基于智能网络磁盘存储系统的任务调度算法

Task Scheduling Algorithm Based on Intelligent Network Disk Storage System

智能网络磁盘(IND)存储系统现有的调度算法虽然使任务加权总完成时间最优,但是会导致大量任务长期得不到执行,从而引起"饥饿"现象.针对这个问题,文中提出了一种可控制调度算法(CSA),把任务划分成步骤链并根据步骤链的执行时间及权重计算出ρ因子,在选择未执行步骤链时综合考虑ρ因子的大小及ρ因子所在任务的已执行时间.实验结果证明,CSA使加权总完成时间与IND现有算法相当,任务的平均周转时间及短任务的完成时间均优于IND现有算法,可以消除"饥饿"现象,较好地兼顾效益性和公平性.

Although the existing scheduling algorithm of intelligent network disk （ IND） storage system minimizes the weighted time cost of tasks,it may result in a starvation phenomenon because many tasks cannot be executed for a long time. In order to solve this problem,an controllable scheduling algorithm （ CSA） is proposed,which divides the tasks into several step chains,computes the ρ factor of tasks according to the execution time and weight of step chains,and selects the unexecuted step chain according to both the ρ factor and the execution time of the tasks of ρ factor. Experimental results indicate that,as compared with the existing IND algorithm,the proposed CAS is of similar total weighted time cost,lower average turnaround time cost of tasks and completion time cost of short tasks, and that it effectively eliminates the starvation with balanced efficiency and fairness.