What it is: Designing so that a failure in one component (or one AZ) does not break the whole system.

What it’s for:

• Reduce impact and recover faster.
• Improve reliability without overbuilding everything.

How to apply:

• Spread across multiple AZs.
• Use separate subnets per AZ.
• Use load balancers and health checks.
• Avoid single points of failure.

Exam cues:

• “prevent a single failure from taking down the app” → fault isolation (often Multi-AZ + redundancy).

Hard words:

• *isolation* /ˌaɪsəˈleɪʃn/: cô lập
• *redundancy* /rɪˈdʌndənsi/: dự phòng
• *single point of failure* /ˌsɪŋɡl pɔɪnt əv ˈfeɪljər/: điểm lỗi duy nhất