Paper Title: Cloud Storage Accelerate Layer (CSAL): Enabling Performance and Capacity Scaling

Paper Abstract: Cloud service providers (CSPs) offer a range of storage services. CSPs tend to extend storage capacity to meet requirements of high-growth technologies, such as big data analytics that require not only high performance but also expanding data volumes. Advantages of high-density NAND flash enable both large capacity and high performance. However, endurance and write performance per TB of NAND flash are decreasing. We propose CSAL, a Flash Translation layer (FTL) in SPDK for new types of media tiers with Intel Optane and QLC/ZNS Flash. CSAL utilizes high performance and endurance Intel Optane media as performance tier to accelerate QLC/ZNS NAND flash as capacity tier. The key insight is to shape random writes to NAND/ZNS friendly writes. CSAL not only eliminates internal fragmentation of the NAND SSDs, but also absorbs frequently updated user writes in the frontend tier thus improving system-level endurance. Cloud tenants usually have different I/O behaviors resulting in performance unpredictability. In this presentation, we will show and analyze the real world I/O behaviors from big data services of Alibaba Cloud to demonstrate predictable performance using CSAL.

Paper Author: Mariusz Barczak, Software Architect, Intel

Author Bio: Mariusz joined Intel in 2013.Currently he is a cloud software architect in the Xeon Memory Group (XMG). His work area is finding innovations for storage software running on Xeon. In particular caching solutions, software defined storage, virtualization, and storage analytics. His recent work is focused on mixed media solutions combining Intel Optane technology with other storage components like QLC SSD drives, to deliver efficient and durable storage. Mariusz has 12 years of experience in software development especially in Linux operating system environment.