Overview
Explore a technical conference presentation from the SNIA Compute+Memory+Storage Summit that delves into dynamic computational partitioning for storage systems. Learn how recent hardware developments have introduced diverse compute units while bringing network and storage bandwidths closer to memory subsystem capabilities. Discover the concept of "decomposable queries" and their implementation in a prototype computational storage system (CSS) that enables dynamic, just-in-time partitioning of computation across storage device hierarchies. Examine the challenges and findings from microbenchmark evaluations in bioinformatics use cases, including unexpected performance issues with Seagate's research KV drives. Understand how this approach allows for transparent integration of future computational storage device (CSx) improvements and accommodates new device types, making it suitable for production deployment.
Syllabus
SNIA CMSS23 - Decomposing Compute to Grow Computational Storage
Taught by
SNIAVideo