[Backup/DR] kt cloud Disaster Recovery Design: Multi-AZ and Multi-Region

Multi-AZ is the basic availability structure that distributes services across multiple Availability Zones (AZ) within a single Region. Since latency within the same Region is a few milliseconds or less, some services can aim for near RPO 0 using synchronous replication and synchronous commit.

Multi-Region, on the other hand, is a DR strategy to prepare for a full Region outage. Since inter-Region latency is tens to hundreds of ms, asynchronous replication is generally used, which means unreplicated data may remain in the event of a failure, and if two Regions perform write operations simultaneously, a conflict resolution design is needed.

In actual design, operations matter more than infrastructure. Traffic switchover may not happen immediately due to DNS TTL and caching, so a Global Load Balancer or a separate switchover procedure is needed.

• Sessions should be moved to a distributed cache or global storage rather than tied to a specific Region.
• Replication strategy should be handled differently for core transactions like orders versus logs/analytics data.
• Operating models typically converge on Active-Passive or Warm Standby.

E-commerce usually sets Multi-AZ as the default and places a Warm Standby in another Region. Finance prioritizes data integrity and uses Active-Passive Multi-Region, while global SaaS uses Geo-Partition to address regional latency and failure response together. kt cloud is progressively expanding this availability/DR structure centered on the Next architecture.