Re: [lwip-users] TCP Throughput. Great work

[David Empson]

----- Original Message ----- 
From: "Chris Williams" <address@hidden>
To: "Mailing list for lwIP users" <address@hidden>
Sent: Friday, June 18, 2010 9:04 PM
Subject: [lwip-users] TCP Throughput. Great work


> In message <address@hidden>, David Empson
> <address@hidden> writes
> > Everything looks normal up to packet 39. Between packet 2 and 39 you're
> > transferring about 5 MB/s. The two window updates indicate points at which
> > the application on the PC read data from the TCP stack, allowing the PC to
> > report that it had more space to receive data.
> ......
>
> Can I just say what a great analysis of a wire shark output this is.
> detailed and yet concise!

Thanks for the kind words.

> You should set up a web page to teach us all how to use wireshark better.

Hard to say whether I could do that usefully. Much of what I spotted in that 
capture was due to intuitively spotting patterns and significant numbers 
(which I'm good at), and that quickly led to an analysis of the likely 
problems. I wasn't looking any deeper than the main list of packets and the 
detailed information about the TCP headers to see the sequence numbers and 
packet lengths (for packets where the summary line in the list of packets 
hides that level of detail).

There can be interaction with the implementation of TCP stacks, for example 
I had to assume the window size from the PC end was behaving "normally", 
i.e. decreasing as long as data was arriving but not being delivered to the 
application, and an increase implied the data was being read.

I've looked at other captures that people have posted but some problems are 
harder to analyse or not enough information is available merely from the 
packets transmitted and received without also having matching debug output 
from LWIP (which I find difficult to parse, so I usually don't have time to 
put enough effort into it).

Cheers