RELENG_8 ignoring TCP window size? [Was: Re: Help for TCP understanding wanted, ACK-MSS-Window [Was: Re: best practice to watch TCP parms of established sockets]]

[Stephen Hurd]

Stephen Hurd wrote:
>> Some experimental results:
>> When rsyncing with windows, and FreeBSD is receiver, I see the same 
>> ACK ever two segemnts, but speed is at 72MB/s.
>> When FreeBSD is sender and Windows is receiver, it looks more I 
>> expected. There are about 20 data segments before a ACK is returned. 
>> And there are  TCP Window Update Segments, reflecting smaller 
>> receiver buffers on the windows side. But this happens at a 
>> throughput of 82MB/s!!! So the windows machine is behaving like I 
>> understand the TCP flow control.
>> Any explanation why the FreeBSD machine seems to ignore window size?
>
> The idea of delayed ACKs is to allow an ACK to be sent with data if 
> there will be data sent right away, not to combine ACKs... leaving out 
> ACKs makes calculation of RTT problematical which causes performance 
> problems all over the place... maybe the dearth of ACKs from the 
> windows system is causing the problem?

If the problem is the ACKs from the Windows system, you should be seeing 
a large number of retransmits from the FreeBSD system (on the order of 
one retransmit per five packets).  Is this what you're seeing?  If you 
have a capture you could share covering a few seconds, I could take a 
look and provide a better opinion.