一种基于V-TGRU模型的资源调度算法

提出了一种在私有云计算环境下基于机器学习V-TGRU模型进行资源预测的算法。通过统计历史记录，将其与当前工作负载下不同任务的先验资源使用情况相结合，同时考虑工作负载特性、主机特征和同一资源池中任务之间的亲和性等因素，动态预测多任务的资源占用情况，并根据预测结果和任务运行现状进行多目标任务优化调度。实验证明，此算法能有效完成对资源的预判选择、减少调度次数、节约调度时间、节省云计算资源和带宽，保障应用任务稳定运行。

A resource scheduling algorithm based on V-TGRU model

This paper presents a new algorithm for resource prediction based on machine learning model V-TGRU in private cloud computing environment. The algorithm makes statistics of historical records and combines the prior resource usage of different tasks under the current workload, at the same time, considering the workload characteristics, host characteristics, the affinity between tasks in the same resource pool and other factors. The multi factor data matrix is further standardized and coded. The standardized coded data are modeled by V-TGRU to dynamically predict the resource occupation of multi tasks, and carry out multi-objective task optimal scheduling. Experimental results show that this method can effectively complete the pre-judgment and selection of resources, reduce the scheduling time and times, save cloud resources and bandwidth, and ensure the stable operation of application tasks.