Uneven Innovation At AWS re:Invent 2022
Some 14 Forrester analysts attended this year’s AWS re:Invent in Las Vegas. The conference was a mix of innovation in some areas — particularly compute and networking, along with a co-engineering approach to industry cloud — but plateaus in others, such as hybrid and edge. Efficiency and consolidation were underlying themes, from amping up AI/ML and analytics offerings to streamlined platform operations.
As usual, there was a blizzard of announcements, and Forrester analysts will follow up on highlights in additional blogs. Here are Forrester’s initial observations of re:Invent.
Compute: AWS is pressing its advantage in custom silicon and multiple instance types. It emphasized its ability to provide low costs and high performance in the same offering. It is no longer accurate to say that cloud compute is commodified at rough parity across the hyperscalers. The announcements of Nitro v5, C7gn (powered by Nitro v5), HPC7g, Inf2 for EC2, and SimSpace Weaver indicate AWS’s push to address HPC workloads, focus on bigger and faster compute, and keep pace with AI innovation coming out of Microsoft Azure and especially Google Cloud Platform (GCP) while continuing to offer lower-priced instances for mainstream workloads running on ARM-based Graviton chips.
Network: AWS’s Elastic Network Adapters (ENA) Express announcement enhances ENA, the proprietary network adapter that the company rolled out in 2016 inside EC2 to decrease latency, improve network reliability, and increase bandwidth consumption allowed by customers within an EC2 instance. The solution combines ENA with Scalable Reliable Datagram (SRD), an alternative to Transmission Control Protocol (TCP) and InfiniBand. If executed as claimed, ENA and SRD have the potential to address data latency and performance concerns for data-heavy apps. An interesting aspect of SRD will be the protocol’s ability to send packets across multiple links and reassemble out-of-order packets. This could significantly increase network utilization that is horribly low, around 20%, thereby lowering networking infrastructure costs across the entire network.
Storage: AWS continues to iterate on its broad cloud storage stack by improving performance, resilience, and controls. Specifically, the announcement of SRD for Elastic Block Storage (EBS) addresses potential latency issues for EBS and makes for higher, more predictable performance for latency-sensitive applications. Elastic File Storage (EFS) elastic throughput eliminates limits for high-throughput applications. S3’s new failover controls for S3 improves multiregion resiliency, and AWS Backup is adding support for Redshift and CloudFormation.
Hybrid and edge: In the past 12 months, hybrid has been a central focus for AWS. As with the other hyperscalers, AWS recognizes that it has a last-mile problem. The value proposition for cloud as a general-purpose tool is obvious to the common enterprise. The focus is now on reaching those workloads that have been difficult to migrate to public cloud environments. As such, AWS views “hybrid” as AWS capabilities and support that are available outside of its own data centers. On the main stage, several customer success stories featured AWS Outposts, Elastic Kubernetes Service (EKS) Anywhere, and the products that comprise the Snow “family” (which is actually a disparate set of hardware). There were no Snow-specific or hybrid announcements, however. Instead, the company showcased its expansion of infrastructure with 11 new regions, 33 new availability zones, and 15 new Local Zones in EMEA. The lack of Snow or hybrid announcements isn’t necessarily a problem for AWS, as customers are still ramping up in these areas, but shows that the hyperscaler conquest of the edge is by no means certain.
Application development: AWS seeks to expand its cloud constituency with product announcements aimed squarely at developers. This year’s entries include Code Catalyst, a rather broad framework of tools that addresses three of the main pain points developers face when building complex applications in the cloud. Code Catalyst leverages Devfile to provide a more streamlined approach to establishing consistent cloud-based dev environments. At the same time, Blueprints reduces the friction of setting up the resources for application delivery and deployment and allows developers to deploy CI/CD pipelines already natively integrated with AWS services. Each of these capabilities enters a crowded space but, taken together with the advantage of native AWS integrations, can be considered a value add for teams looking to run their workloads in AWS.
For serverless pros, two announcements stood out: AWS Application Composer, which addresses serverless complexity for developers with drag-and-drop approaches for the often tricky process of configuring and building the resources and, yes, infrastructure around serverless apps; and Lambda SnapStart, which leverages cached snapshots to reduce the cold start issue, something Forrester identifies as one of the main impediments to mainstream serverless adoption.
Database, analytics, AI/ML: The AWS approach in these categories has been that more is always better. The theme of this year’s re:Invent announcements, however, was ease of adoption and integration with easier data ingestion into Amazon Redshift, more data connectors for AppFlow, AWS Glue Data Quality and Amazon Athena for Apache Spark using Jupyter Notebooks, and improved scale with Amazon DocumentDB. SageMaker got touch-ups plus a preview version of geospatial data, a nod to pressure from Google Cloud. There were a series of boosts for QuickSight as a BI platform to keep data built in AWS from being shipped off for analytics on Azure PowerBI.
Industry clouds: AWS emphasized its industry solutions in manufacturing and industrial, media and entertainment, financial services, games, and telecommunications, as well as particularly underscoring automotive and healthcare. AWS highlighted partnership priorities with big customers that can benefit from AWS’s co-engineering approach to high-priority industry solutions. Automotive companies were the most visible, with a mainstage highlight on Ferrari’s use of AWS simulation and HPC services. There was less talk about the other verticals outside of automotive and healthcare on the mainstage, but AWS has put forward a list of 20 total verticals that it plans to pursue.
Partners: The AWS relationship with global systems integrators got a new twist with the announcement of a deal with Atos that will move information technology outsourcing contracts toward cloud migration with an end run around the contractual barriers that have inhibited such efforts in the past. It’s a deal we’ll examine more closely in a future blog. More generally, re:Invent shows that AWS continues to be the foundation for the bulk of SaaS and tech companies generally. It’s likely that many of the capacity expansions and innovations that AWS provides to its customers are based on the requirements of AWS’s tech customers. This year’s re:Invent repeatedly pointed to how cost-conscious AWS customers can save money by counting their purchases as part of their committed cloud spend.
Marketplace: AWS Marketplace continues to simplify discovery and sourcing — especially for long-tail applications. At re:Invent, AWS announced several enhancements to the Marketplace offering, including a broader selection of data and software providers with supporting services, a streamlining of contracts to include language addressing common security elements, and expanding governance and risk management by validating that vendors have implemented security controls and monitoring those controls over time. These enhancements are aimed at removing the friction commonly associated with the sourcing and procurement process, which can lead to shadow procurement and increased risk exposure. It also speeds up the purchase cycle by pre-vetting vendors before listing them on the Marketplace, which goes a long way in easing the resources and time required for due diligence of new vendors while mitigating third-party risk. Buyers should note that this approach is only for products running on AWS, so it inevitably won’t be able to help everywhere. Larger purchases will likely still demand 1:1 attention from legal, sourcing, and security.
Sovereignty: AWS is opening more and more cloud regions, locations, and availability zones throughout the world, Europe being no exception. Although this strategy taps into one of the five requirements for sovereignty, it does not do much to address the other four. Specifically, it is not clear how AWS is going to address the need from non-US organizations that their data and applications be not open to the US authorities in case of investigations (a point made more urgent by the CLOUD Act). re:Invent did not address in detail the implications of the US-European Union Trans-Atlantic Data Privacy Framework released early this year. It remains unclear how AWS — as well as other US hyperscalers — will continue to support European clients in the face of rising sovereignty concerns.
Security: AWS is finally in the security analytics game, like its main hyperscaler competitors Google Cloud and Microsoft Azure. Its approach is a bit different, however. Earlier this year, AWS announced that it is a founding member of the Open Cybersecurity Schema Framework, which is building a vendor-agnostic standard for events. The effort includes vendors from across the security industry in a variety of domains. The announcement at re:Invent of Security Lake is a related next step, building the foundation for security data management. AWS announced several partners that will provide security analytics on top of Security Lake, which can give AWS customers some flexibility on which vendor they partner with for security analytics. Security Lake, though, is still in preview and has a number of important roadmap items it needs to build before it’s enterprise-ready. Moreover, the AWS announcement for runtime security for EKS is important as AWS customers scale out cloud-native infrastructure. Yet this seems only to match what third parties have offered for Kubernetes for some time. Additionally, AWS’s play in K8s security is largely focused on its own infrastructure — a marked difference from other cloud service providers that increasingly offer multicloud security offerings. AWS’s biggest differentiator in cloud security remains Nitro instances.
Given the size and scope of re:Invent, the above only scratches the surface. Check Forrester analyst blogs for further analysis.