中文摘要：

为了有效解决RED算法拥塞响应滞后以及在聚集类之间公平性不足的问题,提出建立一种基于早期拥塞指示反馈的区分服务网络模型,并在该模型上采用优先级早期随机检测算法。在网络入口节点对分组进行分类、聚集与监测,各队列按优先级设置不同队列长度阈值,当平均队列长度或各队列长度达到相应阈值范围时,立即触发不同程度拥塞指示包向源端反馈。使用改进的NS进行了仿真实验,实验结果表明该算法使得拥塞指示能尽可能快地到达TCP源,有效地降低路由器中的丢包率。提高网络利用率,并且为不同的聚集类提供不同优先级服务,保障了聚集类的公平性。

英文摘要：

In order to solve the problem of RED algorithm in both response delay and lack of fairness between aggregations, this paper proposed the establishment of early congestion notification feedback based on DiffServ network model, and used the priority early random detection algorithm. In the entrance of the DiffServ network it classified, aggregated and monitored the arriving packets, and set each queue according to their priority to a different queue length threshold. When the average queue length or some one queue length reached the corresponding threshold range, it sent varying degrees of congestion indication packet immediately to the TCP source. It gave some experiments in the paper using modified NS simulator. The experiments show that the algorithm ensures the congestion notification can reach the TCP source as quickly as it can, so effectively reduces the packet loss rate of the router and improves network utilization. At the same time ,the algorithm can provide different priority services for different aggregations, and ensure the fairness.