arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 1612.03079cs.DCcs.LG
收藏

Clipper:一个低延迟的在线预测服务系统

Clipper: A Low-Latency Online Prediction Serving System

UC Berkeley(加州大学伯克利分校);The University of Chicago(芝加哥大学)

Daniel Crankshaw, Xin Wang, Giulio Zhou, Michael J. Franklin, Joseph E. Gonzalez, Ion Stoica

摘要:

机器学习正被部署在越来越多的应用程序中,这些应用程序需要在繁重的查询负载下进行实时,准确和鲁棒的预测。然而,大多数机器学习框架和系统只解决模型训练而不是部署。本文介绍了一个通用的低延迟预测服务系统Clipper。Clipper介于最终用户应用程序和各种机器学习框架之间,引入了一种模块化架构,以简化跨框架和应用程序的模型部署。此外,通过引入缓存、缓存和自适应模型选择技术,Clipper在不修改底层机器学习框架的情况下减少了预测延迟,提高了预测吞吐量、准确性和鲁棒性。我们在四个常见的机器学习基准数据集上评估Clipper,并证明其能够满足在线服务应用程序的延迟,准确性和吞吐量需求。最后,我们将Clipper与TensorFlow Serving系统进行了比较,并证明我们能够实现相当的吞吐量和延迟,同时支持模型组合和在线学习,以提高准确性并提供更强大的预测。

英文摘要:

Machine learning is being deployed in a growing number of applications which demand real-time, accurate, and robust predictions under heavy query load. However, most machine learning frameworks and systems only address model training and not deployment. In this paper, we introduce Clipper, a general-purpose low-latency prediction serving system. Interposing between end-user applications and a wide range of machine learning frameworks, Clipper introduces a modular architecture to simplify model deployment across frameworks and applications. Furthermore, by introducing caching, batching, and adaptive model selection techniques, Clipper reduces prediction latency and improves prediction throughput, accuracy, and robustness without modifying the underlying machine learning frameworks. We evaluate Clipper on four common machine learning benchmark datasets and demonstrate its ability to meet the latency, accuracy, and throughput demands of online serving applications. Finally, we compare Clipper to the TensorFlow Serving system and demonstrate that we are able to achieve comparable throughput and latency while enabling model composition and online learning to improve accuracy and render more robust predictions.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑