TY - GEN
T1 - A joint management middleware to improve training performance of deep recommendation systems with SSDs
AU - Wu, Chun Feng
AU - Wu, Carole Jean
AU - Wei, Gu Yeon
AU - Brooks, David
N1 - Funding Information:
We would like to thank Yunho Jin, Samuel Hsia, Udit Gupta, Mark Wilkening and Jeff (Jun) Zhang for their thoughtful comments and suggestions. This work was supported in part by the Ministry of Science and Technology, Taiwan, under Grant No. 110-2917-I-564-025, the Application Driving Architectures (ADA) Research Center, a JUMP Center cosponsored by SRC and DARPA.
Publisher Copyright:
© 2022 ACM.
PY - 2022/7/10
Y1 - 2022/7/10
N2 - As the sizes and variety of training data scale over time, data preprocessing is becoming an important performance bottleneck for training deep recommendation systems. This challenge becomes more serious when training data is stored in Solid-State Drives (SSDs). Due to the access behavior gap between recommendation systems and SSDs, unused training data may be read and filtered out during preprocessing. This work advocates a joint management middleware to avoid reading unused data by bridging the access behavior gap. The evaluation results show that our middleware can effectively improve the performance of the data preprocessing phase so as to boost training performance.
AB - As the sizes and variety of training data scale over time, data preprocessing is becoming an important performance bottleneck for training deep recommendation systems. This challenge becomes more serious when training data is stored in Solid-State Drives (SSDs). Due to the access behavior gap between recommendation systems and SSDs, unused training data may be read and filtered out during preprocessing. This work advocates a joint management middleware to avoid reading unused data by bridging the access behavior gap. The evaluation results show that our middleware can effectively improve the performance of the data preprocessing phase so as to boost training performance.
KW - data arranger
KW - data preprocessing
KW - deep recommendation systems
KW - hardware/software co-design
KW - log-structured merge (LSM)
KW - solid-state drives (SSDs)
KW - training performance
UR - http://www.scopus.com/inward/record.url?scp=85137422205&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85137422205&partnerID=8YFLogxK
U2 - 10.1145/3489517.3530426
DO - 10.1145/3489517.3530426
M3 - Conference contribution
AN - SCOPUS:85137422205
T3 - Proceedings - Design Automation Conference
SP - 157
EP - 162
BT - Proceedings of the 59th ACM/IEEE Design Automation Conference, DAC 2022
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 59th ACM/IEEE Design Automation Conference, DAC 2022
Y2 - 10 July 2022 through 14 July 2022
ER -