DNS — AWS
Service mapping
- Route 53 — managed authoritative DNS on a global anycast network; the default.
Hosted zones (public/private), health checks, and routing policies cover every
generic option:
- Simple / Weighted / Latency / Geolocation / Geoproximity / Failover / Multivalue-answer routing policies map 1:1 to the generic policies (geoproximity adds a bias knob to shift traffic between regions).
- Alias records — Route 53's apex-alias (the generic
ALIAS); point the zone apex at an ELB/CloudFront/S3/API Gateway target with no charge for alias queries. - Health checks — endpoint, calculated (combine checks), or CloudWatch-alarm based; gate any policy so unhealthy targets drop out.
- Route 53 Resolver — for hybrid/VPC DNS (inbound/outbound endpoints); not traffic steering.
- Global Accelerator — anycast IP + health-based failover that reacts faster than DNS (no TTL wait) by steering at the network layer; reach for it when DNS TTL-bound failover is too slow.
When to pick which
Route 53 alias at the apex onto an ELB/CloudFront is the standard front door. Latency routing for speed, geolocation for residency, failover for active-passive DR, weighted for canary. Use Global Accelerator instead when you need sub-TTL failover or a fixed entry IP.
Limits / things that bite (verify against current docs)
- Health-check failover is still TTL + propagation bound — Global Accelerator exists precisely to beat that.
- Latency records route by the resolver's AWS-measured latency, not the user's.
- Alias targets are limited to specific AWS resource types.
- Per-hosted-zone record counts and health-check counts have soft quotas.
Pitfalls
- Putting a
CNAMEat the apex instead of an alias record. - Expecting Route 53 failover to be instant (it is not — consider Global Accelerator).
- Geoproximity bias misconfigured, silently overloading one region.
- Single hosted zone with no secondary-provider plan for a critical property.