VMware upgrades software-defined edge for AI workloads

VMware announced a number of improvements to its software-defined edge products at its annual VMware Explore conference today. Improved connectivity and traffic-management features are aimed at helping enterprises to deploy AI at the edge.

On the connectivity side, the VMware VeloCloud Edge 710 appliance will now have a combined fixed wireless access and satellite connection support, as will the new VMware VeloCloud Edge 720 and 740 appliances, the company announced.

In addition, Symantec points of presence will now be integrated with VMware VeloCloud SASE, while VMware Edge Compute Stack, a runtime and orchestration platform, is now optimized to support AI workloads and traffic patterns.

Edge computing will play a “pivotal role” in the deployment of AI applications, according to research firm IDC, which expects worldwide spending on edge computing to reach $232 billion in 2024, an increase of 15.4% over 2023.

Traditional applications are centered in a company’s data center or cloud. Many AI workloads – especially autonomous AI workloads like video inference cameras, industrial control systems and other operational technology applications – require local processing.

“Edge AI is everything that’s dealing with AI that is outside of the main public and private data centers,” says Sanjay Uppal, general manager for the software-defined edge division at Broadcom, which completed its acquisition of VMware in late 2022. “And what we’ve done is we have revamped our VeloCloud portfolio to enable edge AI workloads.”

Take, for example, VMware’s SASE product. “You get single-vendor SASE from us,” says Uppal. “But now this single-vendor SASE has the ability to support the new networking that is coming out from these edge AI workloads.”

Similarly, the VeloCloud Edge Compute Stack has a different orchestration mechanism than what companies are normally used to with data centers, he says. “Because what is getting orchestrated is distributed around sometimes hundreds, sometimes thousands, sometimes tens of thousands of locations.”

There’s added complexity given that sometimes, the data collected at the edge is processed at the edge, and sometimes it has to be correlated and aggregated centrally, he adds.

There are now several customers for the latest version of VeloCloud Edge Compute Stack in various stages of implementation, Uppal says, with hundreds of edge deployments.

“We’re aligned with Broadcom’s Edge AI approach,” says Keith Bradley, vice president for IT and security at Nature Fresh Farms, in a statement. “We rely on IoT devices at the edge – in our greenhouses and other facilities – to monitor and capture data used to keep millions of growing plants healthy.”

Nature Fresh Farms uses 5G and broadband to connect facilities across Canada and US via the VMware VeloCloud SD-WAN. And it uses VMware VeloCloud SD-Access to get consistent connectivity, performance, and security for its IoT devices in these remote locations.

“We’ve improved our farm-to-fork time to deliver better tasting, higher quality produce to our customers,” Bradley says.

VMware is also using generative AI itself in its product. For example, it has incorporated genAI into its dynamic multipath optimization algorithm, which learns and responds to the ebb and flow of edge workloads.

“We’ve trained DMPO on thousands of workloads and applications,” says Tal Klein, head of marketing for software-defined edge at Broadcom.

Orchestrating these workloads properly can be critical, he says.

Take, for example, a retail chain that has video cameras at its stores. “Think of the sheer quantity of video traffic a camera produces when it is streaming video,” he says. “Now multiply that by tens of cameras.”

But not all this traffic is equal. “The most important moments are when an item is being scanned or a face is being recognized,” Klein says. “It’s critical in those moments that the resources necessary to achieve each workload’s business purpose are available.”

Similarly, in a factory, if an industrial robot can’t get the right resources at the right time, the manufacturing line could stall, Klein says.

Read more VMware news

• With Project Cypress, VMware brings generative AI to cyberdefense
• Customer concerns loom as VMware Explore event approaches
• VMware by Broadcom: Product, service and support news
• Broadcom extends vSphere 7 support six months
• Broadcom bolsters VMware Edge Compute Stack
• Broadcom repackages VMware SASE, adds Symantec security