Can HPE use Juniper and AI to converge networking?

Can HPE use Juniper and AI to converge networking?

Hewlett Packard Enterprise (HPE), at its annual Discover event this week, followed the industry trend toward blanketing all news and keynotes with artificial intelligence (AI), which is set to also bolster the vendor’s operations should its pending $14 billion purchase of Juniper Networks reach completion.

HPE’s biggest AI news of the week was tying chip giant Nvidia into new AI computing and private cloud services that promise to make it easier for enterprise customers to drive AI into their operations and, more importantly, drive cost efficiencies from AI.

However, that AI focus is also set to push the need for more powerful networking services.

Neil MacDonald, EVP, and GM for compute, high-performance computing (HPC) (HPC), and AI, at HPE, during an investor relations session at this week’s event, explained that AI systems involve four different networks.

This includes “fabrics that connect accelerators together” and “fabrics that connect sets of accelerators together.” The other two are data center networking that interconnects these fabric connections, and “then you have whatever you’re using to get to your data and your storage.”

MacDonald noted that this diverse set of networking needs means that there will not be just a single networking solution.

“I think sometimes when this thread of conversation comes up, it comes up through the lens of what’s the one binary answer,” MacDonald said. “There isn’t going to be one binary answer.”

MacDonald noted that this will leave open the opportunity for both Ethernet and InfiniBand to serve different needs. The former has long been the backbone of enterprise networking due to its widespread adoption and cost-effectiveness, while the latter remains appealing because of AI needs.

“The biggest shift that’s going to emerge is that InfiniBand, which has been used in some of the highest performance systems in the world, including some of the highest performance AI systems in the world, is something of an alien technology for many enterprises and as a result, Ethernet is much more absorbable,” MacDonald said. “But classic Ethernet is not sufficiently performant for these workloads.”

MacDonald’s assessment was backed by a recent list of the world’s most powerful supercomputers, which showed InfiniBand used more often for supercomputing networking compared with Ethernet.

This has led to new options like Nvidia’s Spectrum-X Ethernet, which is based on its Mellanox acquisition, and the HPE Slingshot interconnect system, which is used in the Frontier supercomputer.

“Over time, you’ll see richer, more capable Ethernets among some of those fabrics, but it’s still going to remain very heterogeneous at the different layers because they’re designed for different things,” MacDonald added.

Customers increasingly want to buy software that is secure by design. Here’s what that means for CISOs.

HPE and Juniper remain high on AI networking

HPE CEO Antonio Neri pivoted that heterogeneous architecture to where Juniper’s assets can come into play, which is to “integrate fully down with Nvidia as we go forward and make it easy to manage.”

That focus and ease of management should align with Juniper’s recently updated AI-Native Networking Platform, which gained capabilities designed to bolster the vendor’s SD-WAN and secure access service edge (SASE) platforms. Those updates are based on Juniper’s Mist AI platform, which has been a central growth point for Juniper and core to HPE’s acquisition ambition.

Neri explained that he sees the combined networking assets, AI integration, and access to HPE’s GreenLake hybrid-cloud model as convergence points.

“I envision a convergence of the data center networking and the stack over time and integrated with Nvidia,” Neri said. “Today, they have almost two different control planes, and that’s why this experiment – I’ll call it, although it’s a true product now – is important because as we integrate with GreenLake, we can integrate the rest of the architecture over time.”

As for progress on the deal, Neri said it remains on track to close by early 2025, and “we don’t see so far any issues.” The exec added that some countries have already approved the deal, “but, obviously we need to wait for the United States, European Union, and that’s going fairly well.”

“Nothing gives us the pause to think something’s going to go wrong,” Neri said.

Author

Related posts

NONSTOP TBC 2024 UPDATES

Success always attracts a crowd

Why choose HPE Education Services?