Cisco has launched an all-in-one AI knowledge heart resolution co-developed by Nvidia that gives enterprises the server, storage, networking tools, and AI software program they should deploy generative AI and different AI purposes on-premises, the corporate introduced in the present day (June 4).
At its Cisco Dwell convention in Las Vegas, Cisco executives stated the brand new product, referred to as ‘Cisco Nexus HyperFabric AI clusters’, will function Cisco networking gear, Cisco compute options working on Nvidia GPUs and DPUs, Nvidia AI Enterprise software program, and VAST knowledge storage.
The purpose is to make it simple for enterprises to implement and handle new AI knowledge heart infrastructure in on-premises knowledge facilities, colocation amenities, and edge areas. To that finish, enterprises can centrally handle and monitor your entire resolution with Cisco’s cloud-based administration software program.
Cisco executives declare that Cisco Nexus HyperFabric AI clusters would be the first knowledge heart product line that may be cloud-managed. The corporate stated choose prospects may have early trial entry through the 2024 fourth quarter with basic availability anticipated shortly after.
In an interview with DCN this week, Murali Gandluru, Cisco’s vice chairman of product administration for knowledge heart networking, stated the collectively developed resolution with Nvidia will permit enterprises to “partake within the AI evolution” and to “simply design, deploy, function, and handle full-stack AI infrastructure in a plug-and-play mannequin.”
Analysts’ Takes: Cisco vs. the Competitors
With in the present day’s announcement, Cisco hopes to seize a share of the new marketplace for AI knowledge heart infrastructure. Business analysts say Cisco has the potential to succeed.
It helps that the corporate has partnered with Nvidia, whose GPUs dominate the AI market. However the firm’s networking experience can also be a plus, stated Zeus Kerravala, principal analyst of ZK Analysis.
“Relating to AI infrastructure, Cisco is considerably of an underrated firm,” Kerravala advised DCN. “One factor that’s not nicely understood within the trade is that the community performs an essential function in how AI performs.
He added: “You possibly can have all the pc clusters you need, however if you happen to can’t switch the info throughout them in a well timed method, then all the cash you spend on GPUs and AI servers goes to waste. Your networking must be in lockstep. Cisco remains to be the world’s largest networking vendor, and so they’ve bought techniques of all sizes.”
Cisco’s competitors on this house consists of server and networking distributors, Kerravala stated. Arista Networks final week announced it’s collaborating with Nvidia to develop a software program agent to higher handle networks and servers in AI clusters.
Dell Technologies, Hewlett Packard Enterprise, and Lenovo even have partnerships with Nvidia to develop generative AI {hardware} options. Cisco partnered with Nvidia in February, and Cisco HyperFabric is the primary product stemming from the connection, defined Kevin Wollenweber, senior vice chairman and basic supervisor of Cisco’s networking, knowledge heart, and supplier connectivity division.
Vijay Bhagavath, IDC’s analysis vice chairman of cloud and knowledge heart networks, stated Cisco is nicely positioned out there and has the suitable elements to succeed. Cisco has an enormous base of enterprise prospects, together with Fortune 500 firms, that it may promote to, he stated.
“Cisco has an actual footprint that they will go to. They’ll inform their present prospects, ‘In case you’re all in favour of working LLMs in your knowledge facilities, you may leverage our HyperFabric,’” Bhagavath stated. “It has all the suitable piece components: leaf-and-spine switches that may do 400 to 800G, the compute working on Nvidia GPUs, the cloud automation administration mannequin, and so they have all of the Nvidia options they will leverage – DPUs, NICs, and NIMs microservices.”
Use circumstances for Cisco’s resolution embody enterprises that need to implement AI purposes internally as a result of they’ve numerous knowledge that’s personal and confidential that they don’t need to expose to the cloud, Bhagavath stated.
For instance, some firms might have specialised use circumstances, equivalent to drug discovery or automobile design, the place they’ve specialised purposes and big quantities of knowledge that they need to hold in-house. Different enterprises might have to run an open-source giant language mannequin internally for regulatory compliance causes, he stated.
Extra Particulars on Nexus HyperFabric
Nexus HyperFabric AI clusters have been developed for enterprises which can be searching for to undertake AI infrastructure for generative AI and AI inferencing workloads however need to deploy it in a a lot less complicated means, Cisco’s Wollenweber advised DCK.
A key issue that permits simplicity is cloud administration capabilities which can be just like how Cisco Meraki has enabled enterprises to configure, handle, monitor, and troubleshoot their wifi and native space networks by way of a single dashboard, Gandluru stated.
For instance, Cisco HyperFabric, the corporate’s cloud-based administration software program, automates some duties – equivalent to community underlay configurations – and permits enterprises to centrally monitor the efficiency of workloads, Wollenweber stated.
Cisco HyperFabric will embody Cisco compute options which can be based mostly on Nvidia’s MGX server specification, which is a modular and versatile server structure that helps totally different kind elements and Nvidia {hardware}, a Cisco spokesman confirmed.
The answer will run on Nvidia Tensor Core GPUs, beginning with the Nvidia H200 NVL GPU. It’s going to additionally assist the Nvidia BlueField-3 DPU (knowledge processing unit) processor, which offloads processing duties like community, safety, and storage administration to lighten the load on CPUs, and BlueField-3 SuperNIC community accelerators, the corporate stated.
Different elements of Cisco HyperFabric embody:
-
Cisco Nexus 6000 sequence switches delivering 400G to 800G (gigabit ethernet) efficiency, and Cisco Optics household of QSFP-DD modules.
-
Nvidia AI Enterprise, a software program suite of AI instruments, frameworks and pretrained fashions designed to make it simpler for enterprises to develop and deploy AI workloads. It consists of Nvidia NIM microservices.
-
VAST Knowledge Platform, which affords unified storage, database and a data-driven perform engine for AI.
“We’re not simply taking Nvidia’s merchandise, placing them into servers and constructing a material, after which slapping them collectively. It’s a totally converged resolution,” Wollenweber stated.
Cisco Nexus Dashboard Enhancements
Not each enterprise desires to make use of cloud-based administration instruments, nonetheless. For these organizations, Cisco stated it has improved the Nexus Dashboard, a software program software for managing knowledge heart networks on-premises.
New enhancements embody centralizing the administration of Nexus NX-OS and ACI cloth controllers into one person expertise, Gandluru stated.
