VMware Cloud on AWS Is Breaking Boundaries with Disaggregated Architecture and Deeper Support for ML/AI Workloads

A few years back, when we partnered with AWS, our goal was clear: provide a service for vSphere workloads to rapidly and seamlessly tap into the value and innovation offered by the AWS Cloud.  Fast forward to today, first-time customers are often amazed by the speed and efficiency with which they can migrate large numbers of vSphere workloads, even entire data centers, to VMware Cloud on AWS. Customers like Kingston University, VanEck, S&P Global, and thousands of others have benefited from our hybrid cloud approach.

But more importantly, beyond leveraging the same VMware skills, tools, and processes, customers quickly realize that the VMware Cloud on AWS service directly and profoundly influences both their IT organization and their overall business. In a 2022 Forrester customer survey, VMware Cloud on AWS demonstrated substantial savings, with a 53% reduction in infrastructure and operations costs compared to on-premises data centers. Additionally, after evaluating hundreds of customer opportunities, the VMware Cloud Economics team found that a typical comparison of VMware Cloud on AWS against a functionally equivalent traditional public cloud alternative demonstrates a 35% lower total operating cost over a three-year period. (And notably, this savings excludes the additional cost savings on Microsoft operating system licenses and one-time migration expenses.) The value we deliver also extends into the public sector, particularly with the VMware Cloud on AWS GovCloud service engineered for the needs of US government agencies. This service has been both FedRAMP High and IL5 authorized and has experienced a remarkable 160% growth in consumption over the past year.

To meet the evolving needs of our customers, we set another goal in the early days of our partnership: to continually and rapidly deliver innovation and enhancements.  Today, we introduce significant improvements every quarter.  This includes new enterprise capabilities such as the recent networking and storage improvements, new AWS regions such as Bahrain, Zurich, Melbourne and Hyderabad that we added this year, and new commerce capabilities such as the free 30-day trial introduced over the summer.  In all, over the past few years we have delivered 24 major releases and many smaller security patches and are operating the service in 26 AWS Regions around the world.

Customer requests and feedback are the most important factors in determining the innovations we introduce to the market. Recently, customers have emphasized the importance of our traditional strengths—protecting, extending, or migrating their vSphere workloads to the public cloud. They’ve asked us to broaden our support for an even wider array of workloads. At the same time, there’s a growing interest in further leveraging AWS capabilities, particularly for ML/AI workloads. Today, at AWS re:Invent 2023 we’re thrilled to announce our commitment to fulfilling both these requests with ground breaking innovations for the VMware Cloud on AWS service.

Key highlights include:

New disaggregated storage architecture – we are breaking the link between compute and storage.  Until now, the VMware Cloud on AWS service relied on a hyperconverged architecture, in which instances have high-performance SSD drives directly attached to the host as primary storage. The new architecture now allows us to provide instances which access primary storage over a network and therefore do not need locally attached storage. The new architecture offers several benefits:

• Instance variety – the new architecture provides the opportunity to leverage a diverse range of AWS instance types. Since the majority of AWS instances lack locally attached storage, this opens the door to future possibilities of adding instances tailored for specific use cases, such as GPU instances or memory-heavy instances.
• Flexible and scalable primary storage options – customers can now choose to use either VMware Cloud Flex Storage or AWS FSx for NetApp ONTAP as primary storage. This allows them to allocate the precise amount of storage they currently need and easily scale it up (or down) to optimize their capacity or rapidly adopt to changing requirements.
• Cluster specialization – with the new architecture, customers can mix hyperconverged and disaggregated clusters. This means they can create dedicated clusters for applications that demand high-performance I/O, as well as clusters optimized for applications requiring substantial CPU processing. This flexibility leads to improved performance and total cost of ownership (TCO).

Support for the diskless AWS M7i.metal-24xl instance – the first AWS bare metal instance without locally attached storage that leverages the new disaggregated architecture. The M7i instance enables VMware Cloud on AWS to better support for the following use cases:

• Smaller footprint to get started – an instance with less memory and storage capacity than existing options makes it easier for new customers or those with smaller environments to embark on their VMware Cloud on AWS journey, ensuring a better fit for their requirements and budget.
• Optimized for CPU demanding workloads – applications that heavily rely on CPU resources, such as ML/AI, video and image processing, can now benefit from the new instance without unnecessary storage overhead.
• Tailored for lower performance or capacity storage workloads – applications with less stringent storage demands—such as secondary databases or those requiring less storage (e.g., application servers, management services)—can operate more cost effectively with the networked attached storage.
• More efficient Ransomware and Disaster Recovery: a more cost-effective isolated recovery environment and failover capacity will enable more customers to confidently and quickly recover from ransomware attacks and other disasters.

Support for Intel 4th Generation Intel Xeon Scalable processors – the latest CPU from Intel available on the M7i.metal-24xl instance offers built-in accelerators to improve performance efficiency for cutting-edge workloads such as ML/AI, analytics, or HPC workloads.  The recently released support for VMware’s Virtual Hardware version 20 (vmx-20), which provides access to advanced CPU capabilities, now allows customers to harness the power of Intel’s Advanced Matrix Extensions (AMX) accelerator. This enhancement significantly boosts the performance of deep-learning training and inference on the CPU, making it a great fit for workloads such as natural-language processing, recommendation systems, and image recognition.

Learn more about these new capabilities by visiting a more detailed product blog.