Please ensure Javascript is enabled for purposes of website accessibility

Service Providers’ GPUaaS Opportunity

DriveNets’ Senior Director System Engineering, David Watson, shares how DriveNets empowers service providers to monetize AI with automated, hyperscaler-inspired network operations and a high-performance, multi-tenant AI fabric enabling scalable GPU-as-a-Service.

How DriveNets empowers service providers

Full Transcript
Yeah, it comes down to a couple of things.

First, it’s the highest performing AI fabric on the market today. So when we start looking at where it came from, this technology, our AI fabric, is built off a proven DriveNets Denos operating system, which has been deployed at many customers globally from that perspective. So this allows for not only them to deploy at scale and be revenue generating, but it allows for one, multi-tenancy, which is critical for a GPU as a service, allowing them to basically scale not per customer, to scale the common infrastructure across all their customers.

Now we allow this through end-to-end VOQ scheduling within the fabric itself and not relying on any external technologies to accomplish that. Once again, simplifying the deployment for our customer from there. The last thing is it’s deployed.

So when you go through and look at it, the solution’s deployed across many NeoClouds at this point, some LLM providers that’s in test, and it’s based off common technology deployed at some of the largest tier ones in the globe.

Continue reading

eBook

DriveNets AMD System Reference Architecture

This Reference Architecture (RA) document provides an end-to-end blueprint for building a high-performance AI GPU cluste ...

Read more

White Papers

System-Level Optimization of AI Infrastructure with AMD Instinct GPUs

Extracting maximum value from an AI cluster requires an uncompromising, end-to-end system-level optimization journey.

Read more

White Papers

Faster LLM Inference on AMD Requires Rethinking All-Reduce

Extracting maximum ROI from AMD AI clusters requires moving beyond out-of-the-box software bottlenecks.

Read more