Multi-region deployment на собеседовании системного аналитика

Зачем multi-region

Latency. Users глобально — closer region = faster.

DR. Region-wide failure. Other region survives.

Compliance. Data residency (EU данные в EU, RU данные в RU).

Capacity. Distribute load.

Active-passive

One region active. Other standby. Traffic switches при failover.

Pros: simpler, traditional DR.

Cons: standby resources idle, RTO measured в minutes.

Region A (active) ←→ replication → Region B (passive).
Failure A → DNS / LB switch к B.

Active-active

Both regions serve traffic.

Pros:

• All capacity used.
• Lower latency (closest region).
• Faster failover.

Cons:

• Data consistency complex (multi-master writes).
• Conflict resolution.

User EU → EU region.
User US → US region.
Cross-region replication для shared data.

Data replication

DB cross-region.

• Async replication (default). Eventual consistency.
• Sync (rare across regions — too slow).

Conflict resolution.

• Last-write-wins.
• Application-level merge.
• CRDT-style.

Storage replication. S3 cross-region replication. Etc.

Tradeoffs

Cost. 2× infrastructure. Plus cross-region traffic.

Complexity. Operations, debugging, deployments.

Consistency. Strong cross-region — slow / impossible. Eventual — application aware.

В РФ multi-region — Yandex Cloud (несколько zones), VK Cloud. Рассматривают cross-country DR.

Связанные темы

• CAP теорема для SA
• Eventual consistency для SA
• Capacity planning для SA
• SLA SLO SLI для SA
• Подготовка к собесу системного аналитика

FAQ

Это официальная информация?

Нет. Статья основана на cloud architectural practices.

Тренируйте системный анализ — откройте тренажёр с 1500+ вопросами для собесов.