自适应云设计与运维模式格局综述:目标、策略、工具、评估与数据集视角
An Overview on the Landscape of Self-Adaptive Cloud Design and Operation Patterns: Goals, Strategies, Tooling, Evaluation, and Dataset Perspectives
AI总结:
本文系统综述近八年111篇自适应云设计与运维模式研究,从目标、策略、工具、验证与数据集等维度梳理现状,并提出面向SaaS应用智能体的联邦学习蓝图。
摘要:
云原生应用通过使用微服务和模块化架构,显著推动了在线服务的发展与可扩展性。然而,在云环境中实现适应性、弹性和高效的性能管理仍是一项关键挑战。本工作系统回顾了过去八年中关于自适应云设计和运维模式的111篇出版物,并按目标、控制范围、决策方法、自动化水平和验证方法对其进行分类。分析表明,性能优化在研究目标中占主导地位,其次是成本降低和安全性增强,而可用性和可靠性研究不足。反应式反馈回路占主流,而通常利用机器学习的主动式方法正越来越多地应用于预测性资源供给和应用管理。面向资源的适应策略较为常见,但直接的应用级重配置仍然稀少,代表着一个有前景的研究空白。我们进一步编目了用于验证的工具、平台和30多个可公开访问的数据集,并发现数据集使用分散,缺乏事实标准。最后,我们将研究发现映射到自适应应用的通用应用级和系统级设计上,包括为SaaS应用智能体提出的联邦学习方法。该蓝图旨在引导未来工作走向更智能、具备情境感知能力的云自动化。
英文摘要:
Cloud-native applications have significantly advanced the development and scalability of online services through the use of microservices and modular architectures. However, achieving adaptability, resilience, and efficient performance management within cloud environments remains a key challenge. This work systematically reviews 111 publications from the last eight years on self-adaptive cloud design and operations patterns, classifying them by objectives, control scope, decision-making approach, automation level, and validation methods. Our analysis reveals that performance optimization dominates research goals, followed by cost reduction and security enhancement, with availability and reliability underexplored. Reactive feedback loops prevail, while proactive approaches-often leveraging machine learning-are increasingly applied to predictive resource provisioning and application management. Resource-oriented adaptation strategies are common, but direct application-level reconfiguration remains scarce, representing a promising research gap. We further catalog tools, platforms, and more than 30 publicly accessible datasets used in validation, and that dataset usage is fragmented without a de facto standard. Finally, we map the research findings on a generic application and system-level design for self-adaptive applications, including a proposal for a federated learning approach for SaaS application Agents. This blueprint aims to guide future work toward more intelligent, context-aware cloud automation.
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。