[c-nsp] mitigate output drops on Cisco 2960G platform

[Gert Doering]

Hi,

On Fri, Sep 20, 2013 at 07:23:17PM +0100, Phil Mayers wrote:
> (However: I wonder if the various techniques being discussed elsewhere - 
> like the new Linux kernel packet pacing, disabling TCP segmentation 
> offload and reducing nic ring, kernel and app buffers - might alleviate 
> the generation of microbursts?)

That actually sounds *very* interesting.  Especially TCP segmentation
has bit me elsewhere as well, on 1G->100M speed translations, with an
AIX machine firing out something like a 15-packet-burst back-to-back
(due to "one big TCP segment being stuffed into the card") and that
overflowing a stupid switch on the path...

(That particular problem was "solvable" by moving the AIX box to a 100Mbit
port, which immediately made all users behind WAN links much more happy,
as they did not have packet loss at noticeable RTTs anymore...)

Getting off-topic, but the whole idea of people actually working on this
is something I find exciting :-)

gert
-- 
USENET is *not* the non-clickable part of WWW!
                                                           //www.muc.de/~gert/
Gert Doering - Munich, Germany                             gert at greenie.muc.de
fax: +49-89-35655025                        gert at net.informatik.tu-muenchen.de
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 305 bytes
Desc: not available
URL: <https://puck.nether.net/pipermail/cisco-nsp/attachments/20130920/54a2c194/attachment.sig>