[c-nsp] 7301 NPE-G1 gigabit overruns

[Rodney Dunn]

On Wed, Feb 11, 2009 at 04:26:07PM +0200, Tassos Chatzithomaoglou wrote:
> We had a similar case with microbursts and lots of small packets and 
> after moving to NPE-G2, we're very happy with it.
> Traffic goes up, CPU is blocked at 75%, still no sign of drops.

Good to know. It all depends on the load.

The direction forward is hw based forwarding for gige line rate
traffic (ASR1000, 76xx, etc..)


> 
> Regarding the pause frames, i'm too very confused about them being 
> supported.
> http://markmail.org/message/y7bjuudmbrfofp6z

I've done some research and there appears to be some issues on
how the MAC chip can account for the pause frames sent/received
although it does appear to support flow control.

But my *opinion* on that is flow control is a wasted effort becaus
you can't quantify the burst and therefore you just buffer on the
other side and risk drops there.

Rodney

> 
> --
> Tassos
> 
> Lars Lystrup Christensen wrote on 11/02/2009 15:47:
> >Hi John
> >
> >We had the same issue a few years ago on a Cisco 7204VXR with the 
> >NPE-G1 processor. I believe we were told, that we had hit the max 
> >capacity of the router.
> >
> >______________________________________
> >
> >Med venlig hilsen / Kind regards
> >
> >Lars Lystrup Christensen 
> >Director of Engineering, CCIE(tm) #20292
> >
> >Danske Telecom A/S
> >Sundkrogsgade 13, 4 
> >2100 K?benhavn ? 
> >
> >-----Original Message-----
> >From: cisco-nsp-bounces at puck.nether.net 
> >[mailto:cisco-nsp-bounces at puck.nether.net] On Behalf Of john douglas
> >Sent: 11. februar 2009 14:36
> >To: cisco-nsp at puck.nether.net
> >Subject: [c-nsp] 7301 NPE-G1 gigabit overruns
> >
> >Hi All,
> >
> >I am hoping someone may have some guidance.
> >
> >We have some 7301 NPE-G1 running 12.2(33)SRC3 connected to Catalyst
> >3560 switch, dot1q trunking with a few Vlans. I find on the 7301
> >router whenever I try to drive a given gigabit sub interface beyond
> >about 200mbps I begin to see Input Errors (Overruns) incrementing on
> >the physical interface and begin exhibiting packet loss. Under the
> >show controller information I see rx_overrun and rx_int_drop
> >incrementing - all other Rx error values under "Statistics" section
> >are zero. On the switch I see pause input incrementing, but no 
> >errors.
> >Cpu is 40%.
> >
> >I have read other postings about this being attributed to microburst
> >of traffic but I am still a little unclear I am trying to work out
> >
> >1) Does the G1 correctly support pause frames?
> >
> >2) Are there any knobs we can tweak - or we simply hitting the limits
> >of the silicon?
> >
> >The short term solution has been to spread the vlan sub interfaces
> >over the 3 gig interfaces however this leaves me in a situation where
> >I cannot really use EEM to migrate sub interface config from one
> >gigabit interface to another should there be a failure, as the target
> >interface will being exhibiting overruns and loss.
> >
> >Any advice appreciated,
> >
> >Kind regards,
> >
> >John
> >_______________________________________________
> >cisco-nsp mailing list  cisco-nsp at puck.nether.net
> >https://puck.nether.net/mailman/listinfo/cisco-nsp
> >archive at http://puck.nether.net/pipermail/cisco-nsp/
> >_______________________________________________
> >cisco-nsp mailing list  cisco-nsp at puck.nether.net
> >https://puck.nether.net/mailman/listinfo/cisco-nsp
> >archive at http://puck.nether.net/pipermail/cisco-nsp/
> >
> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/