All things related to multicast

["Roberts, Michael J. (IATS)" <>]

I can verify this was a problem for Nortel 8600 non-E revision cards.
It would spike the CPU to 100%, and anything (OSPF) that relied on the
CPU would fall apart.  This was fixed by Nortel in the E blades several
years ago.

We ran into an interesting problem on our campus where someone was using
a multicast group suggested by a software vendor.  It turned out this
multicast group fell within a reserved range that automatically forced
the CPU to process the packet, and not the ASIC on the phy blade.  You
may want to check out the group that is being used and make sure there
is nothing unique about the address that the Supervisor engine is
flagging for CPU processing.  

-mike

-----Original Message-----
From: Charles Spurgeon 
[mailto:]
 
Sent: Wednesday, February 08, 2006 9:03 AM
To: John Kristoff
Cc: 

Subject: Re: Notes from BOF

On Tue, Feb 07, 2006 at 04:53:51PM -0600, John Kristoff wrote:
> On Tue, Feb 07, 2006 at 05:35:11PM -0500, Jonathan S. Thyer JSTHYER
wrote:
> > >High (90+%) cpu on 6509s with sup2's when a multicast app is
> > >sending with a TTL=1.
> > 
> > I have observed this issue on our campus network.  Do you guys know
if 
> > Cisco has a bug id on this?
> 
> Yep, it's part number SUP720.  :-)  Seriously though, I'm not sure 
> if they have a bug id for it.  I've always been under the impression
> that this was an accepted, inherent limitation in the hardware for
> this platform in having to process switch TTL=0/1 packets, but
> perhaps someone else has additional details.

I recall hearing about a problem with Ghost setting mcast TTLs to 1
and causing high router CPU rates due to the TTL expiring and causing
the CPU to get involved for every packet.

There's a brief description of the issue here: 
http://www.unt.edu/dcsmt/multicast_ghost_application_supp.htm

I was told by a Cisco SE that the Sup720 platform could handle mcast
TTL decrement to zero without high CPU load but I don't know the
details (perhaps a rate limiter on TTL expired responses?)

Thanks,

-Charles

Charles E. Spurgeon / UTnet
UT Austin ITS / Networking

 / 512.475.9265