摘要
Unikernel作为云计算领域的前沿技术,具有启动速度快和资源占用少的特点。但是,在云环境大规模任务调度场景下,目前Unikernel缺少根据其任务特点所定制的调度机制,这阻碍了Unikernel进一步发挥其性能优势。针对此问题,首先总结了Unikernel的任务特点。在此基础上,提出了一种新的面向Unikernel的多级反馈队列调度机制M-RSF及其数学模型,在不影响Unikernel结构特点的前提下,能够更有效地对云环境下Unikernel的任务进行调度。同时,基于对任务特性的深入分析,提出了一种新的负载模型,该负载模型不仅可以准确刻画Unikernel的任务特点,还可以为M-RSF中调度策略的参数设置较合理的取值。最后,对M-RSF调度机制在OSvUnikernel上进行了实现和验证,实验结果表明,M-RSF可以有效减少OSvUnikernel在大规模任务调度时的平均等待时间和平均周转时间,相比于未经改动的OSv Unikernel,平均等待时间和平均周转时间减少达15%以上。
Unikernel,as a cutting-edge technology in the field of cloud computing,is characterized by its fast start-up speed and minimal resource usage.However,in the context of large-scale task scheduling in cloud environments,Uniker‐nel currently lacks a customized scheduling mechanism based on its task characteristics,which hinders the further exploi‐tation of its performance advantages.To address this issue,the distinctive task characteristics of Unikernel were initially summarized.Subsequently,a novel multi-level feedback queue scheduling mechanism for Unikernel,referred to as M-RSF,along with its mathematical model,were proposed.Without affecting the structural features of Unikernel,this mechanism could more effectively schedule Unikernel tasks in cloud environments.Meanwhile,a new load model was put forward based on an in-depth analysis of task characteristics.The model not only accurately depicted the task charac‐teristics of Unikernel but also provided a more reasonable value setting for the scheduling policy parameters in M-RSF.Lastly,the implementation and validation of the M-RSF scheduling mechanism on the OSv Unikernel were conducted.Experimental results indicate that M-RSF can effectively reduce the average waiting time and average turnaround time of the OSv Unikernel during large-scale task scheduling,achieving a reduction of more than 15%compared to the unmodi‐fied OSv Unikernel.
作者
董博南
杨秋松
李明树
DONG Bonan;YANG Qiusong;LI Mingshu(University of Chinese Academy of Sciences,Beijing 100049,China;National Engineering Research Center for Fundamental Software,Institute of Software Chinese Academy of Sciences,Beijing 100190,China)
出处
《通信学报》
EI
CSCD
北大核心
2024年第5期54-69,共16页
Journal on Communications
基金
中国科学院战略性先导科技专项基金资助项目(No.XDA-Y01-01,No.XDC02010600)。