Deep Learning Training in Facebook Data Centers: Design of Scale-up and Scale-out Systems

Abstract

Large-scale training is important to ensure high performance and accuracy of machine-learning models. At Facebook we use many different models, including computer vision, video and language models. However, in this paper we focus on the deep learning recommendation models (DLRMs), which are responsible for more than 50% of the training demand in our data centers. Recommendation models present unique challenges in training because they exercise not only compute but also memory capacity as well as memory and network bandwidth. As model size and complexity increase, efficiently scaling training becomes a challenge. To address it we design Zion - Facebook's next-generation large-memory training platform that consists of both CPUs and accelerators. Also, we discuss the design requirements of future scale-out training systems.

arXiv comment10 pages, 14 figures; adjusted Fig. 10, added reference; fixed typos

References in corpus (11)

• Caffe: Convolutional Architecture for Fast Feature Embedding
• MXNet: A Flexible and Efficient Machine Learning Library for Heterogeneous Distributed Systems
• Revisiting Distributed Synchronous SGD
• Horovod: fast and easy distributed deep learning in TensorFlow
• Deep Learning Recommendation Model for Personalization and Recommendation Systems
• Rosetta: Large scale system for text detection and recognition in images
• Evaluating Modern GPU Interconnect: PCIe, NVLink, NV-SLI, NVSwitch and GPUDirect
• Distributed Deep Learning Using Synchronous Stochastic Gradient Descent
• Deep Learning Inference in Facebook Data Centers: Characterization, Performance Optimizations and Hardware Implications
• Revisiting Network Support for RDMA
• ShadowSync: Performing Synchronization in the Background for Highly Scalable Distributed Training

Cited by in corpus (12)

• ReCycle: Resilient Training of Large DNNs using Pipeline Adaptation
• Check-N-Run: A Checkpointing System for Training Deep Learning Recommendation Models
• RIBBON: Cost-Effective and QoS-Aware Deep Learning Model Inference using a Diverse Pool of Cloud Computing Instances
• TT-Rec: Tensor Train Compression for Deep Learning Recommendation Models
• Understanding Training Efficiency of Deep Learning Recommendation Models at Scale
• Time-based Sequence Model for Personalization and Recommendation Systems
• RecSSD: Near Data Processing for Solid State Drive Based Recommendation Inference
• Accelerating Recommender Systems via Hardware "scale-in"
• Quantifying and Improving Performance of Distributed Deep Learning with Cloud Storage
• Supporting Massive DLRM Inference Through Software Defined Memory
• Alternate Model Growth and Pruning for Efficient Training of Recommendation Systems
• AutoFL: Enabling Heterogeneity-Aware Energy Efficient Federated Learning