“True AI” is all about building networks that support efficient training and inferencing: Arrcus CEO
Arrcus, a hyperscale networking software company and a specialist in core, edge and multi-cloud routing and switching infrastructure, has focused on developing a narrative around and products for creating the intelligent data fabric that will connect distributed artificial intelligence (AI) workloads.
In a recent interview with Shekar Ayyar, CEO of Arrcus, during Mobile World Congress 2024 in Barcelona, Spain, the executive noted that most of the industry focus is on generative AI-enabled applications. “Not too many people are thinking about the infrastructure underneath. And very quickly we’re going to find out that that infrastructure is, in fact, not sufficient for what you need to do true AI.”
“And true AI is really doing efficient training, but also efficient inferencing. You’ve got to be able to take everything that happens in these stovepipes of data centers and then bring them into multiple data centers that are connected together and then from there to the edges,” Ayyar added.
Ayyar explained that this would require a network that is flexible and that can extend across from core to edge as well as across multiple clouds. “And so at Arrcus, what we do with our operating system in particular is to actually have that distributed fabric and then bring this to the benefit of telecom carriers, enterprises, as well as cloud companies so that you can have this interconnected data center architecture for AI,” the executive said.
Operators should realize that they’re sitting on an asset which is incredibly valuable in their geographically dispersed points of presence. “And each one of these things can in fact be converted to a compute and network node that then becomes an intelligent connected point in the AI architecture.”
“Now, unfortunately, operators historically have just handed over the keys to somebody, and whether that key is to a hyperscaler cloud or to other forms of intermediaries that are dealing with enterprises. And part of the challenge is that they have not really had the marketing muscle or the sales and go-to- market muscle to be able to take services that are other than just voice and data to the customer. But this is an opportunity to doing that now with the 5G network relatively getting mature. And then on the other side, technologies like Arrcus coming to their benefit in order to do things like multi-cloud, multi-edge transformation—they can take this and then build out their network ground up into something that then becomes useful for deployment of artificial intelligence,” said Ayyar.
During MWC 2024, the company made a number of announcements including the enhancement of its ACE-AI platform to address the growing demand for unified networking fabric for distributed AI workloads. The platform supports new industry-leading Broadcom 800G switching platforms.
The company also highlighted its collaboration with SoftBank in advancing automated network slicing and Multi-Access Edge Computing (MEC) infrastructure, which was possible through the utilization of SRv6 Mobile User Plane (SRv6 MUP) technology.
And further, Arrcus unveiled a strategic collaboration with Red Hat to deliver innovative 5G, edge, and multi-cloud networking solutions
