一种多策略要素的数据访问调度算法

Multi-policy element scheduling algorithm for data access

考虑到任务的重要性、截止时间和资源分布等因素,设计了一种多策略要素的调度算法(MPES),以解决不完全独立的多源分布式气象水文数据库的访问控制问题.该算法为不同优先级的任务设定不同调度窗口,并对节点的安全级别、内容属性和负荷情况进行匹配判定,选择最佳服务节点,以优化系统公平性和整体效率.MPES算法根据队列优先级和可利用的服务资源,动态计算和调整调度窗口;优先级越高的队列,调度窗口越大,意味着可被服务的任务越多.在每个队列调度窗口时间内的任务被轮流执行.对于同一队列中的任务,根据最小松弛度优先调度策略,决定其进入调度窗口的次序,保证接近截止期的任务先执行.仿真试验结果表明,在不同的网络负荷下,MPES算法得到的分布式数据库访问任务的服务效率和公平性较MCT算法和Min-Min算法均有明显提高,尤其是高负荷情况下,总服务时间减少了11.4%～12.3%.

Considering task＇ s essentiality, deadline and resource distribution, a multi-policy element scheduling （MPES） algorithm is designed in order to solve the access control problem of incomplete independent multi-source distributed meteorological and hydrological database. In this algorithm, different scheduling windows are assigned for the tasks with different priorities, and the secure lever, data attribute and load situation are matched to select the best service node, which can optimize the system＇ s fairness and efficiency. According to the priority of every queue and available service re- sources, the size of scheduling windows is dynamically computed and adjusted. The higher the prior- ity is, the bigger the window is, which means more tasks can be got service. The tasks entering the scheduling windows can be executed alternately. For the tasks in the same queue, the MLLF （modi- fied least laxity first） policy is taken to arrange the order of entering the scheduling window, which guarantees the task close to the certain deadline can be served first. The simulation results show that the service efficiency and fairness for the tasks of distributed database access obtained by the MPES algorithm are superior to those obtained by the MCT（ minimum completion time） algorithm and the Min-Min algorithm under different network loads. Especially, the total service time decreases 11.4% to 12.3% under high loads.