Auto-RP Setup

[dtom]

We have 20 sites each with their own local network connected to a central site via WAN connection. Initially, there will be a multicast host and clients (that want to connect to this multicast host) at each site. Clients later down the road may want to connect to other multicast hosts (other sites). We want to use pim sparse mode with auto-rp. We need some suggestions on how to set this up. Each router as an RP? How may RP's do we need? How many mapping agents do we need?  Where should the RP's and mapping agents reside?

[Richard Burts]

From your description I would say that you certainly do not want each router to be an RP. For 20 remote sites and one central site I would think it would be fine to have one RP at the central site.

There might be some question about whether you should configure a second RP to be redundant in case there is a problem with the primary RP. Whether that makes sense or not depends on how the network is set up. Do remote sites communicate directly with each other? Or does traffic from a remote site go to the central site and then is forwarded to the other remote site? If all traffic from the remote site goes to central site (tyypical hub and spoke network) is there a redundant path from the spoke network to the central site?

If spokes can communicate directly with each other or if there is a redundant path from the spoke to the central then there may be benefit in configuring a second RP. If each spoke has only one path to the central router, then I would say configure 1 RP, put it on the central router. There would be no benefit in a second RP in that case.

HTH

Rick

[dtom]

Thanks for the answers.  So, the purpose for using multiple RP's is for redundancy?

[Richard Burts]

There could be several purposes for having multiple RPs. I believe that redundancy is the most common, but not the only one. You could set up multiple RPs for the purpose of load sharing. I have seen a network which had several (fairly large) sites. Each site had several multicast sources. The configured their multicast so that sources at site A used one range of multcast addresses, sources at site B used a different range of multicast addresses, and sources at site C used a different range of multicast addresses. They configured one RP at site A which was responsible for the A range of addresses, an RP at site B that was responsible for the B range of addresses, and an RP at site C that was responsible for the C range of addresses. So each RP carried part of the load.

HTH

Rick

[dtom]

ok thanks for the reply