[Paper Review] Improving the Reliability of Next Generation SSDs using WOM-v Codes

Shehbaz Jaffer, Kaveh Mahdaviani+1 more|arXiv (Cornell University)|Jul 23, 2022

Advanced Data Storage Technologies4 citations

TL;DR

This paper proposes WOM-v codes—a non-binary, voltage-based write-once-memory coding scheme—to significantly extend the endurance of next-generation QLC SSDs by reducing Program/Erase (P/E) cycles. Using a FEMU-based simulator and two key optimizations (efficient garbage collection and encoding), the authors achieve 4.4x–11.1x reduction in erase cycles with minimal performance overhead, dramatically improving SSD reliability and lifetime for real-world workloads.

ABSTRACT

High density Solid State Drives, such as QLC drives, offer increased storage capacity, but a magnitude lower Program and Erase (P/E) cycles, limiting their endurance and hence usability. We present the design and implementation of non-binary, Voltage-Based Write-Once-Memory (WOM-v) Codes to improve the lifetime of QLC drives. First, we develop a FEMU based simulator test-bed to evaluate the gains of WOM-v codes on real world workloads. Second, we propose and implement two optimizations, an efficient garbage collection mechanism and an encoding optimization to drastically improve WOM-v code endurance without compromising performance. A careful evaluation, including microbenchmarks and trace-driven evaluation, demonstrates that WOM-v codes can reduce Erase cycles for QLC drives by 4.4x-11.1x for real world workloads with minimal performance overheads resulting in improved QLC SSD lifetime.

Motivation & Objective

• To address the limited endurance of QLC SSDs, which suffer from significantly fewer P/E cycles than previous-generation SSDs.
• To design and implement non-binary, voltage-based WOM-v codes that enable multiple writes per memory cell without erasing.
• To evaluate the impact of WOM-v codes on real-world workloads using a high-fidelity FEMU-based simulator.
• To optimize WOM-v code performance through efficient garbage collection and encoding techniques.
• To demonstrate that WOM-v codes can extend QLC SSD lifetime with minimal performance overhead.

Proposed method

• Design and implement non-binary WOM-v codes that map multiple data writes to distinct voltage levels in memory cells, deferring erasures.
• Develop a FEMU-based simulation test-bed to emulate real-world SSD workloads and evaluate WOM-v code performance under realistic conditions.
• Introduce an optimized garbage collection mechanism tailored for WOM-v codes to reduce erase operations and improve write amplification.
• Implement an encoding optimization that improves code efficiency and reduces the number of required erase cycles.
• Use microbenchmarks and trace-driven evaluations to validate the effectiveness of the proposed optimizations.
• Measure performance overhead and erase cycle reduction across diverse workloads to quantify reliability gains.

Experimental results

Research questions

• RQ1 Can WOM-v codes effectively reduce the number of P/E cycles in QLC SSDs without incurring significant performance penalties?
• RQ2 How do WOM-v codes perform under real-world workloads compared to conventional SSD wear-leveling techniques?
• RQ3 What is the impact of optimized garbage collection and encoding on WOM-v code endurance and performance?
• RQ4 To what extent can WOM-v codes extend the usable lifetime of QLC SSDs in practical deployment scenarios?
• RQ5 What is the trade-off between code complexity, performance overhead, and erase cycle reduction in WOM-v-coded SSDs?

Key findings

• WOM-v codes reduce erase cycles in QLC SSDs by 4.4x to 11.1x across diverse real-world workloads, significantly extending device endurance.
• The proposed garbage collection optimization effectively reduces write amplification and minimizes unnecessary erasures in WOM-v-coded SSDs.
• The encoding optimization improves code efficiency, enabling higher data density and lower erase frequency without increasing complexity.
• Performance overhead from WOM-v codes remains minimal, with negligible impact on I/O latency and throughput in microbenchmarks.
• Trace-driven evaluations confirm consistent gains across multiple real-world workloads, validating the scalability and robustness of the approach.
• The combination of WOM-v coding and optimizations results in a substantial improvement in SSD reliability and lifetime, making QLC SSDs more viable for long-term storage.