INFRASTRUCTURE LEARNINGINFRASTRUCTURE LEARNING
インフラ大全インフラ大全IT INFRASTRUCTURE / LEARNING
Infrastructure learning guide
Find a lesson or lab/

Esc to close · Search terms stay on your device

LEARNING日本語

Topic lessonTopic: Audit coverage and recovery design

AWS fault domains: choosing between Multi-AZ and Multi-Region

Start with what can fail together. Trace a fictional image submission through compute and storage to see which business routes survive a failure.

What you will learn

Trace dependencies for the business transaction and evaluate which routes survive an AZ-wide or Region-wide outage.

Prerequisite

This is a fixed architecture exercise. The scenario does not estimate actual AWS recovery time or interpret service guarantees.

Find the dependencies shared by redundant paths

AWS provides isolation across Availability Zones and, when necessary, across Regions. Whether a surviving path completes the transaction still depends on placement, capacity and shared dependencies.

Redundant entry paths can share a single dependency
Redundant entry paths can share a single dependency1: Worker A — A uses shared data — Shared database. 2: Worker B — B uses shared data — Shared database. 3: Shared database — Check the outcome — Business check123Worker AWorker BShared databaseBusiness checkRedundant entry paths can share a single dependency1: Worker A — A uses shared data — Shared database. 2: Worker B — B uses shared data — Shared database. 3: Shared database — Check the outcome — Business check123Worker AWorker BShared databaseBusiness check
  1. Worker A → Shared databaseA uses shared data
  2. Worker B → Shared databaseB uses shared data
  3. Shared database → Business checkCheck the outcome
Worker A
East Region, AZ A
Worker B
Same Region, AZ B
Shared database
Both use storage in AZ A
Business check
Submit and read back

An entry point survives but the write path does not

A submission needs an entry point, a worker and a database write
OutageWhat remainsBusiness status
AZ B unavailableWorker A and databaseMay continue if A can handle the load
AZ A unavailableWorker BCannot write to shared database
East Region unavailableOnly the static page is in WestSubmissions cannot continue
Surviving capacity unmeasuredA complete route existsPeak-load operation remains unverified

A second Region does not remove a dependency on the original Region when, for example, the write target or a startup dependency remains there.

Choose isolation that matches the requirement

Name the business outcome. List viewing, submitting and delayed processing separately. Give each outcome its own observable pass condition.

Follow the whole route. For each failure scope, inspect entry points, compute, data, identity and external dependencies. Check the capacity of the paths that remain.

Record deployment and capacity evidence. Keep normal synthetic results separate from the measured capacity of the surviving system during a fault.

Check your decision

Choose an option to see the result and explanation.

Why does losing AZ A stop submissions?
Read the answer and reason

Answer: The shared database is unavailable · Worker B still needs the database in AZ A.

West only hosts a static page. Can submissions continue there?
Read the answer and reason

Answer: Check compute and data separately · There is no path to complete the write.

A complete route survives but has no peak-load evidence. What is its status?
Read the answer and reason

Answer: Peak-load continuity is unverified · Route survival and capacity are distinct outcomes.

Check, then explain in your own words

  1. Define the business outcome.
  2. Trace dependencies for each failure scope.
  3. Measure surviving capacity and completed transactions.
Check your understanding: Why do compute resources in two AZs leave availability uncertain?

Both paths may depend on the same data or service.

Availability depends on isolation of the resources needed to complete the work.

Sources and conditions checked

The official references used in this lesson are collected here with their scope, conditions and corresponding sections. Independent expert review and live AWS comparison have not been performed.