AWS Global Accelerator performs instant failover through continuous health monitoring and intelligent traffic routing to maintain high availability and low latency for applications deployed across multiple AWS Regions or endpoints.
Key aspects of how instant failover works in Global Accelerator:
- Health Checks: Global Accelerator continuously monitors the health of all configured application endpoints using TCP, HTTP, or HTTPS health checks. If an endpoint becomes unhealthy, it is immediately marked as such[3][5].
- Instant Traffic Redirection: Upon detecting an unhealthy endpoint, Global Accelerator instantly stops routing new connections to that endpoint and redirects traffic to the next best healthy endpoint, which could be in the same or a different Region. This failover happens within about 30 seconds from detection of the failure[2][9].
- Routing Logic: The service routes client traffic to the closest healthy endpoint based on network latency and endpoint health. If no healthy endpoint with a positive traffic weight is available, Global Accelerator will route traffic to a random endpoint in the closest endpoint group to the client, effectively *failing open* to maintain connectivity[2].
- Connection Handling: Global Accelerator terminates client TCP connections at AWS edge locations and establishes new TCP connections to healthy endpoints almost concurrently. However, existing active connections to an unhealthy endpoint are not broken immediately; they continue until closed or timed out, after which new connections are routed to healthy endpoints[2][3].
- Traffic Dial and Weights: You can configure traffic weights and traffic dials on endpoint groups to control traffic distribution. Failover respects these settings, attempting to route traffic away from zero-weight endpoints. If all healthy endpoints have zero weight, the failover mechanism resorts to routing to the nearest endpoint group regardless of weight[2].
- Global Network Backbone: By routing traffic over the AWS global network, Global Accelerator ensures low-latency and congestion-free paths, which enhances the speed and reliability of failover[2].
- Static IP Addresses: Global Accelerator provides two static IPv4 addresses that serve as fixed entry points for client traffic. These IPs remain constant even when failover occurs, eliminating DNS caching delays and enabling near-instant failover without client-side changes[4][7].
- Failover Speed: Although marketed as "instant," failover typically completes in under a minute, with about 30 seconds needed to detect and redirect new connections to healthy endpoints. This is faster than DNS-based failover mechanisms, which can be slowed by DNS caching[9].
In summary, AWS Global Accelerator achieves instant failover by continuously monitoring endpoint health, dynamically routing new client connections to the nearest healthy endpoints via static IP addresses, and leveraging the AWS global network for optimal performance. This design minimizes downtime and provides seamless failover across Regions or availability zones without requiring client-side DNS updates or manual intervention[2][3][5][7][9].
Citations:
[1] https://docs.aws.amazon.com/global-accelerator/latest/dg/about-endpoints-endpoint-weights.unhealthy-endpoints.html
[2] https://docs.aws.amazon.com/global-accelerator/latest/dg/introduction-how-it-works.html
[3] https://tutorialsdojo.com/aws-global-accelerator/
[4] https://aws.amazon.com/blogs/networking-and-content-delivery/maximising-application-resiliency-with-aws-global-accelerator/
[5] https://cloudvisor.co/aws-guides/aws-global-accelerator/
[6] https://aws.amazon.com/blogs/networking-and-content-delivery/creating-disaster-recovery-mechanisms-using-amazon-route-53/
[7] https://aws.amazon.com/global-accelerator/features/
[8] https://stackoverflow.com/questions/62638339/how-to-improve-failover-speed-for-aws-global-accelerator-or-route53-latency-rout
[9] https://aws.amazon.com/global-accelerator/faqs/