Supporting Massive DLRM Inference Through Software Defined Memory
arXiv:2110.11489
Abstract
Deep Learning Recommendation Models (DLRM) are widespread, account for a considerable data center footprint, and grow by more than 1.5x per year. With model size soon to be in terabytes range, leveraging Storage ClassMemory (SCM) for inference enables lower power consumption and cost. This paper evaluates the major challenges in extending the memory hierarchy to SCM for DLRM, and presents different techniques to improve performance through a Software Defined Memory. We show how underlying technologies such as Nand Flash and 3DXP differentiate, and relate to real world scenarios, enabling from 5% to 29% power savings.
14 pages, 5 figures
References in corpus (6)
- Deep Learning Recommendation Model for Personalization and Recommendation Systems
- Wide & Deep Learning for Recommender Systems
- FBGEMM: Enabling High-Performance Low-Precision Deep Learning Inference
- First-Generation Inference Accelerator Deployment at Facebook
- Post-Training 4-bit Quantization on Embedding Tables
- JIZHI: A Fast and Cost-Effective Model-As-A-Service System for Web-Scale Online Inference at Baidu