multicast issues (pim sm based)

[joneswill]

Hi

I am having some problems with pim sm.

In my primary site I have two 4507s.  I have an anycast address for the RP which I have created on both core switches.  My reason for doing this is, if I lose one coreswitch, the other coreswitch will take over as the RP for the site.  If I lose both switches the other site will become the primary RP (via MSDP).  This makes sure that both sites mcast traffic remains local, and so I dont flood my wan links (between the two sites) with mcast traffic.

Can someone confirm that this is correct?  Should I configure both my 4507s at my primary site as PIM RPs, each with the same anycast IP address?

The gateway for all my vlans at my primary site is the core switches (hsrp handles which one is primary).  So effectively, my 4500s double as RPs as well as first/last hop routers for all internal lan traffic. 

What I have noticed is when mcast traffic ramps up I can see quite painful microburst issues which can cause brief saturation on portchannels that connect blade chassis's to my cores.  The portchannels are terminated onto oversubscribed line cards so there has always been a level of contention, however with the introduction of multicast, it appears that this has got a lot worse - to the point of causing delay/seconds worth of outages at peak times of mcast.

From what I can see it looks as if we've got to the end of the road with the capacity of our existing supervisors (sup-iv's) and their limit of 6gb linecards.  But I wanted to chuck this out there to see that Im not doing anything stupid first.

any ideas welcome.

thanks

[Amit Aneja]

The two core switches on the primary site are your RP & they will be communicating to the other site's RP via MSDP & if both of them are down, then there is no communication for multicast.

[joneswill]

If both my local site RPs are down, clients will use the RP from my other site, across our wan.  This is working and tested.

bumping for any answers to my question?  right now I'm assuming my only option is new sup's and line cards.

[boss.silva]

Hello,

Have you analyzed what type of traffic do you see traversing the port-channels with a sniffer? This way you can see exactly the behaviour when too much traffic is seen on the interfaces.

Also, please post your config so we can see exactly how this is set up.

The only problem i can see here is that perhaps you are also counting on routing metrics to choose the RP to be one of the 4507's, but if you have equal metrics, you could have some issues with load balancing depending on your configuration.

If you are using the HSRP Virutal IP address as the RP, i don't see any issues, but if not, i would take a better look at the design.

Regards,

Bruno Silva.