摘要
针对工作在不可预测环境下具有不确定性任务特征的实时调度系统的服务区分问题,利用队列理论对权重队列实时调度系统进行多变量控制建模,并结合模型预测控制方法将控制问题转化为受限多变量优化问题,进而基于标准的最小二乘求解器设计预测控制调度框架以及算法。通过调整任务类的权重因子来改变任务类的计算资源分配,以实现为调度任务类提供相对延迟时间比保证。仿真试验结果表明,所设计方法在任务到达速率和执行时间变化且不能预知的情况下,仍然能为系统提供良好的相对延迟时间比控制。
To provide service differentiation guarantee for real-time schedule system with uncertain task set characteristic, quequeing theory is applied to construct the multivariable control model of a weighted queueing schedule system. Based on model predictive control approach, a predictive control scheduling framework and algorithm are designed which can keep relative delay ratio on target by dynamically adjusting computation resource allocation of scheduled tasks. The simulations results demonstrate that the predictive control schedule can achieve desired relative delay ratio specification even when task execution times and arrival rate are variable and unpredictable.
出处
《计算机工程》
CAS
CSCD
北大核心
2007年第5期7-9,12,共4页
Computer Engineering
基金
国家"863"计划基金资助项目(2003AA412010)
关键词
实时调度
延迟
预测控制
权重
服务区分
Real-time schedule
Delay
Predictive control
Weight
Service differentiation