Paper Title: Implementing Computational Storage in an NVMe-oF-Based System

Paper Abstract: NVMe-oF has emerged quickly as the best way to network storage and make flash resources available anywhere in a system. However, it can also serve as the interconnect for a computational storage system, in which the resources include both compute and storage. The advantages are greater scalability, lower burden on central computing resources, lower network capacity requirements, and fewer large data transfers. Typical applications include offloading storage services, such as encryption, erasure coding, or RAID implementations, large IoT systems, cybersecurity, and advanced AI computing.

Paper Author: Patrice Couvert, Technical Product Manger, Kalray

Author Bio: Patrice Couvert serves as Technical Product Manager for the Kalray Data Center Business Unit. With over 20 years of industry experience, Patrice leads product definition, creates white papers and marketing materials, and ensures that products meet customer demands and expectations. He regularly delivers presentations at high-tech conferences, including the Flash Memory Summit and SNIA Storage Developer Conference. Before joining Kalray, Patrice worked on power management, system architecture, and microcontroller design at ST Microelectronics. He holds a patent in high-speed packet processing.