SAP-C02 Practice Question 46: Global UDP Traffic Entry
AWS Certified Solutions Architect – Professional (SAP-C02) · Test 5 · question 6 of 10
A studio runs the backend for a multiplayer game on EC2 instances behind Network Load Balancers in three Regions. Game clients hold a long-lived UDP session for gameplay and a TCP session for matchmaking, and they find the backend by resolving a single hostname served by Amazon Route 53 latency records with health checks.
Three problems have accumulated. When the studio drains a Region for maintenance, console clients keep connecting to it for 20 minutes or more because their resolvers and the consoles themselves cache the DNS answer. Players on some distant ISPs report jitter that tracerouting shows is caused by a poor path across the public internet before traffic reaches AWS. And several enterprise and ISP partners want to allow-list a small, fixed set of destination IP addresses in their firewalls, which the current design cannot provide.
What should the solutions architect implement?
Show answer and explanation
Answer:
A standard accelerator solves all three problems with one construct. It provides static IP addresses that are anycast from the AWS edge network and stay assigned to the accelerator for as long as it exists, which is exactly the small fixed set of addresses the partners want to allow-list. Traffic enters the AWS global network at the edge location closest to the player and is routed over that network to the optimal Regional endpoint, which removes most of the public-internet path that is causing the jitter. And because the client always talks to the same anycast addresses, a Region drain is handled inside Global Accelerator; AWS states that the service reacts instantly to changes in health or configuration to keep traffic directed to healthy endpoints, with no DNS record to propagate and no cached answer to wait out. Network Load Balancers are supported endpoints, so the UDP gameplay traffic is carried.
Option A does not fit the protocol. CloudFront is an HTTP/HTTPS content delivery network; it cannot proxy the game's UDP sessions at all.
Option C keeps DNS in the failover path, which is the cause of the 20-minute drain. TTLs are advisory, resolvers and game consoles cache beyond them, and health check quality was never the issue. It also still gives partners no stable IP addresses.
Option D pushes the failover burden into a client that ships to consoles and updates slowly, still depends on DNS re-resolution, and does nothing about the internet path quality or the allow-list.
Reference: What is AWS Global Accelerator?