AWS Global Accelerator

Exam relevance: Domain 2 (29%). Know when to use Global Accelerator vs CloudFront — a common exam trap.

1. Overview

AWS Global Accelerator improves availability and performance of applications by routing traffic through the AWS global network using anycast static IP addresses. It is not a CDN — it does not cache content.

2. How It Works

User → Nearest AWS Edge Location (anycast IP)
            └── AWS Global Network (private backbone)
                        └── Optimal endpoint (ALB, NLB, EC2, EIP)

• Two static anycast IP addresses assigned per accelerator (global, never change)
• Traffic enters the AWS network at the nearest edge location
• Routed over AWS private backbone — avoids public internet congestion
• Endpoints can be in multiple regions — Global Accelerator routes to the healthiest, lowest-latency endpoint

3. Global Accelerator vs CloudFront

Key rule: If the question mentions static IPs, TCP/UDP (non-HTTP), or instant failover → Global Accelerator. If it mentions caching or HTTP content delivery → CloudFront.

Use both in one architecture without guilt: CloudFront in front of cacheable content, Global Accelerator in front of dynamic APIs. Choosing only CloudFront for a latency-sensitive API, or only Global Accelerator for a content-heavy site, leaves the other half of the requirement unaddressed.

4. Traffic Dials and Endpoint Weights

• Traffic dial: percentage of traffic sent to an endpoint group (per region) — 0–100%
• Endpoint weight: relative weight within an endpoint group
• Use traffic dials for blue/green deployments or gradual regional traffic shifts

Accelerator
├── Endpoint Group: us-east-1 (traffic dial: 90%)
│   ├── ALB-1 (weight: 50)
│   └── ALB-2 (weight: 50)
└── Endpoint Group: eu-west-1 (traffic dial: 10%)
    └── ALB-3 (weight: 100)

5. Health Checks and Failover

• Global Accelerator performs health checks on endpoints every 10 or 30 seconds
• Unhealthy endpoints are automatically removed from routing
• Failover to healthy endpoint in same or different region in ~30 seconds
• No DNS TTL dependency — uses anycast IPs, so failover is near-instant

6. Exam Scenarios

Scenario 1: A gaming company needs low-latency UDP traffic routing to game servers in multiple regions with static IPs for firewall whitelisting. What is the solution?

Use Global Accelerator. It supports UDP traffic, provides two static anycast IPs for whitelisting, and routes traffic over the AWS private backbone to the nearest healthy game server endpoint.

Scenario 2: A company wants to perform a blue/green deployment across two regions, gradually shifting traffic from us-east-1 to eu-west-1. What is the mechanism?

Use Global Accelerator traffic dials. Set us-east-1 endpoint group to 100% and eu-west-1 to 0%. Gradually increase eu-west-1 dial while decreasing us-east-1. This shifts traffic without DNS changes or TTL delays.

Scenario 3: A company needs to serve static website content globally with caching. Should they use Global Accelerator or CloudFront?

CloudFront. Global Accelerator does not cache content. CloudFront caches static content at edge locations, reducing origin load and improving latency for HTTP/HTTPS content.

7. Exam Tips

• Global Accelerator = static IPs + TCP/UDP + non-HTTP + instant failover
• CloudFront = HTTP/HTTPS + caching + edge compute
• Global Accelerator does not cache — it only routes
• Anycast IPs mean failover does not depend on DNS TTL — much faster than Route 53 failover
• Traffic dials = regional traffic shifting; endpoint weights = within-region load distribution
• Both Global Accelerator and CloudFront include Shield Standard DDoS protection

8. Official References

• Global Accelerator Developer Guide
• Global Accelerator vs CloudFront
• Traffic Dials