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

[Nick Hilliard]

On 09/02/2011 19:10, schilling wrote:
> I am familiar with these features. I talked with Cisco TAC several
> times, they are not recommending the storm control since it can not
> differentiate control data from user data, this might cause
> instability of layer 2 network.

This is true on core ports, which is one of the reasons why it's 
important to constrain the size of your layer2 domains.  However, storm 
control is critical on access ports.

> port-security to only allow specific
> mac address might be helpful, but will not be useful for a hub.

Hub?  Urgh, you need to remove this dangerous item from your network, 
pronto! :-)

> So there is no good way to prevent rogue hub/switch from messing with
> our network?

No.  Hubs are trouble, and unless you control the rogue switch, and the 
switch has decent quality port security features, then that will also 
cause trouble.

> So the best we can do is to reduce the fault domain, if something
> messed up, just let it mess up a small area of network?

You need to reduce your fault domain as part of a controlled redesign of 
your network, which will involve partitioning of the network into much 
smaller areas, installation of equipment which has the features and 
functionality that you need, removal of older equipment which is 
actively causing trouble, creation of access policies and templates for 
access and core ports, examination of dot1x (this is a contentious 
point), right down to creation of policies for dealing with people who 
feel that this restructuring is going to impinge on their carefree 
lifestyles.

Also, don't use VTP unless you like living dangerously.

Hyping your network with an MPLS core and using EoMPLS / AToM will give 
you lots of string to hang yourself with.  There are plenty of 
legitimate design reasons to use MPLS as a transport for your L2 core, 
but dealing with edge stability problems is not one of them.

Nick