全负载场景中最优调度算法长时平均性能分析

Analysis of long-term average performance of optimal scheduling scheme in full-load scenario

首先分析了全负载场景中轮询调度、最大速率调度、比例公平调度和速率累积分布调度这4种常用调度算法.结果显示,速率累积分布调度在保证公平的基础上可以得到最好的效率,是4种调度算法中的最优算法.然后采用概率推导法给出了该调度算法的长时平均性能分析,即以轮询调度为比较基准的多用户分集增益的理论推导.该分集增益可通过短时统计结果预测长时平均性能,且可适用于任意实际场景.计算机仿真结果验证了对于该调度算法所产生的多用户分集增益理论分析的准确性,理论分析结果与实际调度结果的误差低于0.1%.

Four well-known scheduling schemes in full-load scenario, namely, round robin schedu- ling, max-rate scheduling, proportional fairness scheduling and cumulative rate distribution based scheduling are investigated. Results show that the last one achieves the best efficiency with a relativebetter fairness guarantee, which is the optimal scheme among all the four ones. Then, the probabili- ty deduction method is adopted to analyze its multi-user diversity gain, which uses round robin scheduling as the benchmark. The result can be used to predict the long-term average performance ofthe scheme through short-time statistical results and is applicable to all practical scenarios. The accu- racy of the analysis results are verified by numerical simulations, in which the error between theory analysis and actual scheduling is less than O. 1%.