What it is
AWS Global Accelerator is a networking service that improves the availability and performance of internet-facing applications. It provides two static anycast IP addresses that act as a fixed entry point to your application, regardless of where the endpoints are.
How it works
- Users connect to the nearest AWS edge location through the anycast IPs.
- From there, traffic travels over the AWS global backbone instead of the public internet for most of the path.
- Global Accelerator sends the traffic to the best healthy endpoint using health checks and routing policy.
Building blocks
- Accelerator: the resource with the two static IPs and a DNS name.
- Listener: TCP or UDP ports (or port ranges) it accepts.
- Endpoint group: one per Region, with a traffic dial to shift traffic between Regions.
- Endpoints: Application Load Balancers, Network Load Balancers, EC2 instances or Elastic IPs.
- A custom routing variation maps users to specific EC2 instances and ports, useful for gaming or VoIP sessions.
Typical uses
- Multi-Region failover without waiting for DNS TTLs to expire, because the IPs never change.
- Static IPs that customers can allow-list in their firewalls.
- Lower latency and jitter for global users, and for non-HTTP traffic (TCP and UDP).
- Blue/green and canary shifts using traffic dials.
Global Accelerator vs CloudFront
CloudFront caches content and is best for HTTP(S) delivery. Global Accelerator does not cache; it accelerates and fails over arbitrary TCP and UDP traffic and gives static IPs. They are often used together.
Watch-outs
- Billing has a fixed hourly charge per accelerator plus a data transfer premium (DT-Premium) by direction and Region pair.
- Client IP preservation depends on the endpoint type and configuration.
- It is an entry-point service. It does not connect your VPCs to each other.
Quotas, limits and pricing change often. Confirm current values in the AWS documentation before designing around them.