VMware edges towards its multi-cloud vision, with a side quest to virtualize AI

Explore VMware has kicked off its annual Explore conference with a series of announcements that advance its core strategy of using virtualized compute, storage, and networks to allow workloads to run with consistent security wherever their owners choose to have them execute.

As such, this year's conference represents steady advancement for VMware rather than the kind of ecosystem-advancing stuff it has pulled off in the past, for example innovations such as support for offloading housekeeping workloads into SmartNICs, or when the vendor almost reinvented storage with virtual volumes.

VMware enlisted the star power of Nvidia boss Jensen Huang for the most interesting announcement this year: the VMware Private AI Foundation with Nvidia, a cut of the virtualization giant's Cloud Foundation package dedicated to packaging AI workloads.

Intended for release in "early 2024," the package will give single VMs the ability to access up to 16 GPUs and allow users to employ different models and frameworks for AI workloads that use their own data. Nvidia NeMoTM and Llama 2 were mentioned as among the options.

VMware will also package virtual machines of workloads the company feels users will want to roll out quickly alongside their AI efforts. A vector database will be among those VMs, as a means of providing inferencing capabilities without having to refer to large language models. A repository of-pre-packaged models is also in the works.

Networks will be tweaked to pump data direct from virtual storage to GPUs without having to bother central processors, while integration between vSphere and Nvidia's NVSwitch will mean virtualized multi-GPU rigs shouldn't experience bottlenecks.

The compute, storage, and networking elements of the above explain why a cut of Cloud Foundation was needed to deliver this new offering – Foundation-grade offerings are effectively VMware's cloud-in-a-box. Now it has an AI cloud in a box coming to market. But for now, it's not in actual clouds. Dell, Lenovo, and HPE have signed up to offer reference designs for this package, but public cloud and service providers are promised as an undefined "future" offering.

Closer to the core

VMware's core products also got some attention.

VSAN is now available in a "Max" edition that disaggregates the virtual SAN, so storage and compute can be scaled independently. vSAN Max can reach up to 8.6 petabytes of capacity and 3.6 million IOPS per cluster. VMware's rival, Nutanix, did this in May.

The NSX virtual networking range also has a new member: NSX+, to overlay network and security policies, and perform enforcement, plus a new network detection and response (NDR) as-a-service offering, across multiple cloud environments.

VMware Ransomware Recovery does what it says on the box, with various techniques to snapshot and failover VMs, and protect data, before ransomware scum can shut them down.

The ransomware offering, NSX+, and VSAN Max are all offered as part of VMware Cloud, the portal-like service that allows VMware customers to buy credits they can apply to a variety of services that can run in many locations.

The three new services therefore represent VMware nudging users towards its cloudy subscription service. As all software vendors are wont to do these days.

AI for DevOps

It's 2023, so AI news is never far off. In this case it applies to VMware's efforts to court developers with its Tanzu portfolio, which simplifies deployment and management of Kubernetes clusters, and containerized apps, across multiple platforms. Tanzu facilitates infrastructure-as-code, the DevOps dream of allowing coders access to the resources they need without having to ask ops for assistance or wait for them to twiddle nerd knobs, and makes that theoretically possible by relying on other VMware products that ops teams already use and understand.

"Tanzu Intelligence" adds AI to do things like analyze app performance and right size Kubernetes clusters, and adds AI/ML-powered contextual troubleshooting that mines log files and uses a conversational AI interface to suggest changes. VMware believes cost savings will follow.

An Intelligent Assist service includes a generative AI chatbot VMware says will "improve operator workflows and expands to include an admin interface for Platform Engineering."

Administrators of virtual desktop environments and apps are also getting an AI helper that offers "anomaly detection that informs IT of potential experience issues for frontline devices and VDI environments," and for mobile and desktop devices too.

VDI users will also get enhancements to the App Volumes product that allows on-demand delivery of apps to persistent virtual desktops. Allowing installation when users want the app means virtual desktop images can be smaller because they don't have to include all allowed apps. VDI infrastructure can therefore shrink.

Living more comfortably on the edge

Also announced at the show was the VMware Edge Cloud Orchestrator, a tool that manages compute infrastructure, networking, and security, on edge rigs that run VMware’s Edge Stack, a version of VMware's core offerings designed to run in diverse locations where hands-on IT help is seldom if ever available. This is actually a significant release as it shrinks VMware's core tech to run even on modest kit powered by two-core Intel Atom CPUs. It also allows remote management chores such as software distribution and device management, on fleets of literally tens of thousands of edge machines.

The Orchestrator draws on VMware's other software-defined datacenter products, but has required plenty of fresh development effort: it's not a cut down version of the VMware stack, but a reworking of it to bring it to different environments.

• UK clears Broadcom buying VMware, but deal yet to scale Great Regulatory Wall of China
• Nutanix reckons you can stuff AI into a box – a box it manages, that is
• Five Eyes nations detail dirty dozen most exploited vulnerabilities
• UK greenlights Broadcom's $61B VMware takeover

To connect those locations, VMware announced a Private Mobile Network offering that sees it manage virtual networks carried by Betacom, Boingo Wireless, and Federated Wireless. The three carriers are billed as the "initial" operators with whom VMware will work.

As if to show what this is all good for, VMware introduced a "Retail Edge" – an encapsulation of its edgy offerings designed among other things to help retailers "accelerate the deployment of next-generation AI apps."

Which brings us back to AI, again. And probably again later, as The Register understands further news related to binary brainboxes will be revealed before the conference is done. Clearly, VMware wants its most faithful conference-attending followers to know it can apply all its best tricks to this year's hot workload.

But the piece of news that will most impact VMware's future – how Broadcom behaves when it acquires the company, as now seems all but certain – remains beyond the sight of even an AI. ®