Clear Ventures, General Catalyst, Hitachi Ventures, Liberty Global, Lightspeed, NVIDIA, Prosperity7 Ventures invest total of $30 million in hyperscale network fabric specialist Arrcus
With a recent $30 million round of funding, including $15 million from NVIDIA, Arrcus CEO Shekar Ayyar detailed two priorities: porting Arrcus software to NVIDIA’s platforms and increasing customer engagement and go-to-market partnerships.
In an interview with RCR Wireless News, Ayyar discussed the synergies between NVIDIA’s position in the red hot artificial intelligence (AI) market and his firm’s capabilities around tightly coupled distributed computing architectures. “People are realizing there is going to be a constrained set of compute resources…that they have got to work with to extract out the maximum efficiency from that architecture,” he said.
“You can sort of throw GPUs at a problem,” he continued, “but that’s not actually always the answer. You’ve got to be smarter about how you bring compute nodes together.” He said the Arrcus networking fabric complements what NVIDIA does with its MGX architecture for modular server designs. The combination, he said, “lets our customers leverage the best of compute resources wherever they’re distributed.”
As for what this means for communications service providers (CSPs), Ayyar said it’s a chance for them “to jump in and see how they could leverage their distributed infrastructure for something meaningful and useful. We provide the right kind of software” to seamlessly connected compute resources in the public cloud, colocation facilities and on premises data centers.
“You take these three different places where…high performance compute can reside…then what is missing in the equation today is a single network fabric that connects these,” he said. “You now have the ability to connect the dots between these.”
Ayyar referenced work Arrcus is doing with Aramco Digital around multi-cloud networking to support “common visibility across their assets…Arrcus can do that. So an operator can take that and essentially productize that as a service. That’s almost an easy here and now thing. Then you can go in an do more value-added interesting and complex things like provide your….customer a low latency network that is 5G-based that allows for better and more efficient traffic flows…We can minimize costs in terms of cloud transitions whether that’s data migrations or data transactions.”
Interestly, what Ayyar is described also came up at the recent TM Forum DTW—Ignite show in Copenhagen. A panel of operator CTOs discussed smart traffic management from the public cloud and into 5G networks as a way for the end user to save on cloud egress fees, among other nascent service areas.
He also called out Softbank as “an example of an operator that gets it.” Arrcus and Softbank are collaborating on automated network slicing and multi-access edge computing infrastructure with the use of SRv6 Mobile User Plane (SRv6 MUP). The high-level concept with SRv6 MUP is to streamline the integration and interaction between mobile networks and IP routing. Softbank is “leading the charge there and they’re enabling these types of services and solutions for their customers. I just wish others would be similar across the globe.”
Big picture, Ayyar said, “Everybody talks about AI but it almost immediately goes down to how many GPUs are you buying and what does your stack look like. That is the right first thing, but then you need to go and graduate from there and say, ‘How are you building the right network architecture to leverage the power that you have either bought or created?'”
“Everybody talks about AI but it almost immediately goes down to how many GPUs are you buying and what does your stack look like? That is the right first thing but then you need to go and graduate from there and say how are you building the right network architecture to leverage the power you have either bought or created.” then how do you make it efficient. Inferences is where really the dollars ultimately come from; people pay for results…
