@Atul_Choudhary wrote:
Totally!! STP itself has nothing to do with broadcast storm, it just prevents the loop which can cause broadcast storms.
You don't need a L2 loop to have a broadcast storm, nor does a L2 loop, and its adverse impact, need any broadcast packets.
That said, certainly a L2 loop can contain broadcast packets/frames, and such would make a L2 loop even more impactful.
Again, what a L2 loop does is flood the loop's links with traffic, the same traffic looping round and round, forever, as L2 doesn't have the equivalent of a TTL. (Actually, if any host is still injecting traffic, some traffic will be dropped as switch buffers overflow. But that will not break the loop or reduce the overall volume of traffic.)
Again, what a broadcast packet/frame causes, every host must open it to determine it's of interest. Switches, that aren't visible (no IP nor MAC) as hosts within a L2 domain, would just not open such packets/frames, to them, it's just hardware forwarded.
The combination, though is particularly bad!
@Atul_Choudhary wrote:
Have seen numerous issues when worked in Cisco TAC where these storms choked out the network links, brought the CPU utilization high and made it very difficult to even get the access of switches to troubleshoot.
I don't doubt that at all! But a switch's CPU shouldn't be impacted by a L2 loop, even if 100% broadcasts, unless the switch has an IP or MAC on the L2 loop. Otherwise, to the switch, it's just a 100% busy port.
The impact of broadcast, and some multicast, traffic to a switch, as a host, is also why so many switches have "storm control settings" and/or CoPP, which have nothing to do with L2 loops, just dealing with floods of certain kinds of traffic.
(I am assuming, switch's CPU isn't used for forwarding/switching "normal" packets. If it were, CPU can be easily overloaded by a high volume of any kind of traffic, as many software based routers are.)
@Atul_Choudhary wrote:
My personal preference is always to use RSTP than MST unless we have a huge number of VLANs.
I've yet to use MST, but I do use rapid-PVST if it's not the default. This because I moved away from L2 topologies, decades ago, when small L3 switches came along. As a general rule, I never, for new or upgrade designs, span VLAN access ports across switches.
Lastly:
Just noticed, this reply and my prior reply was also made in the context of another thread dealing with C1300 lock ups, where STP was thought it might be a possible cause of the issue.
Hence my suggestion to disable STP for a C1300 in my prior reply.
Technically, this is known an oops!
So, my suggestion was specific to a different specific case, vs. this OP.
I do recommended running STP as a safety net.
I also standby what I've written about STP is for dealing with L2 loops, not broadcast storms, but the two can, be combined. Either can be, and often is, very bad! The combination might not be that much worse only because either, alone, can be so bad, it cannot get worse.
