Paper Title: Implementing Computational Storage with Existing SSD Controller Resources

Paper Abstract: Computing and storage needs are ever expanding in the enterprise applications. The NVMe SSD controllers are evolving in functionalities and incorporating various computing resources such as multiple co-processors, hardware accelerators, additional memories, peripherals, etc.. However, these computing and storage resources within a SSD may not be utilized at all times. These resources may be leveraged to offload the host CPU, or remote CPUs in a NVMe-over-Fabric topology to improve the overall system performance, TCO. The NVMe standard also provides mechanisms to take advantage of these resources with compute ability. We will look at an example SSD solution that demonstrates the various ways of identifying and utilizing these compute storage resources.

Paper Author: Ramyakanth Edupuganti, Staff Applications Engineer, Microchip Technology

Author Bio: Ramyakanth (Ram) Edupuganti is working as an Applications Engineer for NVMe controllers in the Datacenter solutions business unit of Microchip Technology. He has contributed to the development of NVMe SSDs and storage products working closely with several enterprise customers. With his passion for new technologies and his Master’s degree in Computer Science engineering background from Sac State university, Ram has special interests in developing newer applications and innovative solutions for challenging problems. Previously he has presented at Flash Memory Summit on "use of Controller Memory Buffers in NVMe Controllers", and now he is excited to share his findings on how the NVMe controller hardware resources may be used for Compute Storage.