# Accelerating Recommender Model ETL with a Streaming FPGA-GPU Dataflow

arXiv: https://arxiv.org/abs/2501.12032 · HTML: https://mlsanity.com/paper/2501.12032/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 23 of its 80 references that are in the map.
- First posted on arXiv 21 Jan 2025.

## Builds on (most cited first, 23 of 23)

- [Wide & Deep Learning for Recommender Systems](https://mlsanity.com/paper/1606.07792/) (2016)
- [AutoML: A Survey of the State-of-the-Art](https://mlsanity.com/paper/1908.00709/) (2019)
- [Deep Learning Recommendation Model for Personalization and Recommendation Systems](https://mlsanity.com/paper/1906.00091/) (2019)
- [A Study of BFLOAT16 for Deep Learning Training](https://mlsanity.com/paper/1905.12322/) (2019)
- [The Architectural Implications of Facebook's DNN-based Personalized Recommendation](https://mlsanity.com/paper/1906.03109/) (2019)
- [NVIDIA Tensor Core Programmability, Performance & Precision](https://mlsanity.com/paper/1803.04014/) (2018)
- [Software-Hardware Co-design for Fast and Scalable Training of Deep Learning Recommendation Models](https://mlsanity.com/paper/2104.05158/) (2021)
- [tf.data: A Machine Learning Data Processing Framework](https://mlsanity.com/paper/2101.12127/) (2021)
- [Contrastive Learning for Debiased Candidate Generation in Large-Scale Recommender Systems](https://mlsanity.com/paper/2005.12964/) (2020)
- [Monolith: Real Time Recommendation System With Collisionless Embedding Table](https://mlsanity.com/paper/2209.07663/) (2022)
- [Understanding Data Storage and Ingestion for Large-Scale Deep Recommendation Model Training](https://mlsanity.com/paper/2108.09373/) (2021)
- [Embedding in Recommender Systems: A Survey](https://mlsanity.com/paper/2310.18608/) (2023)
- [Plumber: Diagnosing and Removing Performance Bottlenecks in Machine Learning Data Pipelines](https://mlsanity.com/paper/2111.04131/) (2021)
- [tf.data service: A Case for Disaggregating ML Input Data Processing](https://mlsanity.com/paper/2210.14826/) (2022)
- [Online Learning for Recommendations at Grubhub](https://mlsanity.com/paper/2107.07106/) (2021)
- [Evaluating Online Bandit Exploration In Large-Scale Recommender System](https://mlsanity.com/paper/2304.02572/) (2023)
- [Profiling and Improving the PyTorch Dataloader for high-latency Storage: A Technical Report](https://mlsanity.com/paper/2211.04908/) (2022)
- [Auto-FP: An Experimental Study of Automated Feature Preprocessing for Tabular Data](https://mlsanity.com/paper/2310.02540/) (2023)
- [Modyn: Data-Centric Machine Learning Pipeline Orchestration](https://mlsanity.com/paper/2312.06254/) (2023)
- [Importance of Data Loading Pipeline in Training Deep Neural Networks](https://mlsanity.com/paper/2005.02130/) (2020)
- [ACCL+: an FPGA-Based Collective Engine for Distributed Applications](https://mlsanity.com/paper/2312.11742/) (2023)
- [Characterizing Off-path SmartNIC for Accelerating Distributed Systems](https://mlsanity.com/paper/2212.07868/) (2022)
- [OSMOSIS: Enabling Multi-Tenancy in Datacenter SmartNICs](https://mlsanity.com/paper/2309.03628/) (2023)

## Abstract (arXiv)

The real-time performance of recommender models depends on the continuous integration of massive volumes of new user interaction data into training pipelines. While GPUs have scaled model training throughput, the data preprocessing stage - commonly expressed as Extract-Transform-Load (ETL) pipelines - has emerged as the dominant bottleneck. Production systems often dedicate clusters of CPU servers to support a single GPU node, leading to high operational cost. To address this issue, we present PipeRec, a hardware-accelerated ETL engine co-designed with online recommender model training. PipeRec introduces a training-aware ETL abstraction that exposes freshness, ordering, and batching semantics while compiling software-defined operators into reconfigurable FPGA dataflows and overlaps ETL with GPU training to maximize utilization under I/O constraints. To eliminate CPU bottlenecks, PipeRec implements a format-aware packer that streams training-ready batches directly into GPU memory via P2P DMA transfers, enabling zero-copy ingest and efficient GPU consumption. Our evaluation on three datasets shows that PipeRec accelerates ETL throughput by over 10x compared to CPU-based pipelines and up to 17x over state-of-the-art GPU ETL systems. When integrated with training, PipeRec maintains 64-91% GPU utilization and reduces end-to-end training time to 9.94% of the time taken by CPU-GPU pipelines.