Paper Title: Architecture of an Ordered Key Value Computational Storage Drive in a HPC System

Paper Abstract: KVCSD is an Key-Value interfaced Computational Storage Drive which supports key order and designed to efficiently support complicated functions such as range query. Scientific data is typically persisted in an order that is not the same as retrieval, creating the need to budget time and resources for costly, massive sorting operations. Additionally, data queries in HPC enabled science almost always require ordered retrieval and frequently return many orders of magnitude less data than the full data source. SK hynix has developed in collaboration with LANL, a new KVCSD which enables faster and more efficient HPC workflows and addresses inefficiencies of server based data queries of current HPC storage systems by moving the ordered queries to near the storage devices. We will share the benefits of Key-Value interfaced CSD in terms of performance and reduced cost in a HPC environment through evaluation results of the HPC storage system integrated with a KVCSD prototype

Paper Author: Woo Suk Chung, Storage Software Team Lead, SK hynix

Author Bio: Woo Suk Chung is a team leader in the advanced storage architecture team at sk hynix, focusing on the development and implementation of future solution for improving perfomance on enterprise SSD and storage system.