lkml.org 
[lkml]   [1998]   [Sep]   [8]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
SubjectRe: Very poor TCP/SACK performance
From
Date
>>>>> "David" == David S Miller <davem@dm.cobaltmicro.com> writes:

David> From: alan@lxorguk.ukuu.org.uk (Alan Cox) Date: Tue, 8 Sep
David> 1998 14:13:48 +0100 (BST)

> In 2.0.x I used 100% overcommit. With ethernet drivers doing
> rx_copybreak that comes out about right in the wash. It was
> originally a finger in air and gues exercise but it turned out
> near enough right

David> Ok, right.

David> The issue is to get the max benefit of SACK with even
David> moderately latent paths we need to up it a bit, this is the
David> crux of all the complaints we are getting.

David> So my question is, can we up the default safely and if so by
David> how much?

I would very much like to see an increased maximum value. I have been
running with significantly higher for a long time due to playing with
HIPPI and I have not experienced anything strange with my Ethernet
interfaces as a result of that.

Most commercial UNIXes seem to have a max limit of either 200000 or
262144, though I think many of them still have a default socket buffer
size of 8KB. Oh and SGI seems to go even higher.

Its not exactly scientific evidence, I admit ;-) Maybe setting the max
limit to 64KB for systems with less than 16MB and 256KB for systems
with more memory?

Jes


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/faq.html

\
 
 \ /
  Last update: 2005-03-22 13:44    [W:0.054 / U:0.492 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site