[j-nsp] How useful is Juniper storm control?

[Ben Dale]

Hi James,

On 15/05/2013, at 12:29 AM, James S. Smith <JSmith at WindMobile.ca> wrote:

> I'm looking for people's experience with storm control on Juniper switches.  We have a pair of EX4500 switches and I notice that storm control kicks in a lot.  I'm concerned that it might be stopping legitimate broadcast and multicast traffic.
> 

I would be concerned too - the fact that you're seeing this on a 4550 (assuming 10G interfaces) - the defaults for storm control are pretty high (80% of link bandwidth for combined BUM traffic).  If you do a lot of multicast on your network, this may not be appropriate though, and there is a knob to specifically disable multicast from being counted:

set ethernet-switching-options storm-control interface all no-multicast

Without thinking too hard (it's late here), I would think with IGMP-Snooping turned on, there shouldn't be too many situations where a loop would cause multicast traffic to increase dramatically.

> So is storm control useful in a spanning tree environment, or is it unnecessary?

I'm of the opinion that it is useful for the same reason that you enabled spanning-tree when you know there are no loops in your topology - it's there for that day when your intern gets their first simultaneous lesson in patching and broadcast domains.  

That said, I think though that you need to spend some time tuning it down for your environment.  Allowing up to 8Gbps of broadcast traffic isn't exactly "control".

Ben