Achieving Success for HPC and AI-Driven Business Outcomes - Paul Bloch, DDN

DDN's Market Positioning & Strategy - DDN is recognized as a key player in high-performance computing, particularly by Nvidia, who has been using DDN exclusively for the past eight years [1] - DDN's technology is integral to Nvidia's testing and development, including platforms like Selene A100, Eos H100 (4,000 GPUs), and GB200 [1] - DDN emphasizes its ability to scale solutions from small implementations (2U) to massive deployments (100,000+ GPUs), validated at 100% [1] - DDN focuses on investing in R&D, engineering talent, and feature development for both Exascaler and Infinia, reinvesting customer dollars back into the company [2] Technological Advantages & Solutions - DDN's solutions offer better GPU efficiency through checkpointing, data loading, and data crunching, with significantly faster write performance compared to competitors [2] - DDN's architecture simplifies deployments with fewer network ports, enhancing stability and scalability, avoiding full mesh requirements seen in competing solutions [1] - DDN provides online upgrades and enhanced visibility into workload and potential issues at the cluster level, extending beyond storage to include network and GPU monitoring [2] - DDN's systems are fully balanced, ensuring that performance scales linearly with added units, aggregating performance and access as the system expands [2] Customer Success & Partnerships - Jump Trading, a high-frequency trading firm, deployed half an exabyte of DDN's platform after switching from competitive technologies [2] - DDN is partnering with Nvidia cloud providers (NCPs) to deliver AI in the cloud as a private cloud solution, offering control over data and latency [2] - Scaleway, an NCP, has found that DDN maintains consistent performance at scale, without issues related to metadata or object size limitations [2] Addressing Industry Trends - The industry is experiencing an accelerated pace of technology change, with new chips emerging every six months to a year, requiring faster time to data, resolution, and production [1] - The scale of deployments is increasing rapidly, with discussions now commonly involving 100,000 to 500,000 GPUs, requiring infrastructure that can handle this scale [1] - Customers demand rapid deployment, expecting systems to be up and running within 60 days or less, emphasizing the need for quick time to results [1]