Radio technology has progressed a great deal in recent years, and we now have high-density radio interfaces through 5G. But while these networks provide high reliability and bandwidth, there is still an ever-present choke point called latency. With applications being hosted all over the world, achieving low latency is not always possible, so the only solution is to bring the infrastructure, and the application itself, closer to end users.
In this article, we’ll review one of the latest AWS services, AWS Wavelength, and show you how Amazon is using it to bring the game to a whole new level.
AWS Wavelength: An Introduction
AWS Wavelength utilizes the high bandwidth and ultra-low latency of 5G in combination with its own services (like compute and storage) to deliver a highly capable infrastructure that can host different application types. AWS Wavelength relies on Wavelength Zones; these consist of AWS infrastructure deployments residing in the data centers of telecommunication providers, located at the edge of 5G networks.
By placing the AWS resources in this way, end users can reach the application servers without ever leaving the telecommunication provider’s network. And because the traffic doesn’t extend to the public internet, latency is kept to a minimum, allowing customers to fully utilize the advantages of 5G.
Achieving ultra-low latency for various 5G applications is certainly one of this service’s many benefits. AWS Wavelength Zones also feature a high-bandwidth secure connection to their parent AWS Region, allowing you to easily access other AWS services when needed. As an AWS customer, you can work in a familiar environment, relying on AWS tools to build, deploy, and secure your product. Additionally, you get flexibility and the ability to scale, as you can create your applications once and easily deploy them to any supported AWS Wavelength Zone.
One obvious downside is that you are severely limited to the 5G zones/regions that are available at the moment. While these regions are technically extensions of AWS, the infrastructure resides within the telecommunication providers’ data centers and the work necessary to provide worldwide coverage is extensive.
AWS Wavelength was released globally in August of this year, but due to the enormous requirements to set up and support it, availability will increase piecemeal, opening more and more regions over time.
At first, AWS Wavelength was available only in the Boston and San Francisco areas, but Amazon has partnered with Verizon to cover the entire U.S., with three more areas already covered: Washington DC, New York City, and Atlanta. Amazon is also working with providers all over the world, and Europe, South Korea, and Japan are expected to have their own AWS Wavelength support in place soon.
Amazon is working hard to achieve maximum coverage as fast as possible. But having to deploy the supporting infrastructure within the telecommunication provider’s data center and also work with different providers around the world takes time. Even with a single provider in the U.S., full coverage won’t be a reality in the coming months, while for the rest of the world, it could be years, especially for regions where telecommunications are already weak.
At the moment, AWS Wavelength allows you to create VPC subnets (and the appropriate carrier gateways), EC2 instances, and EBS volumes. There are many support services available, such as CloudFormation for deploying infrastructure, CloudTrail for logging, and EKS and ECS clusters for containerization. Your EC2 backup instances can also implement AWS Auto Scaling groups, Systems Manager, and CloudWatch. Databases like RDS and DynamoDB are not available in the AWS Wavelength Zones, but due to the high-speed connectivity between them and their parent AWS Region, they can be easily accessed. This also applies to S3 buckets, which are one of the most used resources on the AWS cloud.
For cost monitoring, AWS Cost Explorer is supported.
How Networking Works
When working with AWS Wavelength, you are expanding your VPC, so all of your EC2 instances will be part of your actual AWS Parent Region.
Carrier gateways are a new network component that provides connectivity from the AWS Wavelength Zone to the AWS Parent Region or the internet.
Compute and Instance Availability
The compute offering for AWS Wavelength is very limited at the moment. For your standard compute needs, you only have t3.medium, t3.xlarge and r5.2xlarge instances. If your workload is GPU-heavy, g4dn.2xlarge can be used. You will probably see more and more instance types and sizes available in the future, but as of now, this is one of the limiting factors for running your application via AWS Wavelength.
AWS Wavelength Security
AWS Cloud utilizes a shared responsibility model for security, where customers are responsible for cloud security in terms of which services they opt to use, data sensitivity, compliance, and regulation. On the other hand, Amazon handles cloud security when it comes to the underlying infrastructure, trustworthy external services, and third-party audits.
This model holds for AWS Wavelength as well—there is no direct contact between you and the partnering telecommunications provider in that region.
AWS Wavelength Pricing
AWS Resources running in AWS Wavelength Zones are priced differently than the same services running in AWS Regions, usually at a higher price. For example, the above-mentioned t3.medium instance will cost you $0.0416 per hour in the US-East-1 (N. Virginia), while that same instance will cost you $0.056 per hour in all current AWS Wavelength Zones. The price for a t3.xlarge instance also jumps, going from $0.1664 per hour to $0.224 per hour. This is a difference of over 25%, which is far from insignificant.
There is an even bigger difference in EBS volume pricing . Running one in N. Virginia would cost you $0.10 per GB of provisioned storage per month, whereas in the AWS Wavelength Zones, the price increases by a staggering 50%, costing you $0.15 per GB of provisioned storage per month.
Additionally, there are data transfer costs, most notably going out to the public internet. When transferring data out of AWS from the AWS Region, the cost is $0.09 per GB. For AWS Wavelength Zones, the price goes up to $0.108 per GB.
While it’s reasonable that all the additional work on the AWS side would increase the cost of their services, these numbers are somewhat disheartening. Cost control in the cloud is one of the most important aspects, and price differences such as these could be a deterrent for many. Hopefully, these prices will go down soon, ideally to a level close to that of the AWS Regions.
AWS Wavelength Use Cases
So when and why would you use AWS Wavelength? What would be some good use cases for utilizing this new service?
There are actually many good use cases, for example, live interactive video streaming. With ultra-low latency support, live-streaming high-fidelity audio and high-resolution video is easy enough, but it can also be used to add interactivity into the stream itself.
Another use case would be for online gaming, especially the competitive scene which has been steadily, but rapidly, growing, as real-time gaming completely relies on low latency. Also, with services like AWS Wavelength, you can stream games to end users so that even those with older devices can play, completely negating the weak hardware that was usually a limiting factor.
Then, there are smart factories, machine learning-assisted healthcare, virtual reality, and even the future of intelligent driving.
A bold new move and a step in the right direction.
Because AWS Wavelength connects its worldwide data-center network with those of the various telecommunications providers, Amazon has managed to open a whole new world of possibilities. If you have latency-sensitive applications, you might want to give AWS Wavelength a try. Obviously, this service is in its infancy, and it might not have everything you need right now, but it is bound to grow, and we can only guess where it will take us in the coming years.