Read disturb is a circuit-level noise in high-density solid-state drives (SSDs), which may corrupt existing data in SSD blocks and then cause high read error rates. The approach of read reclaim (RR) is commonly used to avoid read disturb errors by migrating the valid data pages to other free blocks for resetting negative effects of read disturbs, but it affects both I/O responsiveness and SSD lifetime. This paper proposes SWEEP to minimize the number of RR operations. Specifically, it gathers hot read and hot write data pages together, and saves them in a portion of designated blocks. After that, the hot write data pages will be invalidated and the hot read data pages in such blocks are likely to be migrated to other blocks through garbage collection (GC). Consequently, the read disturbs on these hot read data pages will be reset, without additional RR operations. Trace-driven simulation experiments show that our proposal can significantly reduce the number of RR operations by between 1.6 % and 74.4 %, which contributes to a maximum reduction of 29.7 % on total erase operations, indicating a better lifetime of SSD devices. In addition, it can reduce the overall I/O response time by up to 52.9 %, compared to existing optimization schemes for SSDs.

Reliable cloud-based storage systems require accurate failure prediction for Solid-State Drives (SSDs) and Hard Disk Drives (HDDs) to reduce data loss, enable proactive maintenance, support service-level reliability, and lower operational costs. In this survey, we review over 150 prior studies on storage failure prediction and related tasks, and provide a structured overview and evaluation of currently available techniques for storage failure prediction, spanning traditional statistical methods, machine learning, and deep learning approaches. We focus on device-level predictions and compare the performance, constraints, and implementation overhead of prior works in real-world scenarios. Challenges such as data imbalance, fail-slow degradation, and evolving failure patterns are discussed to identify current research gaps, such as the limited interpretability of advanced models, and the need for standardized benchmarks. Our main contribution is the introduction of structured decision frameworks that guide practitioners to choose suitable evaluation metrics, predictive models, and data preparation methods based on certain operational scenarios. These frameworks are complemented by comparative analysis of models, evaluation metrics, interpretability methods and computational overhead across deployment contexts. Our survey discusses open challenges and research directions in the domain, and offers useful insights and a structured methodology for translating research into practical deployment strategies.

In recent years, the rapid development of commercial satellite projects, such as low-Earth orbit (LEO) communication and remote sensing constellations, has driven the satellite industry toward low-cost, rapid development, and large-scale deployment. Commercial off-the-shelf (COTS) components have been widely adopted across various commercial satellite platforms due to their advantages of low cost, high performance, and plug-and-play availability. However, the space environment is complex and hostile. COTS components were not originally designed for such conditions, and they often lack systematically flight-verified protective frameworks, making their reliability issues a core bottleneck limiting their extensive application in critical missions. This paper focuses on COTS solid-state drives (SSDs) onboard the Jilin-1 KF satellite and presents a full-lifecycle reliability practice covering component selection, system design, on-orbit operation, and failure feedback. The core contribution lies in proposing a full-lifecycle methodology that integrates proactive design—including multi-module redundancy architecture and targeted environmental stress screening—with on-orbit data monitoring and failure cause analysis. Through fault tree analysis, on-orbit data mining, and statistical analysis, it was found that SSD failures show a significant correlation with high-energy particle radiation in the South Atlantic Anomaly region. Building on this key spatial correlation, the on-orbit failure mode was successfully reproduced via proton irradiation experiments, confirming the mechanism of radiation-induced SSD damage and providing a basis for subsequent model development and management decisions. The study demonstrates that although individual COTS SSDs exhibit a certain failure rate, reasonable design, protection, and testing can enhance the on-orbit survivability of storage systems using COTS components. More broadly, by providing a validated closed-loop paradigm—encompassing design, flight verification and feedback, and iterative improvement—we enable the reliable use of COTS components in future cost-sensitive, high-performance satellite missions, adopting system-level solutions to balance cost and reliability without being confined to expensive radiation-hardened products.

High-speed non-volatile memory express (NVMe) solid-state drives (SSDs) are shared by multiple tenants in cloud scenarios to improve resource utilization. Tenant-level I/O management is necessary to achieve reliable QoS control during sharing. Unfortunately, our investigation finds that CPUs inevitably participate in I/O management for existing solutions because SSDs are not tenant-sensitive and have limited internal computing resources. It introduces additional CPU costs and latency overhead when serving future faster SSDs. We propose that the Field Programmable Logic Gate Array (FPGA) is a promising alternative for freeing tenant-level I/O management from CPUs. However, implementing tenant-level I/O management using the FPGA requires addressing the following challenges: (1) System compatibility and tenant identification. (2) Efficient FPGA workflows that will not become a bottleneck. (3) Fast I/O management workflow that introduces the lowest additional CPU costs and latency. This paper presents Zero2M, a novel CPU-free system designed to optimize the additional CPU costs and latency overhead in tenant-level I/O management for future faster NVMe SSDs. Zero2M proposes a dedicated FPGA-based NVMe controller to preserve system compatibility and identify tenants using the namespace mechanism in NVMe. It allows I/O management without modifying host software, which existing solutions cannot achieve. The parallelized and pipelined workflows are proposed in the controller to accelerate I/O command processing and prevent the controller from becoming a bottleneck for the I/O management workflow. The read/write speed of the Zero2M controller is 4.65 \(\times\) /4.92 \(\times\) faster than the state-of-the-art hardware-accelerated controller. The I/O management workflow is formulated as a novel parallelized and pipelined accelerator and integrated into the workflow of Zero2M's controller. It optimizes additional CPU costs and latency overhead for tenant-level I/O management. Experiments present that Zero2M reduces an average of 3.01 \(\times\) CPU usage while maintaining the lowest latency overhead (7.62 \(\times\) lower on average) compared to the state-of-the-art solution. It also removes the CPU dependency for tenant-level I/O management for the first time.

Write amplification factor (WAF) is a critical performance and endurance bottleneck in flash-based solid-state drives (SSDs). Multi-streamed SSDs mitigate WAF by enabling logical data streams to be written separately, thereby improving the efficiency of garbage collection. However, despite the architectural potential of multi-streaming, prior research has largely overlooked the design of write buffer management schemes tailored to this model. In this paper, we propose a stream-aware block-level write buffer management technique that leverages both spatial and temporal locality to further reduce WAF. Although the write buffer operates at the granularity of pages, eviction is performed at the block level, where each block is composed exclusively of pages from the same stream. All pages and blocks are tracked using least recently used (LRU) lists at both global and per-stream levels. To avoid mixing data with disparate hotness and update frequencies, pages from the same stream are dynamically grouped into logical blocks based on their recency order. When space is exhausted, eviction is triggered by selecting a full block of pages from the cold region of the global LRU list. This strategy prevents premature eviction of hot pages and aligns physical block composition with logical stream boundaries. The proposed approach enhances WAF and garbage collection efficiency without requiring hardware modification or device-specific extensions. Experimental results confirm that our design delivers consistent performance and endurance improvements across diverse multi-streamed I/O workloads.

Across all fields, experts strive to collect and analyze numerous data to extract meaningful insight. In response to this trend, Hadoop and Spark have emerged, and many organizations have adopted these platforms for big data storage and processing. In addition, data centers with powerful servers are constantly expanding to accommodate the increasing number of data, causing significant costs and environmental problems due to the tremendous energy consumption. Single board computer (SBC) clusters have emerged as a promising alternative for efficient computing. Most SBCs have adopted a microSD slot for data storage; thus effectively processing massive data has some limitations. However, the latest generation Raspberry Pi (RPi), model 5B provides a peripheral component interconnect express (PCIe) interface, enabling high-performance storage media, such as solid state drives (SSD). This paper extensively investigates the practicability and potential of SBCs for terabyte-scale big data processing. We build the SBC Hadoop cluster, adopting the most powerful, latest RPi 5B (8 GB of RAM) with a fast PCIe-based SSD via the PCIe interface, and perform six widely known benchmarks with a large (up to 2 TB) data size. Furthermore, this paper discusses challenges and suggestions, including the effects of input/output (I/O) throughput, central processing unit (CPU) overclocking, power supply, and trim command, which significantly affect SBC Hadoop performance. This comprehensive study concludes that integrating the enhanced computing of RPi 5B with unlocked I/O performance finally paves the way for a practical solution to real-world big data processing on SBC clusters.

Abstract Growing computational demands from seismic imaging applications at bp’s high-performance computing center require recurring infrastructure upgrades to deliver higher performance within fixed power and budget constraints. This article presents a comprehensive hardware-software codesign strategy for migrating a fast Fourier transform (FFT) intensive seismic imaging software framework from traditional central processing unit (CPU) architectures to graphics processing unit (GPU) accelerated platforms. The migration leveraged NVIDIA’s 8-way HGX-H100 nodes featuring fully connected NVLink and NVSwitch to support efficient distributed FFT computation, paired with high-capacity non-volatile memory express (NVMe) solid-state drives (SSDs) enabled by GPUDirect Storage to accelerate node-local input/output (I/O). The legacy Fortran codebase was completely rewritten using a dual-layer design: C++ and Compute Unified Device Architecture (CUDA) for performance-critical kernels to maximize compute and memory efficiency and Python for orchestrating inversion workflows to support rapid research prototyping. Performance validation on production datasets from the Thunder Horse and Herschel fields demonstrated up to 90× runtime speedup and 13× improvement in energy efficiency over CPU platforms, while numerical accuracy was preserved with mean relative errors below 1% for acoustic reverse time migration and full-waveform inversion applications against their CPU references. This successful migration demonstrated the performance and energy efficiency advantages of GPU computing for seismic imaging and provided quantitative justification for bp’s first production-scale GPU cluster.

The channel-level RAID implementation has been introduced to NAND flash-based solid-state drives (SSDs), to fight against channel failures. But, it suffers from unbalanced wear-outs and I/O workloads across channels, due to the nature of in-channel updates on data/parity chunks of data stripes, leading to a decline in I/O performance and a negative impact on the lifespan of RAID-enabled SSDs. This paper introduces an approach to yield wear-out and I/O balances, with the minimal overhead caused by location exchanges of data/parity chunks belonging to the same stripe, when fulfilling write requests or garbage collections (GCs). To this end, we build an assessment model for measuring the balance level of all SSD channels, and then trigger a location exchange of data/parity chunks in the same data stripe, if the exchange operation can be beneficial to wear-out and I/O balances. Besides, we introduce a scheme of pairGC to further minimize the overhead of data location exchanges. Specifically, it pairs the GC operations on different channels and conducts out-of-channel page moves, to achieve the goal of data location exchanges without additional cost. Through a series of emulation experiments on 8 disk traces of real-world applications, we show that our proposal can greatly improve I/O performance by 40.0 % on average, as well as noticeably balance I/O workloads over SSD channels and prolong the endurance of SSDs, in contrast to the state-of-the-art RAID optimization schemes inside SSDs.

External disks (henceforth referred to as disks) are commonly used data storage peripherals for hosts. Verifying the legitimacy of these disks is essential to mitigate security risks, such as privacy breaches and virus propagation, before initiating interactions with a host. To address this challenge, we propose <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">DiskPrint</i>, a novel replay-resistant, few-shot disk authentication system that relies on unintentional electromagnetic (EM) emanations from the internal components of disks. The core idea of <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">DiskPrint</i> is that EM signals emitted during data writing operations can reveal unique hardware discrepancies among different disks. Building on electromagnetic theory, we develop a theoretical model that links EM signals to the underlying electronic components of the disk, demonstrating the feasibility of extracting distinctive disk fingerprints from these emanations. We also propose a set of signal enhancement techniques aimed at mitigating EM interface noise and improving the signal-to-noise ratio (SNR) of the EM measurements. To further strengthen the security of <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">DiskPrint</i>, we introduce a device-agnostic, replay-resistant approach by incorporating randomness into the leaked EM signals. Real-world experiments with 60 disks, spanning both hard disk drives (HDDs) and solid-state drives (SSDs) from seven brands and 14 different models, indicate that <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">DiskPrint</i> achieves an authentication success rate exceeding 99.6% with only three registration samples. A robustness analysis confirms its stability over time, while a security evaluation shows its resilience against various attack scenarios.

As the demand for data storage grows, preventing hard disk failures becomes critical to ensure data security and business continuity. Traditional Remaining Useful Life (RUL) prediction methods often struggle with handling heterogeneous data and sample imbalances, particularly when applied to both hard disk drives (HDD) and solid-state drives (SSD). This paper proposes an effective RUL prediction and intelligent decision-making methodology tailored to multi-class HDD in data center applications. Using Backblaze’s publicly available SMART dataset, we apply this methodology to both HDD and SSD. To tackle data heterogeneity and imbalance, we adopt XGBoost for selecting the critical and common SMART attributes of both HDD and SSD, while utilizing a Transformer model for multidimensional time series analysis. For the training strategy, we combine pre-training and transfer learning. The model is first pre-trained on HDD data and then fine-tuned for SSD tasks, resulting in a unified multi-class RUL prediction model. Our experimental results demonstrate that this approach outperforms traditional models in several evaluation metrics, especially in scenarios where SSD samples are limited. At the system level, we develop an intelligent decision-making system that integrates data collection, model training, failure prediction, and visual display, enabling a closed-loop process from data collection to operational decision-making. This study presents an efficient solution for managing the health of multiple types of hard disks, offering high engineering application value.

This article presents a comparative analysis between two types of data storage devices: HDD (Hard Disk Drive) and SSD (Solid State Drive). Considering the growing demand for fast and efficient storage, this research examines the main differences between these technologies, addressing aspects such as performances, durability, power consumption, and cost. The results highlight that SSDs, although generally more expensive, offer significant advantages in terms of read/write speed, shock resistance, and energy consumption, making them ideal for applications requiring high performance. On the other hand, HDDs still remain relevant due to their lower cost per gigabyte and superior storage capacity. This study provides a comprehensive overview of the characteristics and benefits of each technology, aiding in making informed decisions regarding the choice of the most suitable storage type depending on each need.

E-government applications generate and process large volumes of heterogeneous data that demand high-throughput and low-latency computation. Although Hadoop MapReduce is commonly used for such tasks, its performance is often limited by disk I/O constraints and network delays during the shuffle phase. This study proposes a data address-based shuffle mechanism optimized for Hadoop clusters equipped with Solid-State Drives (SSDs), aiming to enhance data processing performance in e-government applications. The mechanism introduces three key components: address-based sorting, address-based merging, and pre-transmission of intermediate data, which collectively reduce disk I/O and network transfer overhead. Experimental evaluations using Terasort and Wordcount benchmarks demonstrate execution time reductions of 8% and 1%, respectively, with statistical significance confirmed through 95% confidence intervals. Scalability assessments on a simulated 50-node cluster and energy profiling further validate the approach, showing improved performance, reduced network congestion, and a 31% decrease in energy consumption compared to HDD-based systems. The findings establish the proposed mechanism as a cost-effective and efficient solution for large-scale data processing in public sector computing environments.

NAND-flash-based solid-state drives (SSDs) are under constant pressure to deliver higher storage density while minimizing power and performance overhead. As the number of bits stored per NAND flash cell has scaled from single-level cells (SLC) to triple-level cells (TLC) and soon to penta-level cells (PLC), the reduced voltage margins between cell states challenge data reliability, requiring stronger decoding techniques. To maintain reliability and correct error data bits, low-density parity-check (LDPC) codes are widely deployed on these high-density devices and can operate in two modes: hard decoding, which uses threshold-based bit decisions and is relatively power-efficient, and soft decoding, which leverages additional reliability information but imposes higher computational and energy costs. In practice, NAND flash controllers initiate soft decoding when hard decoding fails, thereby preserving data integrity at the expense of latency and power overhead. Current approaches employ read-retry tables to adjust reference voltages and maximize hard decoding success rates; however, such tables cannot fully address diverse bit-error patterns, often unnecessarily invoking soft decoding and incurring significant performance overhead. To overcome this limitation, we propose a novel LDPC-syndrome-based loss function that adaptively adjusts multidimensional reference voltages, significantly reducing unnecessary soft decoding triggers without relying on predetermined read-retry tables or iterative voltage adjustments. Experimental results demonstrate that our proposed loss function effectively reduces the soft decoding trigger rate and the number of page reads, substantially minimizing the performance and power costs associated with soft decoding.

Abstract Space radiation poses a significant threat to the reliability of spacecraft equipment, while traditional static space environment models exhibit substantial inaccuracies in calculating on‐orbit error rates. This study investigates single‐event functional interrupts (SEFIs) in commercial off‐the‐shelf (COTS) solid‐state drives (SSDs) operating in low Earth orbit utilizing real space environment data and AP9 model data. Using PHITS software, the internal radiation environment of the spacecraft was simulated based on external detector measurements and the AP9 model. Findings show that SEFI rates derived from measured spectra and the AP9 model at the 95% percentile closely match on‐orbit observations, whereas results based on spectra deviating from the real environment may differ by more than an order of magnitude. This research highlights the critical influence of the dynamic space environment on single‐event effects and emphasizes the importance of accounting for secondary neutron radiation in spacecraft design.

Cloud technology offers significant advantages; however, its high implementation costs and high hardware requirements pose barriers to small-scale deployments and educational institutions. This study addresses these challenges by investigating the performance of OpenStack deployed via DevStack on a single-node server equipped with an Intel Core i7 processor, 16 GB of RAM, and a 500 GB solid-state drive (SSD) under resource-constrained conditions. We implemented a resource tuning approach by turning off non-essential services (including Cinder, Heat, and Tempest) and adjusting Nova's memory configurations to minimize overhead. Real-time system monitoring was performed using Prometheus and Grafana to examine trends in CPU, memory, and swap utilization across three configurations: default, optimized (RAM=1024 MB), and minimalist (RAM=512 MB). Our empirical results show that the optimized setup enhances system efficiency, decreasing CPU use and memory usage from 86% to 70.90% while maintaining the ability to run up to ten virtual machines with varying operating systems (e.g., CirrOS, Ubuntu 24.04 Server LTS). However, the minimalist configurations, which aim for aggressive swap utilization and reach 100% swap saturation when running 8 VMs under idle workloads, consequently compromise overall system responsiveness despite lower CPU usage. Efficiency in this context is defined as conserving RAM and CPU usage without degrading basic system responsiveness. This highlights a critical trade-off between RAM conservation and overall system responsiveness. This research provides practical insights into designing cost-effective and lightweight OpenStack environments. It establishes a crucial threshold for memory optimization, preventing performance degradation caused by excessive swap usage, particularly in resource-constrained research settings.

In recent years, debates have been persisting about the future survival and development trends of Hard Disk Drives (HDD) and Solid-State Drives (SSD), the two most mainstream storage devices. One prevailing view is that SSD, with their overwhelming speed advantage over HDD, will eventually fully replace HDD. Another perspective argues that competition between mainstream storage devices extends beyond speed, and HDD strengths in other dimensions will secure their continued existence. This study begins with an analysis of the fundamental physical structures and working principles of both devices: first, examining the operational differences between HDD and SSD; next, analyzing the performance gaps caused by these differences; and finally, concluding their respective optimal application scenarios. Based on the market demand corresponding to these scenarios, the study further deduces the future development space and trends for these two mainstream products. Through this research, the author concludes that SSD and HDD each have unique advantages and disadvantages in different application scenarios. Users’ product choices depend on their practical needs and which specific performance metrics they prioritize. Therefore, the two technologies currently coexist in the market not as replacements but as complementary solutions, each dominating its specialized fields. Additionally, future advancements in hybrid storage technologies, feasible new storage mechanisms, and related fields may help HDD and SSD overcome their inherent weaknesses, unlocking opportunities for renewed vitality. These conclusions also provide valuable references for major storage device manufacturers, such as Western Digital, Seagate, Samsung, SK Hynix, in strategizing and planning future product directions.

As machine learning (ML) workloads grow in scale, the carbon impact of data storage is underexplored. Despite the dominance of solid-state drives (SSDs) in ML pipelines for their performance benefits, the environmental trade-offs with traditional hard disk drives (HDDs) are not well understood. We compare the performance and total carbon cost of SSDs and HDDs in ML training workloads. To evaluate carbon impact driven by storage, we use the MLPerf Storage benchmark along with carbon emissions data from two energy grids. We find that although SSDs have significantly higher embodied emissions, their lower operational carbon and faster runtimes make them more efficient for I/O-bound ML workloads—especially once data exceeds memory capacity. While SSDs generally amortize their carbon cost over time, often outperforming HDDs in total emissions, there are caveats. When considering regional energy mix, results suggest that carbon-aware ML infrastructure should consider workload size, memory constraints, and grid intensity—not just device specifications.

Solid State Drives (SSDs) based on the NVMe Zoned Namespaces (ZNS) interface can notably reduce the costs of address mapping, garbage collection, and over-provisioning by dividing the storage space into multiple zones for sequential writes and random reads. The Log-Structured Merge (LSM) tree, which is extensively used in key-value storage systems, converts random writes to sequential writes, hence a suitable scenario to utilize ZNS SSDs. However, LSM tree associated data significantly varies in lifetime due to the levels and merging mechanisms of the LSM tree. Therefore, without an accurate method to estimate data lifetime, data with disparate lifetimes may be placed in the same zone, thus causing low space utilization and high write amplification within the SSD. To address these issues, the article proposes two data overlapping aware optimizations to realize intelligent data placement: a zone allocation scheme and a garbage collection scheme. The key technique of these optimizations is an accurate data-lifetime estimation by considering both the associated tree level of the data and the data overlapping ratio between the data and those in the neighboring level. Using the estimation technique, the zone allocation optimization can place data with similar lifetimes in the same zone. Besides, the garbage collection optimization can reclaim zones in an adaptive manner based on overlapping ratios to reduce the amount of data migration. Experimental results demonstrate that the optimization schemes effectively reduce garbage collection-incurred data copy by average factors of 2.11× and 1.50× in comparison to a conventional work and a state-of-the-art work, respectively. Consequently, the proposed work successfully alleviates the write amplification effect by 18% and 6%, compared to the conventional work and the state-of-the-art work, respectively.

The Zoned Namespace (ZNS) interface transfers most storage maintenance responsibilities from the underlying Solid-State Drives (SSDs) to the host. This shift creates new opportunities to ensure fairness and high performance in multi-tenant cloud computing environments at both hardware and software levels. However, when applications with different workloads share a single ZNS SSD hardware, traditional fair queueing schedulers fail to achieve fairness due to their limited awareness of workload characteristics. Moreover, allowing multiple outstanding requests to access the device simultaneously improves resource utilization but often leads to significant I/O interference among these requests. This interference results in over-throttling, which subsequently degrades the performance of existing fair queueing schedulers. To address the above problems, this paper proposes an efficient and high-performance fair queueing scheduling scheme for ZNS SSD (ZNSFQ) on the host side. Firstly, ZNSFQ introduces a workload-aware fair scheduler that enhances fairness by accurately estimating the I/O cost for each application based on its workload characteristics. Secondly, to optimize performance while ensuring fairness, ZNSFQ designs a request dispatch parallelism adjuster. This adjuster manages the channel-level request dispatch parallelism for each application to minimize I/O interference. Finally, ZNSFQ employs a global adaptive coordinator to alleviate device-level I/O blocking, reducing tail latency and CPU consumption while satisfying fairness and performance. A comprehensive evaluation demonstrates that ZNSFQ significantly enhances fairness and performance compared to the latest fair queuing schedulers. In sequential access scenarios, ZNSFQ enhances fairness by over 38.13% and increases I/O bandwidth by more than 49.24%. Furthermore, in random access scenarios, it reduces CPU utilization by 70.22% while maintaining both fairness and high performance.

ABSTRACTIn modern solid‐state drives (SSDs), managing hot and cold data is key to improving performance and extending lifespan. However, previous Bloom filter‐based methods were often complex and lacked solid empirical validation under high‐performance conditions. To address this, we propose a more efficient multilevel Bloom filter classification strategy tailored to SSDs. Our approach optimizes classification accuracy while reducing computational and storage overhead through careful parameter tuning. By utilizing SSD block‐level parallelism to group similar data access patterns, we enhance garbage collection efficiency and extend block life. Unlike previous studies relying on simulations, we validate our method on real SSD hardware. Experimental results show that our strategy improves SSD performance and endurance, offering valuable insights for future firmware optimization.