Here is a fairly comprehensive write-up of the structure and use of the linux tcp_info by René Pfeiffer but there are a couple of things worth noting:
The author needed to look at these data repeated over time because there are no aggregate stats in that structure.
The author directs you to the tcp.c source as the final authority on the meaning of any of those data.
I'm not sure what you were hoping to get from the Maximum Segment Size, but expect you thought it meant something else.
If you are truly interested in exact measurements of bandwidth you need to use a measurement device which is outside the system being tested as even pulling the ioctls will affect the phenomenon you are interested in knowing about. A passive wire sniffer is the only way to get truly accurate results. Finally, depending on your application, "bandwidth" is a really broad umbrella which flattens many measurements (e.g. latency, round-trip-time, variability, jitter, etc.) into one category.
发布评论
评论（1）
这是 René Pfeiffer 对 Linux tcp_info 的结构和使用的 相当全面的文章，但是有几件事值得注意：
如果您确实对带宽的精确测量感兴趣，则需要使用正在测试的系统外部的测量设备，因为即使拉动ioctls也会影响您有兴趣了解的现象。无源电线嗅探器是获得真正准确结果的唯一方法。最后，根据您的应用程序，“带宽”是一个非常广泛的保护伞，它将许多测量（例如延迟、往返时间、可变性、抖动等）归为一类。
Here is a fairly comprehensive write-up of the structure and use of the linux tcp_info by René Pfeiffer but there are a couple of things worth noting:
If you are truly interested in exact measurements of bandwidth you need to use a measurement device which is outside the system being tested as even pulling the
ioctlswill affect the phenomenon you are interested in knowing about. A passive wire sniffer is the only way to get truly accurate results. Finally, depending on your application, "bandwidth" is a really broad umbrella which flattens many measurements (e.g. latency, round-trip-time, variability, jitter, etc.) into one category.