一种分段平滑的随机早期检测队列管理算法  

An Improved Random Early Detection Queue Management Algorithm

作　　者：赵宇红[1] 白雪冰[1] 张晓琳[1] 

机构地区：[1]内蒙古科技大学信息工程学院,内蒙古包头014010

出　　处：《智能计算机与应用》2014年第3期15-18,共4页Intelligent Computer and Applications

基　　金：内蒙古自然科学基金项目(2011MS0916)

摘　　要：随机早期检测算法(RED)的性能受其参数设置的影响较大,并且该算法中的可设置的参数较多。同时,以早期随机检测算法(RED)计算得到的丢包概率的变化过于激进,在网络负载变化较快时,平均队列长度抖动幅度较大,算法性能不够稳定。为了克服以上问题,这里提出一种新的改进思路——随机早期平滑分段算法(RED-P)。该算法采用二次函数分段计算丢包概率,使得丢包概率变化更加平滑,同时对概率计算公式进行了简化,减少了计算所需的参数量,适度的避免了参数设置对算法性能的影响。经过网络模拟平台NS2的网络模拟仿真实验的对比,结果表明新算法在端到端的延时方面和平均队列长度抖动幅度方面都有所改善,且可获得与原算法几乎接近的吞吐量,提高了服务质量。The performance of the RED algorithm is greatly affected by its parameter settings and there is more parameters need to be set , at the same time , the change of dropping packets probability computing by RED algorithm is too radical , the faster the change of network load is , the more the jitter amplitude of average queue length is , which means the algo- rithm performance is unstable. In order to overcome the problems above , this paper presents a new improvement ideas - Red - P , This algorithm calculates the probability of dropping packets piecewisely by using the quadratic function , making the change of probability of dropping packets more gentle , meanwhile , simplifys the probability calculation formula , de- creases the number of parameters the calculation needs , properly avoids the influence of parameter settings of the algorithm performance. Through the contrast of NS2 simulation experiment , the result shows that the new algorithm is improved on end - to - end delay and jitter amplitude of average queue length , in addition , this algorithm obtains the throughput which is close to the original algorithm , improves the quality of service .

关 键 词：RED 队列管理 拥塞控制 抖动 RED-P 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]