Global Network Optimization with AWS Global Accelerator - Low-Latency Delivery and Failover

Learn how to route traffic onto the AWS global network using Anycast IPs, design endpoint groups, and achieve failover through health checks.

About 6 min readLast updated: 2026-10-31

How Global Accelerator Works and How It Differs from CloudFront

Global Accelerator is a service that forwards user traffic to backends via the AWS global network. Two static Anycast IP addresses are assigned, and users worldwide connect to the nearest AWS edge location. From there, traffic goes through the AWS private network rather than the public internet, so you can expect stable latency and reduced packet loss. While CloudFront specializes in content caching and HTTP/HTTPS delivery, Global Accelerator provides TCP/UDP-level traffic optimization. CloudFront is suited to cases where caching static content or API responses is effective, while Global Accelerator is suited to real-time communication such as game servers and VoIP.

Endpoint Groups and Traffic Control

Endpoint groups are defined per region, and you register ALBs, NLBs, EC2 instances, and Elastic IPs as endpoints. Each endpoint group has a traffic dial (0-100%) that controls traffic distribution across regions. For example, you can configure 70% of traffic to the Tokyo region and 30% to the Singapore region. During region migration, you can gradually shift traffic to the new region by incrementally adjusting the traffic dial, enabling a safe transition. Endpoint weights (0-255) also control traffic distribution among multiple endpoints within the same region.

Health Checks and Failover

Global Accelerator continuously performs health checks on endpoints and routes traffic to healthy endpoints within seconds of detecting a failure. While Route 53 failover depends on DNS TTL and takes from tens of seconds to a few minutes to switch, Global Accelerator switches immediately through changes to Anycast routing. In a multi-region configuration, if all endpoints in the primary region become unhealthy, traffic automatically fails over to endpoints in the secondary region. The static IP addresses remain unchanged even after failover, so no DNS changes or client reconnections are needed.

Global Accelerator Pricing

Global Accelerator pricing consists of a fixed fee and data transfer charges. The fixed fee is approximately USD 0.025/hour per accelerator (about USD 18/month), plus data transfer is charged on a per-GB basis (approximately USD 0.015-0.035 per GB) depending on the source and destination regions (as of August 2026). Compared to CloudFront (no monthly fixed fee, data transfer only), Global Accelerator is more expensive for small traffic volumes due to the fixed cost. Choose Global Accelerator for workloads using protocols that CloudFront does not support, such as TCP/UDP real-time communication and game servers.

Suitable Use Cases

Global Accelerator demonstrates its benefits particularly in uses other than cached delivery of web pages. It suits cases where you want to deliver non-HTTP traffic, such as real-time game communication, voice and video calls, and communication over proprietary protocols, to users around the world with low latency. By carrying traffic from users onto the high-speed AWS backbone from a nearby edge, it achieves more stable low latency than going through the internet. It is also suitable for systems that need a static entry IP. Its value stands out in workloads that involve dynamic communication or need optimization at the TCP and UDP level.

Gradual Traffic Control

With Global Accelerator, you can finely control the proportion of traffic sent to endpoints. Using the dial that adjusts the flow for each regional endpoint group and the weights for each endpoint within a group, you can change how traffic is distributed. With this, you can carry out blue-green style switchovers that move traffic to a new version little by little, or operations that intentionally hold down the load on a specific region. If a problem occurs, you can also cut it off by bringing its proportion close to zero. Being able to control the flow of traffic dynamically helps with safe releases and load distribution.

Static IPs and Client IP Preservation

Global Accelerator provides static entry IPs that do not change. The entry IPs are kept even if you change the configuration behind them, so you can use them as stable destinations to register in firewall allowlists. Because multiple entry IPs are provided, redundancy is also ensured. You can also configure it to pass the original IP address of the connecting user on to the application behind it, so you can perform control and logging based on the access source. Being able to combine static IPs with preservation of user IPs is an important advantage for systems that require IP-based control.

Choosing Between CloudFront and Global Accelerator

Global Accelerator and CloudFront both make use of the AWS global network, but their purposes differ. CloudFront is best suited to keeping cacheable content such as images and pages at the edge and delivering it from there. Global Accelerator, on the other hand, is suited to relaying dynamic communication and non-HTTP protocols quickly and stably without caching. You can sort them out this way: CloudFront for delivering static content, and Global Accelerator for highly real-time dynamic communication or proprietary protocols. By combining them and choosing between them according to the nature of the content, you can optimize overall delivery performance.

Summary

Global Accelerator is a service that improves traffic latency and availability by leveraging the AWS global network. Its key features include stable entry points via static Anycast IPs, failover in seconds, and flexible control through traffic dials. It is particularly effective for global applications that use protocols other than HTTP.

References (Official AWS Resources)

The primary sources for this page are the official AWS website and documentation. Check the official pages below for the latest specifications and pricing.

If this page and the official documentation disagree, treat the official documentation as authoritative.

Building a Service Mesh with AWS App Mesh - Controlling and Observing Microservice CommunicationLearn how to declaratively configure canary deployments, retry policies, and mTLS encryption using Envoy sidecars, and visualize service dependencies with X-Ray integration (AWS discontinued support for App Mesh on September 30, 2026).Service Discovery with AWS Cloud Map - Dynamic Name Resolution for MicroservicesDynamically discover microservice endpoints using DNS-based and API-based approaches. Learn how to automate service registration and deregistration through integration with ECS and EKS.How CloudFront PoP Selection Works - DNS Routing and the Cache HierarchyThis article explains the DNS-based routing that CloudFront uses to route user requests to the nearest PoP, the multi-layer cache structure of edge locations and regional edge caches, and the design factors that determine the cache hit ratio.Dedicated Connection Design - Achieving Stable Private Network Connectivity with Direct ConnectThis article explains design methods for dedicated line connections using AWS Direct Connect and introduces how to achieve stable private network connectivity through the choice between dedicated and hosted connections, redundant configurations, and integration with VPC.Building Dedicated Connections with AWS Direct Connect - Redundancy Design and Traffic ControlDesign redundant dedicated connections and connect to VPCs across multiple regions with Direct Connect Gateway. This article also covers bandwidth aggregation with LAG and MACsec encryption.Elastic IP Address Management and Design - Static IP Usage and Cost OptimizationLearn about Elastic IP allocation, EC2 association, cost implications of unused EIPs, and alternative approaches to consider.Why ELB Is Split into 4 Types - Behind the Design Decisions of ALB / NLB / GWLB / CLBThis article explains why the 4 load balancer types, ALB, NLB, GWLB, and CLB (Classic), coexist instead of being unified, from the layers of the OSI reference model, their performance characteristics, and their history.Global Network Acceleration - Low-Latency Delivery with AWS Global Accelerator and CloudFrontRoute traffic onto the AWS global network early using Anycast IPs to direct clients to the nearest edge location. This guide covers ALB/NLB integration and failover design.