AWS Global Accelerator handles traffic distribution during endpoint failures through continuous health monitoring and automatic failover mechanisms designed to maintain high availability and optimal performance.
**Health Monitoring and Detection
Global Accelerator continuously performs health checks on all configured application endpoints using TCP, HTTP, and HTTPS protocols. These health checks ensure that only healthy endpoints receive traffic. If an endpoint fails these health checks and is deemed unhealthy, Global Accelerator immediately detects this change in status[6][10][13].
**Automatic Failover and Traffic Rerouting
Upon detecting an unhealthy endpoint, Global Accelerator automatically reroutes user traffic away from the failing endpoint to other healthy endpoints. This failover is seamless and typically occurs within about 30 seconds after the health check failure is detected[9][10][13]. The rerouting prioritizes endpoints that are closest to the user and are healthy, leveraging the AWS global network to minimize latency and improve reliability[5][7].
**Failover Logic Across Endpoint Groups and Regions
Global Accelerator organizes endpoints into endpoint groups, often spanning multiple AWS Regions. If all endpoints in a primary endpoint group become unhealthy or have a traffic weight of zero, Global Accelerator attempts to fail over to healthy endpoints in other endpoint groups, ignoring traffic dial settings during this failover attempt. If no healthy endpoints with a weight greater than zero are found after checking the three closest endpoint groups, Global Accelerator "fails open" by routing traffic to a random endpoint in the closest endpoint group to the client, even if it is unhealthy or has a zero weight[6][7].
**Traffic Control Features
Users can configure traffic dials to control the percentage of traffic routed to each endpoint group and weights to distribute traffic among endpoints within a group. During failover, these settings are overridden to ensure traffic is directed to healthy endpoints regardless of dial or weight settings, ensuring availability[6][9][12].
**Maintaining Client Affinity
For stateful applications requiring session persistence, Global Accelerator can maintain client affinity by consistently routing requests from the same user to the same endpoint, even during failover scenarios, until connections are reset[6].
In summary, AWS Global Accelerator ensures resilient traffic distribution during endpoint failures by continuously monitoring endpoint health, instantly rerouting traffic to the nearest healthy endpoints across regions, overriding traffic controls if necessary to maintain availability, and supporting client session persistence to provide a consistent and reliable user experience.
Citations:
[1] https://repost.aws/knowledge-center/global-accelerator-unhealthy-endpoints
[2] https://aws.amazon.com/global-accelerator/faqs/
[3] https://repost.aws/knowledge-center/global-accelerator-failover-different-region
[4] https://docs.aws.amazon.com/global-accelerator/latest/dg/introduction-benefits-of-migrating.html
[5] https://digitalcloud.training/aws-global-accelerator/
[6] https://docs.aws.amazon.com/global-accelerator/latest/dg/about-endpoints-endpoint-weights.unhealthy-endpoints.html
[7] https://aws.amazon.com/blogs/networking-and-content-delivery/deploying-multi-region-applications-in-aws-using-aws-global-accelerator/
[8] https://cloudonaut.io/review-aws-global-accelerator-latency-multi-region-disaster-recovery/
[9] https://docs.aws.amazon.com/global-accelerator/latest/dg/about-endpoints.avoid-connection-collisions.html
[10] https://hands-on.cloud/aws-services/global-accelerator/
[11] https://docs.aws.amazon.com/global-accelerator/latest/dg/about-endpoint-groups-health-check-options.html
[12] https://docs.aws.amazon.com/pdfs/global-accelerator/latest/dg/global-accelerator-guide.pdf
[13] https://cloudvisor.co/aws-guides/aws-global-accelerator/