[c-nsp] EoMPLS or VPLS loop prevention/storm control

[Peter Rathlev]

On Wed, 2011-02-09 at 10:12 -0500, schilling wrote:
> We right now have several bridged campus wide VLAN. It happens several
> times a year where a loop in one of the VLAN will cause our backbone
> to be unavailable. Now we are thinking to better architect the design.
> If we migrate to some platform like ASR9K and use EoMPLS or VPLS, what
> will happen if we have a loop in one of the VLAN?  The simple loop is
> to have a dump switch, connected two ports of it together.

Without knowing much about Nexus, I think OTV would be the right answer
to this if you're about to invest in new equipment. You might not want
Nexus for various other reasons though.

Using EoMPLS could maybe protect you a little more than a direct L2
connection, but not much. You might be lucky that the device closest to
the origin of the loop dies first, and since e.g. LDP signaling would
die with the control plane the other end might survive.

But it's not real protection. EoMPLS is L2 transparent and extends your
broadcast and STP domains just like a trunk would.

If you want to address the problem without investing in new equipment
you could look at mitigating the problems instead with some of the many
mechanisms that exist. STP BPDU guard, STP loop guard, storm-control and
CoPP (for devices capable of this) come to mind. That's where I would
start.

-- 
Peter