集群软件无线电系统中实时信号处理调度研究

Scheduling of Real-Time Signal Processing in Cluster-Based Software Radio Systems

在集群软件无线电系统中,当宽带大容量信号数据进入系统后通过在节点上的并行计算实现对强衰弱信号的高增益、低延迟处理.结合集群软件无线电系统中信号处理的特点,研究了以下任务调度方面的问题:1)提出了一种适合集群软件无线电系统中信号处理的调度器模型.该模型简单、高效,避免了瓶颈问题.2)提出了一种新的包含3个步骤的调度策略——RQBB,其中第1步采用已有的DASAP算法.3)提出了两种启发式算法——MQB和MSD,分别用在RQBB的第2步和第3步操作.MQB是一种公平算法,用于使所有接收的任务具有较高的QoS收益(较高的QoS级别和较小的QoS级别差异),MSD算法用于使系统具有较高的吞吐率并达到负载均衡.通过大量实验对RQBB与DASAP,DALAP算法和RQRB策略进行了比较.实验结果表明,RQBB具有较高的调度成功率,使得所接收任务具有最优的QoS收益,同时使得系统具有较高的吞吐率并达到负载均衡.

In this paper, according to the characteristics of signal processing in cluster-based SR systems, the following scheduling issues are investigated： First, a universal scheduler model suitable for signal processing in cluster-based SR systems is proposed. The model is simple, the efficient and it avoids the bottleneck problem. Second, a novel three-step scheduling strategy-RQBB is put forward, where the existing DASAP algorithm is used in step 1. Third, two heuristic algorithms MQB and MSD are proposed. They are used in step 2 and step 3 of RQBB, respectively. The MQB is a fair algorithm that strives to make all the accepted tasks have high QoS benefit （high mean of QoS levels and small difference of QoS levels）. The MSD is designed to guarantee the system with high throughput and achieve load balancing without violating the timing constraints of accepted tasks. Extensive simulation experiments are performed to compare RQBB with RQRB, DASAP and DALAP. Experimental results indicate RQBB improves QoS benefit better than others and achieve load balancing while guaranteeing high schedulability.