网络带宽测量

 网络带宽测量技术，可以为端的速率控制提供依据。这里主要介绍几种有影响力的带宽测量方法，对其中的原理，则尝试推之。

基于卡尔曼的带宽测量技术

 WebRTC中的拥塞控制算法GCC就是通过卡尔曼滤波技术，推导出网络中排队时延差(queue delay variation)。GCC将网络排队时延差作为网络拥塞信号，进而控制发送端的编码速率。GCC中的卡尔曼滤波方法中的时延帧的长度为单位的。[1]的带宽测量技术以数据包为单位进行的。

符号	含义
$C_i$	链路容量
$X_i$	链路背景流量
$B_i$	用户可用带宽

$$B_i=C_i-X_i\text{\hspace{1em}(2)}$$
 There are some basic assumptions in[1].The link queue is FIFO,no priority queue in intermediate routers and the traffic is scheduled on a FCFS basis.The work in[1] bases the analysis on fluid traffic model. If the sending rate $u<C-X$,there is no congestion.The received traffic rate will exactly be $u$. If $u>C-X$,the received probe traffic will be:
$$r=\frac{u}{u+X}C\text{\hspace{1em}(4)}$$
 [8]尝试在用户侧用packet train技术测量链路容量。
[1]Real-Time Measurement of End-to-End Available Bandwidth using Kalman Filtering(2006)
[2]Pathload: A measurement tool for end-to-end available bandwidth(2002)
[3]pathchirp: Efficient available bandwidth estimation for network paths(2003)
[4]A new end-to-end probing and analysis method for estimating bandwidth bottlenecks(2000)
[5]End-to-End Available Bandwidth: Measurement Methodology, Dynamics, and Relation With TCP Throughput(2003-ton)
[6]An Enhanced Available Bandwidth Estimation Technique for an End-to-End Network Path(2016)
[7] Machine learning for measurement-based bandwidth estimation (2019)
[8] Packet-Dispersion Techniques and a capacity-Estimation Methodology
[9] Locating Available Bandwidth Bottlenecks

网络瓶颈链路容量与最小时延计算

公式来源于[1].
$$C=\frac{W_2-W_1}{D(W_2)-D(W_1)}\text{\hspace{1em}(1)}$$
$$D_{min}=\frac{W_2{D}_1-W_1{D}_2}{W_2-W_1}\text{\hspace{1em}(2)}$$
[1]Throughput metrics and packet delay in TCP/IP networks
[2]What type of distribution for packet delay in a global network should be used in the control theory