Maximizing I/O Throughput and Minimizing Performance Variation via Reinforcement Learning Based I/O Merging for SSDs

Wu, Chao; Ji, Cheng; Li, Qiao; Gao, Congming; Pan, Riwei; Fu, Chenchen; Shi, Liang

doi:10.1109/tc.2019.2938956

Cited by 14 publications

(8 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…[14][15][16][17] Gugnani et al 18 propose a set of strategies for providing QoS performance guarantee with non-volatile memory express (NVMe) SSDs on virtual environment. Wu et al 2 propose a reinforcement learning-assisted I/O merging technique for improved QoS performance on SSDs. However, few considers the QoS violation incurred by OSP in 3D CT-based SSDs.…”

Section: Related Workmentioning

confidence: 99%

“…The blocked I/O requests may experience a long I/O latency, which incurs enlarged I/O performance variation 2 . As shown in Figure 1B, the standard deviation of I/O latency for a SSD with OSP is 55.3% larger than a device without OSP on average for a new device, while the value is 58.9% for a device aged by 70%.…”

Section: Motivationmentioning

confidence: 99%

“…QoS demands all I/O latency to be constrained within a specific range 2 . Although SSDs provide a faster accessing speed comparing with HDDs, they suffer from large I/O performance variation and long worst‐case latency owing to various reasons, for example, garbage collection, read/write interference, process variation 2,3 . The enlarged I/O performance variation and prolonged worst‐case latency could violate the QoS demand and impact the user experience.…”

Section: Introductionmentioning

confidence: 99%

“…While OSP technique is recommended for superior system I/O throughput, this study identifies that OSP could significantly degrade the quality‐of‐service (QoS) performance. QoS demands all I/O latency to be constrained within a specific range 2 . Although SSDs provide a faster accessing speed comparing with HDDs, they suffer from large I/O performance variation and long worst‐case latency owing to various reasons, for example, garbage collection, read/write interference, process variation 2,3 .…”

Section: Introductionmentioning

confidence: 99%

“…2 Although SSDs provide a faster accessing speed comparing with HDDs, they suffer from large I/O performance variation and long worst-case latency owing to various reasons, for example, garbage collection, read/ write interference, process variation. 2,3 The enlarged I/O performance variation and prolonged worst-case latency could violate the QoS demand and impact the user experience. As all programmed pages share one completion time, OSP could spread one prolonged worst-case latency to the latency of all programmed pages.…”

mentioning

confidence: 99%

See 4 more Smart Citations

Machine learning assisted OSP approach for improved QoS performance on 3D charge‐trap based SSDs

Zhu

et al. 2020

Int J Intell Syst

Self Cite

View full text Add to dashboard Cite

Three-dimensional (3D) charge-trap based solid-statedrivers (SSDs) have become an emerging storage solution in recent years. One-shot-programming in 3D charge-trap based SSDs could deliver a maximized system input/output (I/O) throughput at the cost of degraded Quality-of-Service (QoS) performance. This paper proposes reinforcement-learning based one-shotprogramming (RLOSP), a reinforcement learning based approach to improve the QoS performance for 3D charge-trap based SSDs. By learning the I/O patterns of the workload environments as well as the device internal status, the proposed approach could properly choose requests in the device queue, and allocate physical addresses for these requests during one-shotprogramming. In this manner, the storage device could deliver an improved QoS performance. Experimental results reveal that the proposed approach could reduce the worst-case latency at the 99.9th percentile by 37.5%-59.2%, with an optimal system I/O throughput.

show abstract