin da ƙungiya za ta iya aiwatarwa**. Yawan accounts da aka yi rajista kaɗai ba ya nuna requests na lokacin cunkoso.

A ce kasuwanci yana shirin campaign ga tsarin oda. Tambayar ita ce ko customers za su iya duba kaya, saya da samun confirmation yayin campaign, cikin lokaci da budget da ake da su.

## 1. Mayar da growth zuwa workload

Tambayi product lokacin da users za su zo da ayyukan da za su yi. Ƙaruwa a hankali cikin shekara ta bambanta da campaign da ke kawo kowa lokaci guda.

Misali, accounts 100,000 da 1% active a lokaci guda suna nufin concurrent users 1,000. Da 5%, users 5,000 ne. Hasashen misali ne, ba default ratios ba. Har yanzu a kimanta yawan requests da aikin kowace request.

Rubuta nauyin da campaign zai haifar:

- Requests na browsing da search, ciki har da kaya masu shahara.
- Writes na oda da payment a peak times.
- Ƙarin storage da tarihin da ake riƙewa.
- Background work kamar notifications da reports.

Hakan yana ba mu workloads da za a gwada maimakon umarnin ninka servers sau goma kawai.

## 2. A amince da abin da dole ya ci gaba da kyau

Bayyana response time, error rate da completion da aka yarda da su ga muhimman journeys. Bayyana freshness da bukatar recovery inda suka dace. Jinkirin report na iya halatta, amma checkout failure ba ya halatta.

```text
[Campaign assumptions] -> [Expected workload] -> [Service targets]
                                                   |
                                            [Acceptance tests]
```

Manufofin suna haɗa aikin fasaha da sakamakon customers. Suna taimaka wa tantance ko ingantawa ta isa da irin degradation da za a iya jurewa yayin matsin lamba.

## 3. Auna iyakar yanzu kafin zaɓar canji

Yi amfani da production observations da load tests masu wakilci. Duba slow queries, database connections, CPU, memory, queue age da quotas na external services. Haɗa traffic mai karkata zuwa kaya masu shahara maimakon random requests kawai.

Idan database connections ne iyaka, ƙarin application nodes na iya ƙara contention. Idan quota na payment provider ne iyaka, application mai sauri ba zai ƙara wannan quota ba.

Gwada cold caches da rasa instance tare da yanayin aiki na yau da kullum. Muna buƙatar capacity don recovery yayin campaign, ba don gwaji mai cikakkiyar yanayi kawai ba.

## 4. Mayar da binciken zuwa aikin da ke da masu alhaki

| Abin da aka gano | Matakin da za a tantance | Hujjar da ake bukata |
| --- | --- | --- |
| Expensive reads masu maimaituwa | Gyara query ko cache sakamakon da ya dace | Rage aikin database tare da freshness da aka yarda da ita. |
| Background tasks suna gasa da sayayya | Takaita ko ware capacity ɗinsu | Checkout yana lafiya yayin background load. |
| Dependency quota bai isa ba | Nemi ƙari ko sarrafa admission | Quota da aka tabbatar da gwajin halin a iyaka. |
| Capacity bai isa ba bayan ƙananan gyare-gyare | Tantance scaling ko babban canjin design | Amfani da aka auna, migration plan da kuɗi. |

Ba wa kowane aiki owner, budget, dependencies, due date da acceptance test. Haɗa deployment, koyar da operation da shirin on-call, ba implementation kawai ba.

Ba da fifiko bisa tasirin customer da haɗarin delivery. Babban rewrite ko sharding yana buƙatar dalili da aka nuna da isasshen lokacin migration da verification.

## 5. Yi rehearsal na campaign sannan ƙara traffic a hankali

Saki changes a ƙananan matakai masu sarrafuwa kuma gwada journeys tare. Bayyana lokacin ƙara traffic, dakatar da ƙari ko rollback. Canjin data format na iya buƙatar forward recovery maimakon rollback mai sauƙi.

Shirya controlled degradation: jinkirta optional exports, takaita background work ko ƙin ƙarin requests tare da kare critical operations. Gwada cewa waɗannan suna rage load da gaske kuma customers suna samun sakamako bayyananne.

Idan target bai dace da lokaci ko budget ba, ba product zaɓuɓɓuka bayyanannu: kawo users a matakai, rage expensive feature ko ba measured bottleneck kuɗi. Bayyana tasirin customer da rashin tabbas na kowane zaɓi.

**Ka tuna:** growth assumptions → service targets → measured limits → roadmap mai owners → rollout da aka gwada. Tech lead yana bayar da capacity plan da ya haɗa mutane, kuɗi da recovery.

Abin da aka gano	Matakin da za a tantance	Hujjar da ake bukata
Expensive reads masu maimaituwa	Gyara query ko cache sakamakon da ya dace	Rage aikin database tare da freshness da aka yarda da ita.
Background tasks suna gasa da sayayya	Takaita ko ware capacity ɗinsu	Checkout yana lafiya yayin background load.
Dependency quota bai isa ba	Nemi ƙari ko sarrafa admission	Quota da aka tabbatar da gwajin halin a iyaka.
Capacity bai isa ba bayan ƙananan gyare-gyare	Tantance scaling ko babban canjin design	Amfani da aka auna, migration plan da kuɗi.

1. Mayar da growth zuwa workload

2. A amince da abin da dole ya ci gaba da kyau

3. Auna iyakar yanzu kafin zaɓar canji

4. Mayar da binciken zuwa aikin da ke da masu alhaki

5. Yi rehearsal na campaign sannan ƙara traffic a hankali

Tambayoyi masu alaƙa