基于无线信道状态信息的缓存膨胀问题求解方案
Bufferbloat mitigating scheme based on radio channel state information

缓存膨胀问题是指超出链路容量的分组阻塞在缓存中造成网络传输延时增加的现象，蜂窝网络中为提高数据传输的可靠性而在基站上配置的大缓存会进一步加剧缓存膨胀问题。针对这一现状，该文提出了一种基于无线信道状态信息的接收窗口调节方案。该方案直接在接收端利用无线信道的状态信息估算出网络的可用带宽；为削弱信道噪声和传输控制协议的确认分组自时钟机制的影响，采用Kalman滤波算法对可用带宽做进一步的预测；然后利用该预测值动态地计算接收窗口并调节发送窗口的大小。基于网络仿真平台的实验结果表明：相比已有的解决方案，该方案在保持吞吐量不变的情况下，能有效地减小基站大缓存中的队列长度，减少网络延时。
Abstract：Bufferbloat occurs when excessive packets that exceed the network capacity are blocked in the buffer, which increases the network transmission latency. Cellular networks generally use a large buffer for network reliability, which further aggravates this problem. This paper gives a receive window adjustment scheme based on radio channel state information to mitigate bufferbloat in cellular networks. The scheme retrieves the available bandwidth at the receiver side from the radio channel state information. The effects of channel noise and the ACK self-clocking mechanism in the transmission control protocol are modeled by a Kalman filter to predict the available bandwidth. Then the send window size is adjusted according to the calculated receive window. The scheme performance was evaluated on a network simulation platform with the results indicating that this scheme significantly reduces the queue backlog and the latency while still maintaining the original high throughput than previous mechanisms