In collaboration with industry peers, Pure Storage commits to expanding the capabilities of high-performance Ethernet for large-scale AI and HPC initiatives.
Pure Storage has announced that it joined Ultra Ethernet Consortium (UEC), a Linux Foundation initiative dedicated to building a complete, open and accessible Ethernet-based architecture to accelerate critical, data-intensive AI and HPC applications.
As the leading data storage platform serving the most advanced enterprise AI initiatives, Pure Storage commits to further defining and integrating UEC technology standards, while building and delivering a forward-compatible platform to optimise performance for enterprise AI and HPC workloads using Ethernet, supporting rapid innovation and faster time to market.
As enterprises strive to achieve their AI goals, they face significant challenges with existing network solutions that are difficult to manage and scale, hindering the flexibility and efficiency needed for growing AI workloads.
To address these challenges, UEC aims to advance Ethernet technology, offering a scalable and efficient solution to support AI innovation.
Ethernet’s increased adoption in data centres, due to its lower TCO, extensive interoperability and proven reliability, has made it the foundation for many of the world’s largest AI clusters. The advancement of new Ultra Ethernet standards will allow enterprises to maximise their existing AI and HPC investments while deploying performance-intensive applications, driving innovation, boosting productivity and enhancing operational efficiency.
By contributing to the standardisation and growth of high-performance Ethernet for large-scale AI and HPC initiatives, Pure Storage will accelerate platform development to support UEC standards long-term, delivering:
- Faster time to innovation: As AI and HPC workloads are accelerated, AI infrastructure powered by Ultra Ethernet will enable lower job completion times and higher frequency iterations that speed model building, training and inference, translating into faster time to innovation, faster time to market and shorter time to insight.
- Cost-effective, efficient operations: Faster compute, network and link speeds, as well as lower latency and higher network utilisation provides enterprises with high performance AI and HPC networking and storage built on cost-effective and easy to manage Ethernet-based switches and storage.
- Enhanced choice for customers: Customers will have greater choice between IB and Ethernet networks, allowing them to extend their Ethernet networking investments into the AI and HPC applications they are adopting.
Maciej Kranz, GM, Enterprise, Pure Storage, said: “Since its inception, Pure Storage has committed to delivering flexible, reliable and high-performance data storage technologies and services to simplify the enterprise data experience. Joining the UEC provides Pure with the opportunity to work alongside leaders in compute, networking and storage to drive the development of a new ultra-Ethernet architecture for the most advanced AI and HPC workloads.”
Hugh Holbrook, Chief Development Officer, Arista Networks, said: “The UEC is a leader in driving the evolution of Ethernet technologies, accelerating innovation and setting new benchmarks for performance, reliability and scalability in networking. As a member of the steering committee, Arista strongly supports the UEC, and Arista is excited to have Pure Storage’s deep data storage expertise, in collaboration with other industry leaders, to advance Ethernet standards and ultimately empower organisations to meet the ever-growing demands of AI and HPC.”
