Paper Title: Approaches that Combine Optane with QLC Offer Solutions Competitive with TLC
Paper Abstract: QLC media is optimized for $/GB but its lower endurance makes it suboptimal for write-intensive and latency/QoS sensitive workloads. Optane technology is a great complement to Intel QLC technology as it delivers high endurance, low latency, and QoS consistency. Based on the insights from empirical data, we discuss three software approaches that combine Optane SSDs with QLC SSDs to deliver endurance and latency/QoS consistency at cost competitive with TLC. The first approach discusses best practices to separate short lifetime data to Optane while keeping the longer lifetime data on QLC. The second approach comprises a set of selective insertion policies at block and file levels that use Optane's higher available endurance for storing blocks/files responsible for consuming most P/E cycles on QLC while evicting those blocks/files to QLC according to lifetime. The final approach discusses two-layer architecture in which the bottom layer uses Optane for power fail safety and XOR for Zoned Namespace (ZNS) QLC while the top layer reuses the second approach. These approaches deliver benefits compared to TLC that span file systems, databases, and KV stores popular in cloud deployments.
Paper Author: Kapil Karkra, Principal Engineer, Intel
Michal Wysoczanski,
Software Architect,
Intel
Piotr Wysocki,
,
Intel
Author Bio: Kapil Karkra is a Principal Engineer and the lead software architect in the Intel Nonvolatile Memory Solutions Group responsible for the architecture of Virtual RAID on CPU (VROC), Volume Management Device (VMD), NVMe*, and Caching Acceleration Software (CAS) drivers. He also leads host software advanced development activities. Kapil actively drove the proposal for enclosure management (NPEM) in the PCIe* base specification for the NVMe ecosystem. His current focus is around improving QLC performance and endurance using Intel Optane technology. Kapil has spent his entire career advancing storage architectures (e.g., RAID, caching, and storage interfaces) at companies like IBM, Philips, and Intel. He has over 20 years of storage experience and holds two patents with 12 more pending. Kapil holds a bachelor’s degree in Electrical Engineering from National Institute of Technology (NIT) in India and an MBA from Arizona State University.
Author 2 Bio: Michal Wysoczanski is a software architect and tech lead in the Intel Nonvolatile Memory Solutions Group working on Caching Acceleration Software (CAS). His work is focused on caching software algorithms especially in storage virtualization and software defined storage deployments. His current focus is around improving QLC performance and endurance using Intel Optane technology. Michal has 15 years of experience in software development especially in Linux operating system environment. Michal holds a bachelors degree in Computer Science from Koszalin University of Technology in Poland.
Author 3 Bio: Piotr Wysocki is a software architect in the Intel Nonvolatile Memory Solutions Group responsible for driving pathfinding activities for Virtual RAID on CPU (VROC), Volume Management Device (VMD), Caching Acceleration Software (CAS), SPDK based Flash Translation Layer module for ZNS SSDs, and Open Channel SSDs. His current focus is on developing algorithms for data placement and data movement between QLC and Intel Optane SSDs to optimize performance and endurance of a collection of drives. He holds two patents with 22 more pending. Piotr has spent 8 years of of his career at Intel advancing memory and storage technologies. In his previous jobs he worked on embedded software and FPGAs. Piotr holds a masters degree in Electronic Engineering from Gdansk University of Technology in Poland.
|