TY - GEN
T1 - RUSM
T2 - 11th IEEE Non-Volatile Memory Systems and Applications Symposium, NVMSA 2022
AU - Alhasan, Hasan
AU - Chen, Yun Chih
AU - Ho, Chien Chung
AU - Kuo, Tei Wei
N1 - Funding Information:
ACKNOWLEDGEMENTS This work was supported in part by the Ministry of Science and Technologyunder grant no. 109-2221-E-006 -215 -MY3.
Publisher Copyright:
© 2022 IEEE.
PY - 2022
Y1 - 2022
N2 - 3D NAND flash-based storage devices, i.e., Solid-State Drives (SSDs), are gradually regarded as promising candidates to lead the flash industry thanks to their rapidly growing density. However, 3D NAND SSD has relatively high flash command latency, which raises the phenomenon of chip-blocking write, yielding the read long-tail latency problem. Data replication is a viable strategy for increasing data availability. However, data replication brings extra time overhead to read and write data, which reinforces the original chip-blocking write problem. We first reveal that the conventional scheme writes a whole page to flush smaller data portion, resulting into time squandering. Based on this observation, we propose a novel scheme, RUSM (Replicate Using Subpage Merging), which reclaims the improperly used time from the conventional page writing operation to amend the replication mechanism. Through experiments, we show how RUSM controls the chip-blocking write problem and enhances reading performance at low overhead cost.
AB - 3D NAND flash-based storage devices, i.e., Solid-State Drives (SSDs), are gradually regarded as promising candidates to lead the flash industry thanks to their rapidly growing density. However, 3D NAND SSD has relatively high flash command latency, which raises the phenomenon of chip-blocking write, yielding the read long-tail latency problem. Data replication is a viable strategy for increasing data availability. However, data replication brings extra time overhead to read and write data, which reinforces the original chip-blocking write problem. We first reveal that the conventional scheme writes a whole page to flush smaller data portion, resulting into time squandering. Based on this observation, we propose a novel scheme, RUSM (Replicate Using Subpage Merging), which reclaims the improperly used time from the conventional page writing operation to amend the replication mechanism. Through experiments, we show how RUSM controls the chip-blocking write problem and enhances reading performance at low overhead cost.
UR - http://www.scopus.com/inward/record.url?scp=85140974218&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85140974218&partnerID=8YFLogxK
U2 - 10.1109/NVMSA56066.2022.00020
DO - 10.1109/NVMSA56066.2022.00020
M3 - Conference contribution
AN - SCOPUS:85140974218
T3 - Proceedings - 2022 IEEE 11th Non-Volatile Memory Systems and Applications Symposium, NVMSA 2022
SP - 63
EP - 68
BT - Proceedings - 2022 IEEE 11th Non-Volatile Memory Systems and Applications Symposium, NVMSA 2022
PB - Institute of Electrical and Electronics Engineers Inc.
Y2 - 23 August 2022 through 25 August 2022
ER -