The Log-Structured Merge (LSM) tree is a widely used storage engine in key-value systems, but its compaction process is a major performance bottleneck due to high computational and I/O demands. While GPU-accelerated compaction (GPU Compaction) has emerged as an efficient solution, state-of-the-art approaches rely on CPU-centric GPU-SSD direct I/O stacks, suffering from synchronization overhead between IO and GPU Computation and limited GPU-SSD IO bandwidth. To address these limitations, we propose GCLComp, a GPUCentric LSM Compaction Strategy based on GPU-Initiated Direct Storage (GIDS). It is a computational storage architecture application based on GPU and GIDS. GCLComp introduces MiniFS, an efficient user-space Host-GPU shared-driver file storage system specifically designed for GPU-accelerated LSM compaction. It enables GPU-initiated I/O for compaction and higher GPU-SSD IO throughput. This allows the GPU to independently read the input files required for compaction from the SSD, perform compaction, and write the output to SSD, minimizing CPU involvement. GCLComp further employs a GPU IO-Computation Pipeline using CUDA streams to seamlessly orchestrate I/O and computation, eliminating synchronization overhead. The experimental results show that MiniFS improves read and write throughput by 52 % and 34 %, respectively, over the state-of-the-art. Based on that, GCLComp achieves up to 19.16 % higher compaction throughput, 21.98 % higher average Fillrandom throughput, 17.59 % lower average Fillrandom latency, 97.27 % lower Fillrandom P999 tail latency and 31.85% higher ReadWriteMix throughput under the KV store, compared to SOTA. We also find that compaction throughput scales nearly linearly with pipeline batch size by a sensitivity analysis.

Approximate Nearest Neighbor Search (ANNS) is widely used in various fields, including database systems, recommendation engines, and large language models. As data dimension and size continue to expand, many studies explore GPU acceleration for graph-based ANNS. While previous methods use large batch to maximize throughput, they often lead to increased latency. In contrast, small batch is more effective for online lowlatency applications, as it minimizes batch accumulation time. However, employing small batch on GPU presents challenges. First, the query bubble issue in batch processing negatively impacts both latency and GPU utilization. Second, current GPU search methods incur excessive sorting overhead, and introduce additional TopK-merging overhead on GPU. To address these challenges, we propose ALGAS, a low-latency GPU search system designed for small batch. ALGAS employs dynamic batching based on persistent GPU kernel function to optimize query bubble. Additionally, it employs beam extend to reduce sorting overhead, especially effective at high recall rate. It also eliminates TopK-merging overhead via GPU-CPU cooperation. Furthermore, it employs an adaptive GPU tuning scheme to optimize resource utilization. We compare ALGAS with the state-of-the-art graph-based works. ALGAS reduces latency by up to 21.9%-35.4% and increases throughput by up to 27.8%-55.2% under various real-world datasets.

LSM-tree-based Key-value systems are widely used in many internet applications, known for their superior write performance. Compaction operations, responsible for maintaining the pyramidal storage structure of the LSM-tree to ensure acceptable read performance, pose significant performance bottlenecks. The application of high-performance SSDs and lightweight user-space file systems in LSM storage alleviates IO bandwidth bottlenecks, but it amplifies the computational resource consumption of compaction when KV is small and medium, shifting the bottleneck from IO to computation. To mitigate the computational bottleneck of compaction, we propose GPComp, a GPU-accelerated compaction strategy for high-performance SSDs with lightweight user-space file systems. GPComp features efficient GPU compaction units and a CPU-GPU cooperative compaction acceleration strategy. We introduce a user-space file system specifically designed for LSM storage, TopFS-GPU. It implements an SPDK-based SSD-GPU P2P IO stack to enhance data transfer throughput in GPU-accelerated Compaction. It features an asynchronous write-back cache strategy, facilitating mixed read-write workloads in LSM-tree-based key-value systems. Additionally, our pipeline mechanism overlaps GPU computations with SSD-GPU IO, increasing system throughput. Implemented based on LevelDB, GPComp shows up to a 2.65x increase in average write throughput and a 2.32x improvement in mixed read-write throughput, with a P99 tail latency reduction of up to 169.65% compared to state-of-the-art methods.

Proactive drive failure prediction can help operators handle the failing drives in advance, enhancing the storage system dependability. SSD and HDD failure prediction techniques are currently evolving towards a semi-supervised approach. In this paper, we are dedicated to enhancing the methodology for semi-supervised drive failure prediction from these aspects: design more powerful, robust yet generic models, mine drive failure modes and make the prediction model interpretable. Specifically, we propose two semi-supervised drive failure prediction models, DFP-VL and DFP-FL, from the perspective of reconstructing SMART data and learning SMART data's probability density, respectively. They capture the pattern of healthy drives, and failing drives can be detected when they deviate from the normal pattern. We mine two failure modes, slow deterioration and dramatic deterioration. Based on that, we design two failure detectors, ”ThresholdDetector” and ”HybridDetector”, to determine whether a drive deviates from the normal pattern. We evaluate the proposed methods on both SSD and HDD SMART data. DFP-VL is generic yet effective and can interpret the predicted results. DFP-FL has better performance than DFP-VL, but it cannot interpret the results. ThresholdDetector has low complexity and can detect most failing drives. HybridDetector has high complexity and can further improve the detection performance.

As a proactive failure tolerant mechanism in large scale cloud storage systems, drive failure prediction can be used to protect data by early warning before real failures of drives, and therefore improve system dependability and cloud storage service quality. At present, solid state drives (SSDs) are generally widely used in cloud storage systems due to their high performance. SSD failures seriously affect the dependability of the system and the quality of service. Existing proactive failure tolerant mechanisms for storage systems are basically aimed at HDD failure detection and use classification technology (Supervised learning), which relies on enough failure data to establish a classification model. However, the low failure rate of SSDs leads to a serious imbalance in the ratio of positive and negative samples, which brings a big challenge for establishing a proactive failure tolerance mechanism for SSDs storage systems by using classification technology.In this paper, we propose a proactive failure tolerance mechanism for SSDs storage systems based on unsupervised technology. It only uses data of normal SSDs to train the failure prediction model, which means that our method is not limited by the imbalance in SSDs data. At the core of our method is the idea to use VAE-LSTM to learn the pattern of normal SSDs, in which case faulty SSDs can be alerted when their patterns are very different from normal ones. Our method can provide early warning of failures, thereby effectively protecting data and improving the quality of cloud storage service. We also propose a drive failure cause location mechanism, which can help operators analyze the modes of failure by providing guiding suggestions. In order to evaluate the effectiveness of our method, we use cross-validation and online testing methods on SSDs data from a technology company. The results show that the FDR and FAR of our method outperform the baselines by 17.25% and 2.39% on average.