A5081041448- Advanced Data Storage Technologies
- Caching and Content Delivery
- Parallel Computing and Optimization Techniques
- Cloud Computing and Resource Management
- Distributed and Parallel Computing Systems
- Distributed systems and fault tolerance
- Interconnection Networks and Systems
Wuhan National Laboratory for Optoelectronics
2019-2022
Huazhong University of Science and Technology
2019-2022
LSM-tree is widely used in key-value stores for big data storage, but it suffers from write amplification brought by frequent compaction operations. An effective solution this problem separation, which decouples values the and them a separate value log. However, existing separation schemes achieve poor range query performance, especially small pairs, because they focus on mitigating neglect access characteristics of SSD. In article, we propose FenceKV, aims to better performance while...
PCIe devices, such as SSDs and GPUs, are pivotal in modern data centers, their value is set to grow amidst the emergence of AI large models. However, these devices face onboard DRAM shortage issue due internal space limitation, preventing accommodation sufficient modules alongside flash or GPU processing chips. Current solutions either curb device-internal memory usage supplement slower non-DRAM mediums, prove inadequate performance-compromising. This paper introduces Linked Memory Buffer...
Shingled Magnetic Recording (SMR) is one of the most promising techniques that satisfy ever-growing storage volume demands. By overlapping tracks, SMR enormously improves area density, which in turn brings higher volumes. However, sacrifices random write performance for better Current drives propose to remedy this problem by employing an in-drive persistent cache temporally store incoming writes and migrate them their disk destinations later on. Unfortunately, cleaning processes takes up...
Primary and secondary indices in LSM-tree-based key-value (KV) stores play significant roles for real-world applications, but they suffer severe I/O amplification due to compaction operations. Prior works show that KV separation can mitigate the under various workloads either primary or indices. However, range queries of only achieve suboptimal efficiency two reasons: (1) improves insert/update performance by sacrificing queries, (2) may conflict with each other.
In a parity-based SSD RAID, small write requests not only accelerate the wear-out of SSDs due to extra writes for updating parities but also deteriorate performance associated expensive garbage collection. To mitigate problem writes, buffer is often added at RAID controller absorb overwrites and performed same stripe. However, this approach achieves suboptimal efficiency because file layout information invisible block level.This paper proposes RAFS, RAID-aware system, which utilizes...
Log-Structured-Merge (LSM) tree-based key value stores are facing critical challenges of fully leveraging the dramatic performance improvements underlying storage devices, which makes compaction operations LSM become CPU-bound, and slow compactions significantly degrade store performance. To address this issue, we propose LUDA, an with CUDA, uses a GPU to accelerate stores. How efficiently parallelize procedures as well accommodate optimal contract architecture challenge LUDA. Specifically,...
In big data era, LevelDB is widely deployed in enterprise server for massive storage. However, with the growing size of stored data, will inevidablely frequently call background compaction thread to compact between level, resulting write amplifications and therefore degrading performance. this paper, we propose WOKV, a Write-Optimized Key-Value store, mitigate amplification improve system WOKV adopts Least-Rewrite Compaction Strategy induced by operations, uses Multi-Thread Multi-Buffer...
Primary and secondary indices are demanded in real-world applications. Recent works show that key-value(KV) separation is efficient to improve queries of multiple LSM-based KV stores. It stores the value a separate log only keys address primary indices. However, this share-value scheme on results suboptimal efficiency: (1) range all cannot fully exploit bandwidth SSD devices simultaneously (2) put operation inefficient update indices.To above inefficiency, we propose RepKV, replicated store...