High Availability, Scaling and Disaster Recovery
HA design combines multi-AZ deployment, load balancing, and auto scaling. A candidate must understand scaling groups: scaling configuration (instance template), scaling rules (target tracking or step), and the lifecycle (scale out/in, cooldown, health check). SLB integrates with scaling groups to add/remove backends. Multi-AZ means distributing instances across zones and using SLB health checks to fail over.
DR concepts: RPO (data loss tolerance) and RTO (downtime tolerance) drive the strategy. Snapshots, OSS cross-region replication, and multi-region deployments provide different RPO/RTO. Backup vs replication vs failover are distinct.
Traps: assuming auto scaling reacts instantly (cooldown and warm-up delay); forgetting that a scaling group needs a health check to replace unhealthy instances; confusing RPO and RTO; assuming cross-region replication is synchronous (it is asynchronous, so RPO > 0). Also, a single-AZ database undermines a multi-AZ app tier.
Rule of thumb: map requirements to RPO/RTO first - low RTO needs warm standby/failover; low RPO needs frequent snapshots or synchronous replication.
Worked example: Traffic spikes but scaling is slow - the cooldown period and instance warm-up delay the new instances becoming healthy in SLB.