AWS Global Accelerator

The aim of Global Accelerator is to let traffic enter the AWS network as soon as possible.

While CloudFront copies data closer to the user, Global Accelerator moves the user closer to your service by providing a very high performance path to it. It aims at reducing the route via the public internet as short as possible. IT does no caching.

Global Accelerator, once traffic enters the AWS network, routes traffic to the closest infrastructure of your application. If you have servers in multiple regions, traffic within the AWS network will be routed to the closest entrypoint to your application.

It is a Layer 4 product: it works on any TCP/UDP connection (including HTTP/S connections but also MQTT, VoIP, …​).

Global Accelerator uses Anycast IP addresses, edge locations supporting Global Accelerator are assigned the same static IP address. Routers are aware of the anycast nature of those IPs and will route traffic to the closest edge location holding those IPs.x

It works with:

Traffic for standard accelerators is routed to endpoints based on the health (it uses health ckecks) of the endpoint along with configuration options that you choose, such as endpoint weights.
Health checks allow an automated failover within 1 minute from an unhealthy report (⇒ Disaster Recovery).

You get a pair of static IP anycast addresses that your clients can use for whitelisting.

DDoS protection is available thanks to AWS Shield

While Standard Accelerators follow AWS routing logics you can use Custom Accelerators to use custom application logic to direct one or more users to a specific destination and port among many, while still gaining the performance benefits of Global Accelerator (VoIP or game servers that make location their primary criteria for routing traffic for example).