'Go faster' with core cloud services: AWS re:Invent 2020 roundup
December 22 2020
by William Fellows, Owen Rogers, Jean Atelsek, James Sanders
AWS recently held its ninth annual re:Invent conference - the first one that was entirely virtual, taking place over three weeks. Being free to attend ensured more than 400,000 registrations. This report focuses on some of the significant core cloud developments from week one of the event.
The 451 Take
CEO Andy Jassy's opening remarks provided connective tissue from re:Invent Las Vegas to 'Las Virtual' – reminding the audience, as he does each year, that re:Invent is primarily an education, developer-driven and networking event. The cloud may be well into its teenage years now, but most of the opportunity remains ahead, given the single-digit percentage of IT spending it accounts for. While the pandemic may have accelerated the move to, and use of, the cloud, Jassy's plea was stark: Go faster. Communicating a palpable sense of urgency, the message is that in order to access the benefits of the cloud, traditional enterprises must embrace fundamental structural change – not only technology debt and legacy infrastructures, but cultural and organizational inertia – as the key obstacle. This year, AWS is reaching further down into the enterprise with on-premises Anywhere and other offerings, which reaffirms that AWS 'everywhere' is an inevitability.
AWS leads with its strongest suit – VMs
Cloud computing is no longer just a bunch of managed servers and storage sitting in someone else's datacenter, available for rent by the hour. AWS and others are building cloud services up the stack at a breathtaking pace, to meet every enterprise need up to the application layer (and some of that as well).
Despite this, re:Invent opened on the announcements of new virtual machines, which was interesting. The detail of these instances is less important than their prominence and variety. Despite all the excitement about cloud-native, serverless, containers, machine learning, analytics and everything else up the stack, the core of most enterprises' IT estates is servers. The big tickets in enterprise cloud spend are virtual machines and storage. Moving an app to a virtual machine that is cheaper and faster is a far simpler proposition than reengineering it to take advantage of new services further up the stack.
And in the time of COVID-19, quick wins are the order of the day. By putting compute front and center, AWS is demonstrating to enterprises that it is continuing to focus on two key enterprise IT concerns: lower costs and better performance. In fact, the term 'price-performance' was used liberally during the week-one keynote to describe the value of many of the new instances announced.
Moreover, by releasing specialist hardware for specialist jobs – most notably machine-learning instances based on custom silicon – AWS is also showing compute isn't commodified yet. Enterprises want access to the most innovative technology, but to consume it in the on-demand, pay-as-you-go manner they've become accustomed too.
On-premises container options with Anywhere
The rise and rise of 'cloud to ground' and 'cloud all around' offerings points to customer demand for on-premises and multicloud technology delivered and supported by the hyperscalers. They want to bring the efficiencies, automation and velocity of the cloud to wherever their workloads need to be. Kubernetes is the dial tone for this. At its re:Invent 2020 virtual event, AWS called up a raft of new enterprise hardware and software (vs. cloud services) to meet this need.
On the first day of the conference, AWS announced that in 2021 it will roll out two services – Amazon ECS Anywhere and Amazon EKS Anywhere – enabling customers to run AWS containers in the environment of their choice, including other clouds. To allow customers to start prepping their containers for Amazon EKS Anywhere, the company is immediately open sourcing Amazon EKS Distro, its distribution of the open source Kubernetes container orchestration platform.
The two new services will use the same APIs and cluster configurations as in AWS's public cloud, easing control and management of heterogeneous IT environments without requiring installation of Outposts hardware or VMware Cloud on AWS. Outside of AWS and Outposts, EKS Distro is customer managed. It also brings a migration and portability story, enabling customers to use on-premises and the cloud as workloads requirements change (dev/production, among others).
By enabling customers to mimic the AWS EKS experience on any server or VM, Amazon EKS Anywhere joins similar services from fellow hyperscalers Google (Anthos) and Azure (Arc). All require customers to install Kubernetes-based middleware on-premises or on third-party clouds, where the services will run.
AWS acknowledged as part of the announcement that many developers and commercial SaaS providers are using Kubernetes to define and deliver applications across environments. With Amazon EKS Anywhere and Amazon EKS Distro, the company is making it easier for these stakeholders to serve customers across environments, as well as freeing its user base to venture beyond AWS.
With the continuing ascent of serverless containers as a first-class execution option, the extension of AWS Lambda to support containers as a packaging format will be welcome. This will enable users to use tools such as Docker CLI and container registries to package and deploy Lambda functions as container images (although this is not an embrace of the open source Knative serverless container platform).
'Easy button' for cloud-native deployment
Although AWS is largely a portfolio of building blocks, it is also about reducing the integration requirement and removing the need for (expensive) developers to concern themselves with application deployment, stack assembly and maintenance.
The new AWS Proton is a fully managed and automated development and deployment service for container and serverless applications. AWS Proton is essentially a self-service catalog of tools that provides an 'easy button' way to assemble components and publish stacks as reusable schemas. The idea is that teams can use Proton to connect and coordinate the different tools needed for infrastructure provisioning, code deployments, monitoring and updates from containers and serverless to VPCs, firewalls and load balancers. Stacks are patched and updated automatically. AWS is offering pre-built stacks that encapsulate some of its best practices in a number of areas.
At the same time, AWS announced the Amazon Elastic Container Registry (ECR) Public, for sharing and deploying container software publicly. It overcomes the pull-rate limit (at least to free accounts) that Docker's has introduced. Amazon ECR Public integrates with AWS Fargate, Amazon ECR and AWS Marketplace, enabling AWS customers to consume publicly distributed container software. A new website called Amazon ECR Public Gallery is available to allow anyone to browse and search for public container images, view developer-provided details and see pull commands – without needing to sign in to AWS.
AWS has introduced 1U and 2U rack-mountable versions of AWS Outposts. The former pairs an Arm-powered, 64-core Graviton2 processor with 128GB RAM and 4TB NVMe storage; the latter utilizes Intel CPUs to provide 128 cores and is paired with 512GB RAM and 8TB NVMe storage. Amazon's new diminutive Outposts options joins the firm's full-size (42U) fully managed rack configurations that were released to general availability in 2019.
Generally, deployments of AWS Outposts can run a subset of AWS services on-premises, managed identically to AWS-hosted resources in the AWS console. Between the new 1U/2U options and the existing full rack configurations, AWS represents a substantial threat to the (ahem) reinvented business models of the traditional infrastructure vendors – the utility for a 'single pane of glass' to manage AWS and on-premises compute resources is obviated if the on-premises hardware is managed seamlessly inside the AWS Console.