[c-nsp] Cisco Switch Packet Buffering Matrix?

[Gert Doering]

Hi,

On Thu, Jun 03, 2010 at 07:11:29AM -0500, Jeff Bacon wrote:
> It's all ok, even with bursty traffic (sorta), as long as your inputs
> match your outputs, e.g. there's nothing to buffer. It's when that's not
> the case that everything goes to hell.

Well, if there's nothing to buffer, then the buffer size is pretty much
irrelevant :-) - the problem is microbursts (two ingress ports bursting
at full line rate for 100ms, then idling for 900ms - makes an average
rate of 200 Mbit/s at GE ports, but when sending this traffic out via
a single GE port, you'll already see drops...)

> The E/X models are completely different, no idea on those. (Have a
> couple of -Es coming. Only reason I accept them for the purpose is that
> I'm going to use them to break up a 10G inter-datacenter-link into 10 1G
> channels, so there's no reason for anything to have to buffer...)

Mmmmh, to me this very much sounds like "a 10G microburst coming in via
10G port, leaving via 1G port" *will* need buffer space...

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/20100603/3de7836d/attachment.bin>