The workload prediction and resource allocation significantly play an inevitable role in production of an efficient cloud environment. The proactive estimation of future workload followed by decision of resource allocation have become a prior solution to handle other in-built challenges like the under/over-loading of physical machines, resource wastage, Quality-of-Services (QoS) violations, load balancing,VM migration and many more. In this context, the paper presents a comprehensive survey of workload forecasting and predictive resource management models in cloud environment. A conceptual framework for workload forecasting and resource management, categorization of existing machine learning based resources allocation techniques, and major challenges of inefficient distribution of physical resource distribution are discussed pertaining to cloud computing. Thereafter, a thorough survey of existing state-of-the-art contributions empowering machine learning based approaches in the field of cloud workload prediction and resource management are rendered. Finally, the paper explores and concludes various emerging challenges and future research directions concerning elastic resource management in cloud environment.
翻译:工作量预测和资源分配在创造高效云层环境方面发挥着不可避免的重要作用。积极主动地估计未来工作量,然后作出资源分配决定,这已成为处理其他内在挑战的先决解决办法,如物理机载不足/超载、资源浪费、违反服务质量、负负载平衡、VM迁移等。在这方面,本文件全面调查云层环境中的工作量预测和预测资源管理模型。讨论了工作量预测和资源管理的概念框架、基于机器的现有资源分配方法分类以及实际资源分配效率低下的重大挑战。随后,对云量预测和资源管理领域基于增强机能学习的现有最新方法进行了彻底调查。最后,本文件探讨并总结了关于云层环境中弹性资源管理的各种新出现的挑战和未来研究方向。