Dell Technologies, NVIDIA optimize AI Factory approach for CSPs

By Sean Kinney, Editor in Chief September 25, 2024

Collected at: https://www.rcrwireless.com/20240925/network-infrastructure/dell-technologies-nvidia-optimize-ai-factory-approach-for-csps

From core to edge, Dell AI for Telecom includes network optimization capabilities and supports AI services like GPUaaS

Back in May at Dell Technologies World, the ICT giant unveiled its AI Factory with NVIDIA solution which holistically covers data, infrastructure, ecosystem, strategy and use cases. Today the company announced Dell AI for Telecom, an iteration of the product specifically designed to help communications service providers (CSPs) quickly put together the pieces needed to leverage artificial intelligence (AI) for opex reduction by way of increased network automation and monetization through the delivery of new, advanced services. 

In an interview with RCR Wireless News, Dell Vice President of Product Management for the Telecom Systems Business Andrew Vaz explained the value proposition across three dimensions: enabling CSPs to run any AI workload anywhere in their network, using a validated, turnkey hardware/software/services bundle to accelerate AI-based business outcomes, and draw on Dell’s partner ecosystem to realize network optimizations in the core, RAN and out to the edge. 

“AI is a horizontal technology that can be applied to use cases across the network,” Vaz explained. He stressed that the flexible solution covers a full gamut of telco-specific use cases “all the way across the network, from edge all the way to the backend of the network in the core data centers.” 

On the infrastructure side, Dell is bringing a range of AI servers featuring NVIDIA GPUs, including its XE9680, R760xa and PowerEdge XR8000. That last one is tailored for a range of environmentally-intensive edge deployments and can support six NVIDIA L4 Tensor Core GPUs in an ultra-dense 2U form factor. ”We have systems of all sizes, cost points, environmental considerations, that you can literally put at different points of the networks,” Vaz said. “From an infrastructure perspective, we really are a one-stop shop, which is helpful.” 

With a number of partners, Dell AI for Telecom is coming to market with a focus on CSP initiatives around network modernization, AI monetization and customer experience (CX). On the modernization point, Dell has partnered with Kinetica which has built a large language model (LLM) written for telco-specific database queries to facilitate faster troubleshooting and analysis. With specialist Synthefy, the product can be used to stand up network digital twins to support predictive maintenance. Vaz said early results show a 50% improvement in time to root cause analysis with the integrated Kinetica tooling, and a 90% accuracy rate in identifying network component failures with the Synthefy’s digital twin-based approach. 

With regard to AI monetization and CX, Dell is working on the former with Indonesian ICT provider Lintasarta on GPU-as-a-service, which essentially allows enterprises to access third-party AI computing resources following a consumption-based model. Vaz pointed out that governmental data sovereignty requirements could help foster the GPUaaS market in a way that plays to CSPs existing strengths. 

Another component of Dell AI for Telecom is a CX solution co-developed with SK Telecom; essentially the Dell/NVIDIA AI suite, combined with an generative AI chat agent tuned to CSP needs, is integrated into existing BSS software for improved customer engagement and revenue growth. 

For more on the Dell AI Factory with NVIDIA, check out this video interview filmed at the Digital Transformation World—Ignite event held in Copenhagen, Denmark, in June. And for a closer look at the gen AI BSS solution developed with SK Telecom, watch this video also from DTW—Ignite. 

Whichever direction a CSP chooses to go with its AI strategy, Vaz said it has to start with the use cases. “What do you want to do?,” he said. “There’s so many different things we’re seeing from the operators. It’s not a single set of things; it’s multiple. The [radio access network] guys want certain things—they want ease of troubleshooting, power efficiency…The core and OSS/BSS guys are looking at when a customer calls me with a problem, and they pay for a certain policy for us, maybe I should offer them a new policy dynamically based on what they’re saying…How do I do that? Getting clear on which use cases you want is critical.” 

He also made clear that the quality of AI-related outcomes is based on the quality of data fed into an AI solution. “It’s really important to implement a robust data strategy,” Vaz said. “The analogy that I like about AI Factory with a real factory is…when you have a real factory, you’re inputting a bunch of material to go build something and have an outcome. You have to feed the factory. You’re feeding it with data now, so quality of ingredients matters.” Architecturally, and from a cost standpoint, “You want to bring AI to the data, as opposed to sending all your data to the AI.” 

For a deep dive on Dell’s perspective on, and solutions for, CSPs, check out this site. 

Dell Telecom Systems Business SVP and GM Dennis Hoffman summed it up in a statement: “Capitalizing on the multiple opportunities presented by AI has become the most compelling driver of network cloud transformation. Dell AI for Telecom brings together Dell’s AI expertise and infrastructure, with partners across the ecosystem, to help network operators implement AI solutions in and on the network that reduce opex, improve performance and create new edge revenue opportunities.”