DBScholar

Back to papers

Analyzing and Mitigating Data Stalls in DNN Training

Summary: Finds input-pipeline stalls (storage fetch/preprocessing) dominate DNN training across models, datasets, and production hardware. DS-Analyzer provides differential what-if diagnosis, while CoorDL mitigates stalls and delivers up to 5× speedups over DALI. (summarized by gpt-5.6-luna on Jul 24 2026)

Paper ID
h3f0e5d33386356d4
Venue
VLDB
Year
2021
Pagerank
0.00010551698
Overall Rank
1,475 | 90.09%
DOI
10.14778/3446095.3446100
PDF
Download (CC BY-NC-ND 4.0)

Incoming Non-self Citations Over Time

Authors

BibTeX Citation

@article{mohan_vldb21,
        title = {{Analyzing and Mitigating Data Stalls in DNN Training}},
        author = {Mohan, Jayashree and Phanishayee, Amar and Raniwala, Ashish and Chidambaram, Vijay},
        journal = {PVLDB},
        series = {{VLDB} '21},
        volume = {14},
        number = {5},
        pages = {771--784},
        doi = {10.14778/3446095.3446100},
        url = {https://doi.org/10.14778/3446095.3446100},
        year = {2021}
}

Incoming Citations (Sorted by Pagerank)

Showing 16 of 16 citing papers.

Rank Citing Paper Year Venue Pagerank
2,055 tf.data: A Machine Learning Data Processing Framework 2021 VLDB 9.109765e-05
2,682 Accelerating Recommendation System Training by Leveraging Popular Choices 2022 VLDB 8.1384948e-05
3,601 Where Is My Training Bottleneck? Hidden Trade-Offs in Deep Learning Preprocessing Pipelines 2022 SIGMOD 7.1724084e-05
3,812 FastFlow: Accelerating Deep Learning Model Training with Smart Offloading of Input Data Pipeline 2023 VLDB 7.0056415e-05
5,387 GoldMiner: Elastic Scaling of Training Data Pre-Processing Pipelines for Deep Learning 2023 SIGMOD 6.1521259e-05
6,168 Apt-Serve: Adaptive Request Scheduling on Hybrid Cache for Scalable LLM Inference Serving 2025 SIGMOD 5.8603375e-05
6,314 Progressive Compressed Records: Taking a Byte out of Deep Learning Data 2021 VLDB 5.8148251e-05
7,816 Nautilus: An Optimized System for Deep Transfer Learning over Evolving Training Datasets 2022 SIGMOD 5.446446e-05
7,831 FusionFlow: Accelerating Data Preprocessing for Machine Learning with CPU-GPU Cooperation 2024 VLDB 5.4416736e-05
9,076 TensorSocket: Shared Data Loading for Deep Learning Training 2026 SIGMOD 5.2258409e-05
9,079 Scheduling Data Processing Pipelines for Incremental Training on MLP-based Recommendation Models 2025 SIGMOD 5.2258409e-05
10,005 cedar: Optimized and Unified Machine Learning Input Data Pipelines 2025 VLDB 5.0954911e-05
10,047 MEMO: Fine-grained Tensor Management For Ultra-long Context LLM Training 2025 SIGMOD 5.0896901e-05
10,670 Mixtera: A Data Plane for Foundation Model Training 2026 SIGMOD 4.9769913e-05
11,258 GPEmu: A GPU Emulator for Faster and Cheaper Prototyping and Evaluation of Deep Learning System Research 2025 VLDB 4.9769913e-05
11,442 Analyzing Near-Network Hardware Acceleration with Co-Processing on DPUs 2025 VLDB 4.9769913e-05
Previous Page 1 / 1 Next

Outgoing Citations (Sorted by Pagerank)

Showing 1 of 1 cited papers.

Citations counted here include only citations to other VLDB/SIGMOD/CIDR/PODS papers in this database.

Rank Cited Paper Year Venue Pagerank
1,152 Cerebro: A Data System for Optimized Deep Learning Model Selection 2020 VLDB 0.00011796404
Previous Page 1 / 1 Next

Semantically Similar Papers

# Overall Rank Paper Year Venue
1 13,694 Reimagining Deep Learning Systems Through the Lens of Data Systems 2024 VLDB
2 10,447 DFLOP: A Data-driven Framework for Multimodal LLM Training Pipeline Optimization 2026 SIGMOD
3 5,387 GoldMiner: Elastic Scaling of Training Data Pre-Processing Pipelines for Deep Learning 2023 SIGMOD
4 6,844 Serving Deep Learning Models with Deduplication from Relational Databases 2022 VLDB
5 4,828 Accelerating Sampling and Aggregation Operations in GNN Frameworks with GPU Initiated Direct Storage Accesses 2024 VLDB
6 4,899 Comprehensive Evaluation of GNN Training Systems: A Data Management Perspective 2024 VLDB
7 3,812 FastFlow: Accelerating Deep Learning Model Training with Smart Offloading of Input Data Pipeline 2023 VLDB
8 9,076 TensorSocket: Shared Data Loading for Deep Learning Training 2026 SIGMOD
9 3,220 Jointly Optimizing Preprocessing and Inference for DNN-based Visual Analytics 2021 VLDB
10 3,601 Where Is My Training Bottleneck? Hidden Trade-Offs in Deep Learning Preprocessing Pipelines 2022 SIGMOD