Qingheng Gao
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Qingheng Gao has a total of 9 co-authors. Here are the top co-authors by number of shared publications:
- Hang Cheng
ByteDance China, Beijing, China
1 shared publication - Jixi Shan
ByteDance China, Beijing, China
1 shared publication - Jun Song
ByteDance China, Beijing, China
1 shared publication - Rui Shi
ByteDance Inc., Beijing, China
1 shared publication - Teng Yin
Zhejiang University, Hangzhou, China
1 shared publication - Wei Chen
Zhejiang University, Hangzhou, Zhejiang, China
1 shared publication
Latest Publications
- Open Access
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 8399-8409
FLEDA: Forecasting-Based Data Partition Placement for Efficient Deep Learning Recommendation Model Training
In recommendation and advertising systems, model training pipelines rely on high-throughput ingestion of fresh, collected behavioral data. These data are typically organized into time-partitioned, multi-level table paths. The access patterns driven by highly concurrent model training tasks exhibit complex cross-path correlations, multi-scale temporal dependencies, and frequent hot-cold transitions, making effective data placement under limited SSD capacity challenging. In this paper, we propose FLEDA, a forecasting-based data partition placement framework that selectively places daily-level training data partitions onto SSDs under constrained storage capacity. We first model historical data access patterns as multivariate and multiscale time series to capture both temporal dynamics and cross-path dependencies. By leveraging dependencies across two dimensions, we propose a model that extracts multi-scale cross-path temporal features and fuses mixture-of-experts to forecast the future throughput of all table paths. Besides, we design a ranking-based placement policy that balances predicted throughput, recent access history, and SSD residency continuity to maximize coverage of future data accesses while avoiding excessive data migration. Experiments on large-scale, real-world training workloads show that FLEDA consistently outperforms state-of-the-art multivariate time-series forecasting models and significantly improves SSDs' throughput coverage with low data migration overhead.
- 8 August 2026
- Published by Association for Computing Machinery (ACM)
- p. 8399-8409
Open AccessFLEDA: Forecasting-Based Data Partition Placement for Efficient Deep Learning Recommendation Model Training
In recommendation and advertising systems, model training pipelines rely on high-throughput ingestion of fresh, collected behavioral data. These data are typically organized into time-partitioned, multi-level table paths. The access patterns driven by highly concurrent model training tasks exhibit complex cross-path correlations, multi-scale temporal dependencies, and frequent hot-cold transitions, making effective data placement under limited SSD capacity challenging. In this paper, we propose FLEDA, a forecasting-based data partition placement framework that selectively places daily-level training data partitions onto SSDs under constrained storage capacity. We first model historical data access patterns as multivariate and multiscale time series to capture both temporal dynamics and cross-path dependencies. By leveraging dependencies across two dimensions, we propose a model that extracts multi-scale cross-path temporal features and fuses mixture-of-experts to forecast the future throughput of all table paths. Besides, we design a ranking-based placement policy that balances predicted throughput, recent access history, and SSD residency continuity to maximize coverage of future data accesses while avoiding excessive data migration. Experiments on large-scale, real-world training workloads show that FLEDA consistently outperforms state-of-the-art multivariate time-series forecasting models and significantly improves SSDs' throughput coverage with low data migration overhead.