Paper Title: Storage Engine for 3D XPoint and 3D QLC NAND SSDs

Paper Abstract: 3D XPoint fills the performance gap between DRAM and NAND-based SSDs, while 3D NAND QLC SSDs reduce the cost gap between SSDs and hard drives. Combining these two technologies will improve performance and reduce the total cost of ownership (TCO). A new key-value storage engine is introduced that was developed and optimized for managing 3D XPoint and 3D NAND QLC technologies. This hybrid solution is independent from applications and services, allowing the best tuning for performance, QoS, and endurance. A TCO model based on $/GB will be introduced. Furthermore, the ratio of 3D XPoint (here is Intel Optane SSD or equivalent of fast storage SSD) to 3D NAND QLC SSDs controlled by the engine is configurable to address a variety of different workloads. We will also show performance data (IOPS and QoS) Optane+QLCs vs TLCs SSD.

Paper Author: Jack Zhang, Cloud Storage Architect, Intel

Author Bio: Jack Zhang is a chief cloud and enterprise architect at Intel NVM (non-volatile memory) solution group (NSG). He manages and leads SSD solutions/optimizations and next generation Optane solutions for Cloud and Enterprise. He also leads SSD solutions/optimizations for open source storage solutions for software-defined storage, OpenStack, Ceph, and big data. His recent focus is next generation storage at Tier 1 CSP based on 3DX point + 3D QLC NAND SSD. Jack holds Master degree of Computer & Electrical engineering from University of Dayton, He has more than 25 years of design experience of firmware/hardware, software/kernel/driver, system architectures, as well as new technology ecosystem enabling, and market development.