PetPS: Supporting Huge Embedding Models with Persistent Memory

Author:

Xie Minhui1,Lu Youyou1,Wang Qing1,Feng Yangyang1,Liu Jiaqiang2,Ren Kai2,Shu Jiwu1

Affiliation:

1. Tsinghua University

2. Kuaishou

Abstract

Embedding models are effective for learning high-dimensional sparse data. Traditionally, they are deployed in DRAM parameter servers (PS) for online inference access. However, the ever-increasing model capacity makes this practice suffer from both high storage costs and long recovery time. Rapidly developing Persistent Memory (PM) offers new opportunities to PSs owing to its large capacity at low costs, as well as its persistence, while the application of PM also faces two challenges including high read latency and heavy CPU burden. To provide a low-cost but still high-performance parameter service for online inferences, we introduce PetPS, the first production-deployed PM parameter server. (1) To escape with high PM latency, PetPS introduces a PM hash index tailored for embedding model workloads, to minimize PM access. (2) To alleviate the CPU burden, PetPS offloads parameter gathering to NICs, to avoid CPU stalls when accessing parameters on PM and thus improve CPU efficiency. Our evaluation shows that PetPS can boost throughput by 1.3 -- 1.7X compared to PSs that use state-of-the-art PM hash indexes, or get 2.9 -- 5.5X latency reduction with the same throughput. Since 2020, PetPS has been deployed in Kuaishou, one world-leading short video company, and successfully reduced TCO by 30% without performance degradation.

Publisher

Association for Computing Machinery (ACM)

Subject

General Earth and Planetary Sciences,Water Science and Technology,Geography, Planning and Development

Reference56 articles.

1. 2022. Apache Kafka. https://kafka.apache.org/. 2022. Apache Kafka. https://kafka.apache.org/.

2. 2022. Apache MXNet | A flexible and efficient library for deep learning. https://mxnet.apache.org/versions/1.9.1/. 2022. Apache MXNet | A flexible and efficient library for deep learning. https://mxnet.apache.org/versions/1.9.1/.

3. 2022. Compute Express Link. https://www.computeexpresslink.org/. 2022. Compute Express Link. https://www.computeexpresslink.org/.

4. 2022. dmlc/ps-lite: A lightweight parameter server interface. https://github.com/dmlc/ps-lite. 2022. dmlc/ps-lite: A lightweight parameter server interface. https://github.com/dmlc/ps-lite.

5. 2022. DPDK. https://www.dpdk.org/. 2022. DPDK. https://www.dpdk.org/.

Cited by 4 articles. 订阅此论文施引文献 订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献

1. Sorting on Byte-Addressable Storage: The Resurgence of Tree Structure;Proceedings of the VLDB Endowment;2024-02

2. Storage Technology Trends and Development;Data Storage Architectures and Technologies;2024

3. Perseid: A Secondary Indexing Mechanism for LSM-based Storage Systems;ACM Transactions on Storage;2023-11-17

4. Progress on storage systems for disaggregated data centers;SCIENTIA SINICA Informationis;2023-08-01

同舟云学术

1.学者识别学者识别

2.学术分析学术分析

3.人才评估人才评估

"同舟云学术"是以全球学者为主线,采集、加工和组织学术论文而形成的新型学术文献查询和分析系统,可以对全球学者进行文献检索和人才价值评估。用户可以通过关注某些学科领域的顶尖人物而持续追踪该领域的学科进展和研究前沿。经过近期的数据扩容,当前同舟云学术共收录了国内外主流学术期刊6万余种,收集的期刊论文及会议论文总量共计约1.5亿篇,并以每天添加12000余篇中外论文的速度递增。我们也可以为用户提供个性化、定制化的学者数据。欢迎来电咨询!咨询电话:010-8811{复制后删除}0370

www.globalauthorid.com

TOP

Copyright © 2019-2024 北京同舟云网络信息技术有限公司
京公网安备11010802033243号  京ICP备18003416号-3