Subject: Re: Poor tcp performance
To: None <firstname.lastname@example.org>
From: der Mouse <mouse@Rodents.Montreal.QC.CA>
Date: 03/20/2001 00:51:41
>> I'm seeing poor TCP throughput. It looks to me as though the window
>> is 32K but there's never more than about 9K of data in transit.
> Are you going to bother stating what this connection is? I might
> *guess* from the port numbers that it's outgoing mail, but who knows?
Yes, it's an outgoing mail message.
> In addition, a snippet of a trace like this is ENTIRELY useless;
I included it to give some measure of precision to the comment about
32K of window of which only 8-9K is ever used.
> there are all sorts of secondary effects -- such as slow start,
> congestion window monitoring, etc. -- that could cause instantaneous
> behavior like this.
Oh, certainly. I've been watching this connection on and off for most
of its lifetime and I don't think I've ever seen more than 15K
difference between the sequence number in an outgoing packet and the
ack number in a returning packet. I suppose I *could* have just been
unlucky in my choice of moments to check, but that seems improbable.
When I have more spare time (which probably means in a month or two),
one of the things I intend to do is do a more detailed investigation of
this problem. My note today was largely intended to make sure it
wasn't something on the order of "oh, yes, I saw that months ago; it's
the mistake fixed in 1.107->1.108 of tcp_output.c". If I'd meant it to
be a more formal bug report I would have sent-pr it.
7D C8 61 52 5D E7 2D 39 4E F1 31 3E E8 B3 27 4B