总览图：
工具：
默认调度器
https://blog.csdn.net/QI8811481/article/details/157386876?spm=1001.2014.3001.5502
调度器的工作要求 ：
调度器的调度策略
https://blog.csdn.net/QI8811481/article/details/157391094?spm=1001.2014.3001.5502
资源充足情况下：
Container绑定CPU核
https://blog.csdn.net/QI8811481/article/details/157387126?spm=1001.2014.3001.5502
Container绑定GPU (通过Device Plugin实现硬件资源绑定)
https://blog.csdn.net/QI8811481/article/details/157467148?spm=1001.2014.3001.5501
资源不足情况下：
集群内部对象都已经被划分到资源模型中
https://blog.csdn.net/QI8811481/article/details/157385799?spm=1001.2014.3001.5502
依据资源模型实行资源回收-Eviction
https://blog.csdn.net/QI8811481/article/details/157386621?spm=1001.2014.3001.5502
新建的K8S对象根据优先级实现抢占
https://blog.csdn.net/QI8811481/article/details/157392812?spm=1001.2014.3001.5502
资源从不足到充足的情况下：
MoveAllToActiveQueue插件刷新调度机会
https://blog.csdn.net/QI8811481/article/details/157393079?spm=1001.2014.3001.5502
Pod更新触发关联Pod迁移
https://blog.csdn.net/QI8811481/article/details/157393268?spm=1001.2014.3001.5501