Edgecore Networks has deployed high-capacity, open-networking switches for HyperVerge to handle AI training and IP storage workloads. Utilizing Broadcom hardware and the SONiC OS, the solution delivers strict latency control and allows HyperVerge to rapidly self-configure the system, demonstrating the speed and flexibility of open networking.
Besides the switch design meeting bandwidth and feature requirements, open-networking and publicly available standards enabled the HyperVerge team to independently and quickly set up configurations and performance dashboards – a true testament to the benefits of an open ecosystem.
“For AI training workloads, storage bandwidth and latency are as critical as compute. Any bottleneck directly translates into underutilized GPUs. At HyperVerge, we unified all our NVMe storage into a single clustered parallel file system, and Edgecore’s 100G RoCEv2 lossless RDMA fabric powered by SONiC enables us to deliver this storage at near-local NVMe latency and bandwidth across our GPU cluster over the network. This infrastructure powers training workloads for our face recognition, OCR, and custom LLM models used in identity verification, digital KYC, and underwriting intelligence,” said Manideep Kolla, Head of Identity AI at HyperVerge.
“As AI applications find adoption in enterprises, both effective and efficient solutions are required for customer success” said Nanda Ravindran, VP of Product Management and Planning at Edgecore Networks. “By using Edgecore’s high-throughput data center design coupled with the SONiC open-networking NOS, the team was able to quickly design a solution for the HyperVerge datacenter storage cluster. Most notably, the team worked effectively and independently by leveraging open networking public content and ready-to-use frameworks.”
This collaboration underscores the Edgecore Networks commitment to advancing open-networking technologies and delivering high-impact networking solutions for next-generation companies.
