Amazon Web Services Inc. today extended both the scope and the reach of its massive cloud computing services, announcing new computing services and new ways to bring them into the heart of corporate data centers. 

For one, it introduced five EC2 instance types at its virtual re:Invent event, among them the high-speed M5zn series for computationally intensive applications and new G4ad instances featuring Advanced Micro Devices Inc. graphics cards. It also announced four new capabilities for containers, software that enables applications to be deployed easily across many computer systems.

Not least, the cloud giant made additions to its edge computing portfolio, including two smaller AWS Outposts appliances that customers can deploy on-premises in their own data centers.

“We’ve said the vast majority of companies in the fullness of time will have their operations in the cloud,” AWS Chief Executive Andy Jassy (pictured) said in his three-hour re:Invent keynote this morning. But for now, he acknowledged, AWS wants to meet them where they are. “Effectively, they want us to distribute AWS to these edge nodes,” he said.

More flavors of compute

AWS offers customers of its cloud several quite a few choices when it comes to processors. The EC2 lineup has instances powered by Intel Corp. central processing units, ones featuring Nvidia Corp. graphics processing units and a growing number of instances powered by AWS-designed chips such as Graviton2, to name a few of the available options. The cloud giant is now adding yet another type of silicon to the mix: Advanced Micro Devices GPUs.

AWS’ new G4ad instance series, one of the five announced at  re:Invent today, are the first in the EC2 lineup to feature graphics cards from AMD. Specifically, the cloud giant is using workstation-grade Radeon Pro V520 GPUs paired with the chipmaker’s Epic CPUs. The target use case for G4ad is running graphics-intensive workloads such as video rendering systems and streaming services.   

According to AWS, the G4ad offer significantly better price-performance than its previous top-of-the-line graphics-optimized instance, the Nvidia-powered G4dn series. “When compared to G4dn the new G4ad instances enable up to 45% better price performance for graphics-intensive workloads,” AWS developer advocate Steve Roberts wrote in a blog post. “Compared to an equally-sized G4dn instance, G4ad instances offer up to 40% improvement in performance.”

The new M5zn instance series announced alongside the G4ad also targets compute-intensive workloads, though it’s aimed at a different slice of the market. EC2 M5zn instances are powered by customized CPUs from Intel’s Xeon Scalable line that offer top all-core frequency of up to 4.5 gigahertz. Moreover, AWS says that the chips provide the fastest single-thread performance of any Xeon Scalable chip currently available to organizations via the cloud.

Also notable is what the new M5Zn instances don’t include. There’s no local flash storage and there’s a lower ratio of memory to compute power than in the previous compute-optimized instance series that preceded the M5zn. The reductions, AWS said, will help customers running compute-intensive applications avoid the risk of provisioning more capacity with their instances than what their workloads can use.

The M5zn was developed based on users’ feedback. “They have indicated if we built an instance with this set of attributes, it would be an excellent fit for workloads such as gaming, financial applications, simulation modeling applications such as those used in the automobile, aerospace, energy and telecommunication industries, and High Performance Computing,” detailed AWS chief evangelist Jeff Barr.

The three other instance types announced today cater to three different sets of applications. For applications that work with large amounts of information, new D3 instances offer up to 38 terabytes of storage capacity per virtual machine. Meanwhile, applications such as analytics tools that not only hold onto large amounts of information but also regularly access that data can be deployed on the new R5b instance. It can transport data to and from AWS’ EBS block storage service for processing at a speedy 60 gigabits per second.

Rounding out the trio is the C6gn series. Similarly to the RB2 instance, the series is optimized for applications that need to move large amounts of information around quickly. Customers will have up to 38 gigabits per second of EBS block storage bandwidth and  network bandwidth of up to 100 gigabits per second, or two and four times more, respectively than in AWS’ previous generation of network-optimized instances. 

C6gn runs on AWS’ homegrown Graviton2 processors under the hood. And there’s more coming, Jassy said during this keynote: “We are not close to being done investing and inventing with Graviton.”

More options for containers

AWS also announced four new services around containers, which host modern, “microservices”-based applications that can run across many computer systems. It’s allowing customers to run its Elastic Container Service or ECS and its Elastic Kubernetes Service or EKS for managing containers in their own data centers. The idea behind ECS Anywhere and EKS Anywhere is to provide more control to customers while allowing them to use the same tools they employ to manage their Amazon cloud services.

A third new service, Proton, helps manage container and so-called serverless or more automated development and deployment of applications. The aim is to enable infrastructure teams to make application components organized into “stacks” available to development teams for a variety of use cases, and lets them store and reuse them, so when they build new applications, they can reuse common components.

“It’s a game-changer for managing the deployment of microservices,” Jassy said.

Finally, AWS added a public registry to its Elastic Container Registry, which is used to manage their container “images” employed in their applications. Now, with the public registry, developers can store, manage and share container images for anyone to use, without the company having to operate their own container repositories.

Edge computing in focus 

Continuing the string of compute-focused announcements this morning, AWS expanded its portfolio of edge computing services, which allow companies to run applications closer to their users to lower latency. Three products are getting enhancements: AWS Outposts, Local Zones and Wavelength.

AWS Outposts are essentially cloud-in-a-box appliances that enable companies to deploy AWS services on-premises. Two new, smaller appliances — one the size of a pizza box, the other the size of two stacked pizza boxes, Jassy noted — have been added to the lineup that are compact enough to be usable in small branch locations such as retail stores. The first appliance packs 64 vCPUs, 515 gibibytes of memory and 4 terabytes of flash storage into a one rack-unit form factor, while the other provides up to 128 vCPUs, 512GiB memory, and 8TB of storage in a two rack-unit chassis.

The new devices will be coming next year.

AWS today also inaugurated three new Local Zones, or regional data centers, in Boston, Houston and Miami. It shared plans to expand the service to a dozen additional cities next year. Verizon Communications Inc., meanwhile, said against the backdrop of re:Invent that it has teamed up with AWS to bring the cloud giant’s Wavelength service to Las Vegas. Wavelength is based on a similar concept as Local Zones: It lets customers run latency-sensitive applications on AWS hardware co-located near or inside major cities,  at the edge data centers of carriers such as Verizon.

With reporting from Robert Hof

Photo: AWS/livestream