arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2407.20494cs.NI
收藏

多云环境中的MLOPS:典型网络拓扑

MLOPS in a multicloud environment: Typical Network Topology

Boyang Yan

AI总结:

本文探讨了支持多云环境的云原生MLOPS流水线的设计与实现,旨在通过优化网络拓扑和云提供商选择,构建确保高可用性、可扩展性和安全性的基础设施,以简化机器学习模型的部署与生命周期管理。

摘要:

随着人工智能、机器学习和数据科学持续驱动以数据为中心的经济,由于海量数据和计算需求导致在单机上实现机器学习的挑战,促使人们采用云计算解决方案。本文探讨了支持多云环境的、安全的云原生机器学习操作(MLOPS)流水线的设计与实现。主要目标是创建一个稳健的基础设施,以促进安全的数据收集、实时模型推理和机器学习生命周期的有效管理。通过利用云提供商的能力,该解决方案旨在简化机器学习模型的部署和维护,确保高可用性、可扩展性和安全性。本文详细介绍了实现最佳MLOPS环境的网络拓扑、问题描述、业务和技术需求、权衡以及提供商选择过程。

英文摘要:

As artificial intelligence, machine learning, and data science continue to drive the data-centric economy, the challenges of implementing machine learning on a single machine due to extensive data and computational needs have led to the adoption of cloud computing solutions. This research paper explores the design and implementation of a secure, cloud-native machine learning operations (MLOPS) pipeline that supports multi-cloud environments. The primary objective is to create a robust infrastructure that facilitates secure data collection, real-time model inference, and efficient management of the machine learning lifecycle. By leveraging cloud providers' capabilities, the solution aims to streamline the deployment and maintenance of machine learning models, ensuring high availability, scalability, and security. This paper details the network topology, problem description, business and technical requirements, trade-offs, and the provider selection process for achieving an optimal MLOPS environment.

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。

↑