[2022-Oct-26] A Joint Management Middleware to Improve Training Performance of Deep Recommendation Systems with SSDs
Institute of Information Systems and Applications |
|
Speaker: |
Chun-Feng Wu (吳俊峯), Assistant Professor |
Topic: |
A Joint Management Middleware to Improve Training Performance of Deep Recommendation Systems with SSDs |
Date: |
13:20-15:00 Wednesday 26-Oct-2022 |
QR Code: |
|
Link: |
|
Location: |
Delta 105 |
Hosted by: |
Prof. Te-Chuan Chiu |
Abstract
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. Last but not least, I will share some of my representative works and also future plans in the field of memory and storage systems (including OS rethinking and processing-in-memory).
Bio.
Chun-Feng Wu received his Ph.D. degree in Department of Computer Science and Information Engineering from National Taiwan University, Taipei, Taiwan, in 2021. He is now an assistant professor at Department of Computer Science, National Yang Ming Chiao Tung University, Hsinchu, Taiwan. He was a Postdoctoral Scholar at Harvard University before joining NYCU. He served in R&D alternative service at Institute of Information Science, Academia Sinica, Taipei, Taiwan. His primary research interests include memory/storage systems, operating systems, processing-in-memory, and next-generation memory/storage architecture designs. (https://cfwu417.github.io/)
All faculty and students are welcome to join.