The Problem First

You deployed your app in one region (say India). Global users have to reach it over the public internet — which means many unpredictable hops through routers before reaching your server.

More hops = more latency. And public internet congestion makes it even worse.

We want traffic to get onto the fast, private AWS network as early as possible.

What is an Edge Location Server?

Before understanding Global Accelerator, this is important to be clear on:

• Origin server = your actual server (EC2, ALB, S3) — where real files and app logic live. One place, one region.
• Edge location server = a cache/relay server sitting physically close to users in cities worldwide. Hundreds of them globally.

Edge Location Server:
- sits near users (city level, worldwide)
- either stores cached content (CloudFront)
- or receives traffic and forwards it over private AWS network (Global Accelerator)

Users always talk to the nearest edge location first — not directly to your origin. That is what makes both CloudFront and Global Accelerator fast.

Unicast IP vs Anycast IP

This is the core concept behind how Global Accelerator routes users to the nearest edge.

Unicast IP — one server, one unique IP. Each server has a different address.

Server A = 12.34.56.78
Server B = 98.76.54.32
Client must know which specific IP to connect to

Anycast IP — multiple servers share the same IP. Network automatically routes the client to the nearest one.

Server A = 12.34.56.78
Server B = 12.34.56.78   (same IP on both)
Client connects to 12.34.56.78 → routed to nearest server automatically

Global Accelerator gives your app 2 static Anycast IPs. Users worldwide connect to the same 2 IPs and automatically get routed to the nearest edge location — no DNS trickery, no changing IPs.

How Global Accelerator Works

HOPS → Each hop is a router that receives your data packet and forwards it to the next router on the path. More hops = more time spent at each relay point = more latency.