Scholarlink

<p>In cloud computing environments, the dynamic and unpredictable nature of user demands and workloads poses significant challenges to resource scheduling and management. To address these challenges, we propose MDRP (multidimensional data stream based Resource Prediction and intelligent scheduling), a novel framework that integrates real-time multidimensional data stream analysis with intelligent scheduling and proactive resource prediction. MDRP captures diverse system-level and application-level metrics–including CPU, memory, I/O, network usage, task characteristics, and historical behavior–to build a holistic representation of system states. A deep reinforcement learning(RL) module is designed to dynamically adjust scheduling decisions based on predicted workload patterns and resource utilization trends. Moreover, the framework incorporates time series forecasting to anticipate future resource demands, enabling proactive scaling and migration strategies. Experimental results on simulated cloud environments demonstrate that MDRP improves task completion efficiency by 8.1%, reduces resource contention by 12.3%, and enhances system stability by 9.4% compared to existing methods.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

An intelligent scheduling and resource prediction framework using multidimensional data streams and deep reinforcement learning

  • 期刊论文

  • 发表日期：

    2026年6月1日

查看全文
  • Wei Deng,
  • Ziliang Qiu,
  • Xuezhi Liu,
  • Juzheng Huang,
  • Binhua Ren,
  • Siyuan Qin

摘要

In cloud computing environments, the dynamic and unpredictable nature of user demands and workloads poses significant challenges to resource scheduling and management. To address these challenges, we propose MDRP (multidimensional data stream based Resource Prediction and intelligent scheduling), a novel framework that integrates real-time multidimensional data stream analysis with intelligent scheduling and proactive resource prediction. MDRP captures diverse system-level and application-level metrics–including CPU, memory, I/O, network usage, task characteristics, and historical behavior–to build a holistic representation of system states. A deep reinforcement learning(RL) module is designed to dynamically adjust scheduling decisions based on predicted workload patterns and resource utilization trends. Moreover, the framework incorporates time series forecasting to anticipate future resource demands, enabling proactive scaling and migration strategies. Experimental results on simulated cloud environments demonstrate that MDRP improves task completion efficiency by 8.1%, reduces resource contention by 12.3%, and enhances system stability by 9.4% compared to existing methods.