Edge and 5G Computing - Achieving Ultra-Low Latency with AWS Wavelength and Local Zones
Learn about ultra-low latency computing with AWS Wavelength and Local Zones. Covers processing at the 5G network edge, proximity placement in metro areas, use cases, and when to use standard regions instead.
The Demand for Ultra-Low Latency Computing
Standard AWS Regions introduce tens to hundreds of milliseconds of latency from end users, but some applications cannot tolerate this delay. Game streaming (cloud gaming) requires latency below 20 milliseconds for a comfortable experience. In AR/VR applications, latency causes motion sickness. For autonomous vehicle inference, even a few milliseconds of delay directly impacts safety. AWS offers two edge computing options to address these ultra-low latency requirements: Wavelength and Local Zones. Wavelength places AWS compute within 5G carrier networks so that traffic from 5G devices reaches AWS directly without traversing the internet. Local Zones place AWS infrastructure in major metro areas, enabling processing close to end users in those cities.
How Wavelength Works
A Wavelength Zone is AWS infrastructure installed within a carrier's 5G network. Traffic from 5G devices is processed within the carrier network, eliminating the need for internet hops and achieving single-digit millisecond latency. In Japan, AWS partners with KDDI, with Wavelength Zones in Tokyo and Osaka. Wavelength Zones support EC2 instances, EBS volumes, and VPC subnets. You deploy by creating a carrier gateway in the parent Region's VPC (e.g., ap-northeast-1) and adding a Wavelength Zone subnet. Instances in the Wavelength Zone can access regional services like S3, DynamoDB, and RDS, enabling architectures that combine edge processing with regional data storage.
Local Zones and When to Use Each
Local Zones are AWS infrastructure placed in major metro areas (over 30 cities including Los Angeles, New York, Chicago, and Dallas), providing sub-10-millisecond latency to end users in those cities. While Wavelength is specialized for 5G devices, Local Zones deliver low latency to all users in a city regardless of wired or wireless connectivity. More services are available in Local Zones than in Wavelength Zones, including EC2, EBS, ECS, EKS, RDS, and ElastiCache. As a guideline, use Wavelength when you need ultra-low latency (single-digit milliseconds) from 5G devices, Local Zones when you need low latency (sub-10 milliseconds) for users in a specific city, and standard Regions for everything else. In Japan, Local Zones are not yet available, but Wavelength Zones (KDDI) are operational in Tokyo and Osaka. To learn about low-latency design systematically, related books (Amazon) are a useful reference.
Wavelength and Local Zones Pricing
EC2 instance pricing in Wavelength Zones is comparable to standard Regions, but available instance types are limited. Data transfer through the carrier gateway costs approximately $0.05 per GB. EC2 pricing in Local Zones tends to be about 10-20% higher than in Regions. Both options offer fewer services than full Regions, so the cost-effective approach is to place only latency-sensitive components in Wavelength/Local Zones and process everything else in the Region.
Summary - Guidelines for Edge Computing
AWS Wavelength and Local Zones are edge computing options for applications that require ultra-low latency. Wavelength enables single-digit millisecond processing within 5G networks, while Local Zones provide sub-10-millisecond processing in major metro areas. Consider these for use cases where latency directly impacts user experience or business value, such as game streaming, AR/VR, real-time video analytics, and financial trading. When standard Region latency is sufficient, using Regions is recommended for their broader service availability and lower cost.