How Innovations in VMware vSphere® are Redefining Infrastructure to Run Future Applications

 By Krish Prasad, SVP & GM, Cloud Infrastructure Business Group, VMware 

Businesses of all types are on their digital transformation journeys – modernizing and deploying applications to help drive growth and maintain a competitive advantage.

These new applications are driving major shifts in infrastructure architecture. First, they are driving innovation in hardware leading to the emergence of hardware technologies such as GPUs, DPUs, SmartNICs and NVMe that power applications with artificial intelligence (AI), machine learning (ML), and memory-intensive algorithms. Organizations are also looking to adopt a DevOps model that leverages containers and Kubernetes to deliver new IT services and applications quickly, which in turn needs an agile on-demand infrastructure, further complicating infrastructure requirements. Finally, customers are selecting their deployment locations closer to the point of need and using a multi-cloud strategy that extends all the way from public clouds to private clouds to the edge. This is pushing organizations to specialized infrastructure silos with different tools and processes, creating operational and security issues.

Over the years, VMware’s software-defined infrastructure has helped hundreds of thousands of organizations reduce costs while increasing agility through virtualization and consolidation. VMware vSphere, the industry-leading compute platform, powering more than 85 million workloads across private and public clouds. From compute virtualization to storage virtualization, to networking virtualization and beyond, the VMware innovation engine hasn’t stopped.

vSphere, which powers VMware’s multi-cloud solutions, is evolving to unleash the latest advances in silicon-based accelerators, high speed interconnects, and cloud-scale architectures. Today at VMworld 2021, we’re unveiling new Technology Previews that will help our customers deploy consistent enterprise-grade infrastructure across the multi-cloud that will address the requirements of the applications of the future.

Enabling Modern Apps and Multi-Cloud through Kubernetes 

With vSphere 7 we delivered Kubernetes and container support natively in vSphere  – enabling customers to run containers and VMs on the same platform, while enabling IT admins to manage Kubernetes and containers by leveraging the same tools and processes that they used to support VMs. vSphere with Tanzu provides the fastest way to get started with Kubernetes workloads on an exceptional platform to run Kubernetes, offering all the powerful capabilities of VMware infrastructure, such as better security, robust reliability, availability and resiliency.


On vSphere, modern Kubernetes-based containerized applications can run side-by-side with traditional VM-based applications, at speeds equal to or faster than those in bare-metal environments. A third party study ** shows that  vSphere 7 with Tanzu beats bare-metal for Kubernetes workloads in both performance and pod density – vSphere with Tanzu can support 6.3 times more container pods per physical host when compared to the bare-metal alternative!




But that was just the start. Today at VMworld 2021, we announced Project Cascade as a Technology Preview*. Project Cascade will provide a unified Kubernetes interface for both on-demand infrastructure (IaaS) and containers (CaaS) across VMware Cloud – available through an open command line interface (CLI), APIs, or a GUI dashboard.  Project Cascade will address the needs of developers and DevOps in a multi-cloud world. It will enable a true open multi-cloud with an industry standard programmatic interface that is powered by Kubernetes.  Project Cascade will be built on an open foundation, with the open-sourced VM Operator as the first milestone delivery for Project Cascade that enables VM services on VMware Cloud.

Today, we also announced Project Arctic as a Technology Preview*, which is the next step in the evolution of vSphere in a multi-cloud world. Hundreds of thousands of customers run vSphere and many still have large on-premises deployments that continue to deliver value. What if they could continue leveraging their existing investments as well as toolsets and domain expertise, but still benefit from the expanded cloud capabilities with VMware Cloud for a complete hybrid cloud experience?

Project Arctic will bring multi-cloud to the fingertips of vSphere customers, by natively integrating cloud connectivity into vSphere – making vSphere cloud-aware, and making hybrid cloud the “default” operating model.  Millions of IT Admins who use vCenter on-premises will be able to leverage their domain expertise and benefit from expanded cloud capabilities. Customers would be able to instantly access VMware Cloud capacity and deploy VMware Cross-Cloud Services – for example, they would be able to enable disaster recovery to the cloud with just a few clicks, or leverage security services for threat detection and ransomware protection.

Interested in joining a future beta of Project Arctic?  Sign up here

Project Arctic and Project Cascade together will address the needs of two distinct groups in a multi-cloud world: Project Arctic will address the needs of IT operators who need to deliver enterprise IT resources, and Project Cascade will address the needs of developers and DevOps professionals who need to rapidly develop and deploy modern applications.


Reimagining Infrastructure as a Distributed Architecture with Project Monterey

Project Monterey is the next step in the evolution of cloud infrastructure to meet the needs of modern applications, including bare-metal workloads. Project Monterey reimagines the virtual infrastructure as a distributed architecture that is enabled by the Data Processing Unit (DPU) and offers a single, more secure operating model that unifies management for traditional, cloud-native, and bare-metal workloads for our customers. Infrastructure services will be offloaded to the DPU, providing significant network acceleration, improved security isolation, and the ability to run bare-metal workloads managed by vSphere. We announced Project Monterey at VMworld last year in collaboration with several key industry partners and now have an early access program for select customers. To learn more about Project Monterey, read this blog

Democratizing AI for the Enterprise with NVIDIA

We see organizations in all industries leveraging AI and machine learning to drive business value from data.  Customers leaning into AI have been setting up specialized silo stacks on bare-metal to run AI/ML workloads. This approach is not scalable and increases the risk of projects failing to reach production. In response, we have mainstreamed AI in the enterprise by partnering with NVIDIA, the leader in AI, to deliver an AI-Ready Enterprise Platform running on vSphere.

And we’re delivering all the benefits of virtualized AI workloads without performance tradeoffs. Compared to NVIDIA’s benchmarks for AI performance*** run on bare-metal, we’ve seen that with our combined solution the virtualized setup running on vSphere performs at the same level, and scales linearly. In other words, organizations get all the benefits of virtualization without compromising performance.




Enabling Memory Intensive Apps with Project Capitola

As applications consume an ever-increasing amount of memory, the cost of memory to support these applications is becoming significant. Many aspects of cloud and data center infrastructure such as compute, storage, and networking are software defined, but memory remains a hardware component not governed by software intelligence. This has led to underutilization and inefficient use of memory capacity that straddles several price and performance points between technologies such as DRAM, PMEM and NVMe. To address this, we are excited to announce the Technology Preview* of Project Capitola.

Project Capitola is a software-defined memory implementation that will provide the best price-performance ratio for current and next-gen applications by aggregating memory tiers across DRAM, PMEM, NVMe and other future technologies. Project Capitola will be enabled for on-premises data centers and the cloud by bringing together a broad ecosystem of memory and server vendors. I am glad to share that we have joined forces with Intel, one of our lead partners for Project Capitola, to blend our expertise in software defined data center management with Intel’s industry leading Intel® Optane™ PMem on Intel® Xeon™ platforms, as a first step towards delivering on the promise of Project Capitola to our customers.

To learn more about Project Capitola,, read this blog.

Equivalent in Performance Against Bare-Metal for Telco Workloads

As we looked both at the Telco space and needs of real-time applications, where latency must be as low as possible, we implemented optimizations within ESXi inside the VM kernel itself, providing a hard guarantee for real-time applications. With these improvements, we’ve seen vSphere perform extraordinarily well against bare-metal systems. This breakthrough is the result of years of innovation to virtualize some of the most difficult and rigorous workloads out there.

 vSphere is Your Platform for the Future

 As you can see, the vSphere innovation engine keeps revving – we are continuing to evolve vSphere to be your most preferred platform for all enterprise workloads, from traditional, to containerized, to AI/ML, and more.

We’re thrilled to have been able to work with our customers and partners to develop vSphere as a true multi-cloud platform for all types of apps as they embark on their digital transformation journey.


Learn More About Upcoming Innovations in vSphere​

 Don’t Miss These Sessions at VMworld 2021​



*The development, release, and timing of any features or functionality described for VMware’s offerings in this presentation remain at the sole discretion of VMware.

**A Principled Technologies Report, commissioned by VMware: Pod density comparison: VMware vSphere with Tanzu vs a bare-metal approach using Red Hat OpenShift 4.7, August 2021

*** NVIDIA Benchmark performance: