Session: Innovating DC Physical Infrastructure for AI and HPC at Scale
The rapid growth of AI and steady demand for HPC has heightened the demand for computational power, necessitating innovative data center infrastructure solutions. The Open Compute Project (OCP) Community provides a framework to efficiently accommodate sustainable large-scale computational infrastructure tuned for AI. This session explores leveraging OCP tenets to address AI and HPC workload challenges in data centers, focusing on networking, DC densification, and sustainability.
Participants will examine the importance of networking, exploring the OCP Communities’ emphasis on high-speed networking, low-latency interconnects, and advanced fabric technologies for optimizing data transfer within AI clusters. Techniques like open network architectures will be discussed. Densification strategies for AI data centers will be explored, including rack-scale architectures, modular designs, and high-density computing solutions, enabling efficient cooling, power management, and scalability. The discussion will also highlight sustainability considerations, such as energy-efficient designs, renewable energy integration, and heat transfer approaches.
Topics like power management, waste heat recovery, and green certifications will be covered, showcasing sustainable practices in AI and HPC infrastructure. Attendees will gain insights into leveraging OCP to address AI and HPC workload challenges, optimize connectivity, maximize data center density, and promote environmental responsibility.
