如果您提供游戏、媒体、移动或金融等应用程序，则需要一致的、低延迟的用户体验。AWS Global Accelerator通过提高客户机和运行在AWS上的应用程序之间的互联网流量性能，帮助您实现更低的延迟。它使用AWS全球网络将TCP或UDP通信定向到离客户机最近的AWS区域中的健康应用程序端点。
GA的原理和特性
Global Accelerator为每个加速器提供两个静态IP地址，通过AWS边缘位置在全球进行广播。使用加速器，Global Accelerator将TCP或UDP流量路由到应用程序端点（如EC2实例、ALB、NLB或弹性IP地址），改善延迟、抖动、包丢失和吞吐量性能。
可以使用速度比较工具 https://speedtest.globalaccelerator.aws/ 从您自己的位置测试使用GA的性能优势。
Global Accelerator通过在边缘终止TCP，自动提供额外的吞吐量改进，包括：
- 巨型帧支持：在AWS边缘位置和应用端点之间启用巨型帧，每个数据包可多发送和接收6倍的数据
- 大型接收端窗口和TCP缓冲区：调优接收端窗口和TCP缓冲区设置，在更短时间内接收和缓冲大量数据
- 大的拥塞窗口：通过AWS全球网络传输数据，可以扩展TCP拥塞窗口，发送比通过公共互联网更多的数据
性能测试与结论
测量从世界各地的客户端到运行在AWS US East（N. Virginia）地区EC2上的端点的吞吐量，使用Global Accelerator的结果是吞吐量性能提高了60%。
GA的使用场景
为单一区域应用程序提升延迟和可用性
使用Global Accelerator，您的用户流量会离开互联网，通过80+全球边缘站点进入Amazon的私有全球网络，再导向您的应用程序源。可以将流量性能提升高达60%。
多区域应用程序的简化和弹性流量路由
Global Accelerator提供两个来自全球分布式边缘站点的静态任播IP地址，为您的应用程序提供单个切入点，无论其部署了多少个AWS区域。您可以增加或删除源、可用区域或区域，而无需降低应用程序可用性。
用于在线游戏的增强型玩家服务器联网
使用Global Accelerator沿私有AWS网络路由玩家流量，可以提升玩家在线体验，降低游戏内延迟、抖动和数据包丢失。支持Unity和Unreal等常用引擎流量加速。
在AWS上运行服务通信和IP语音
Global Accelerator通过减少呼叫设置时间来提升实时通信（RTC），同时增加呼叫成功比率和质量。
GA定价
使用AWS Global Accelerator，您需要为预置的每台加速器以及流经加速器的主方向流量付费。在账户中运行加速器时，每小时收取 0.025 USD。除了对在AWS区域中运行的应用程序终端节点收取正常的EC2数据传出费用之外，还会收取数据传输附加费。
返回技术博客
If you run applications for gaming, media, mobile, or financial services, you need a consistent, low-latency user experience. AWS Global Accelerator improves the performance of internet traffic between clients and your applications running on AWS by routing TCP or UDP traffic to healthy application endpoints in the nearest AWS region via the AWS global network.
How Global Accelerator Works
Global Accelerator provides two static IP addresses per accelerator, broadcast globally via AWS edge locations. It routes traffic to application endpoints (EC2 instances, ALB, NLB, or Elastic IPs), improving latency, jitter, packet loss, and throughput.
You can test the performance benefits using the speed comparison tool at https://speedtest.globalaccelerator.aws/.
By terminating TCP at the edge, Global Accelerator automatically provides additional throughput improvements, including:
- Jumbo frame support: Enables jumbo frames between AWS edge locations and application endpoints, allowing 6x more data per packet
- Large receive window and TCP buffers: Tuned settings allow faster buffering of large amounts of data from applications
- Large congestion window: Transmitting over the AWS global network allows a larger TCP congestion window than over the public internet
Performance Test Results
Measuring throughput from clients worldwide to an EC2 endpoint in AWS US East (N. Virginia), using Global Accelerator resulted in a 60% improvement in throughput performance.
Use Cases
Improve Latency and Availability for Single-Region Applications
With Global Accelerator, user traffic leaves the public internet and enters Amazon's private global network via 80+ edge locations, then routes to your application. This can improve traffic performance by up to 60%.
Simplified and Resilient Traffic Routing for Multi-Region Applications
Global Accelerator provides two static anycast IP addresses from globally distributed edge locations, giving your application a single entry point regardless of how many AWS regions it's deployed in. You can add or remove origins, AZs, or regions without reducing availability.
Enhanced Player-Server Networking for Online Games
Route player traffic over the private AWS network to reduce in-game latency, jitter, and packet loss. Supports popular engines like Unity and Unreal.
Real-Time Communications and VoIP on AWS
Global Accelerator reduces call setup time and improves call success rates and quality for real-time communication (RTC) applications including VoIP and video conferencing.
Pricing
You pay for each accelerator provisioned and for the dominant direction of traffic flowing through it. Running an accelerator costs $0.025 per hour. In addition to standard EC2 data transfer out charges for your application endpoints, a data transfer premium applies.
Back to Tech Blog