cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
3444
Views
15
Helpful
5
Replies

Multicast Anycast-RP with static RP or with BSR

cedar_lee
Level 1
Level 1

Which one do you think is better for multicast RP redundancy, Anycast-RP with static RP mapping or Anycast-RP with BSR?

And why?

Thanks

Cedar 

1 Accepted Solution

Accepted Solutions

daniel.dib
Level 7
Level 7

I prefer to use a static IP that is anycasted. It's just one command to enter in the router and traffic will go to the closest RP. With anycast RP you will need MSDP of course so that all RP's knows of all the sources.

The argument for BSR is that you don't need to enter any commands on the routers to receive the RP and that you could easily change the RP without adding any commands to the routers. In my experience you would rarely change the RP and preferably your changes should be automated through some form of scripting anyway.

A potential issue with BSR would be if someone by mistake advertises itself as an RP through BSR and has a better priority than the correct RP or the same priority and a higher IP address. If using BSR, make sure the correct RP has the best priority.

When deploying multicast you need to add PIM on the interfaces so adding one extra command for a static RP won't add much to the work that is needed. There's also the added advantage of running less protocols in your routers which means less protocols that can have bugs/exploits.

In the end they are both viable options but I prefer to set a static RP.

Daniel Dib
CCIE #37149
CCDE #20160011

Please rate helpful posts.

View solution in original post

5 Replies 5

daniel.dib
Level 7
Level 7

I prefer to use a static IP that is anycasted. It's just one command to enter in the router and traffic will go to the closest RP. With anycast RP you will need MSDP of course so that all RP's knows of all the sources.

The argument for BSR is that you don't need to enter any commands on the routers to receive the RP and that you could easily change the RP without adding any commands to the routers. In my experience you would rarely change the RP and preferably your changes should be automated through some form of scripting anyway.

A potential issue with BSR would be if someone by mistake advertises itself as an RP through BSR and has a better priority than the correct RP or the same priority and a higher IP address. If using BSR, make sure the correct RP has the best priority.

When deploying multicast you need to add PIM on the interfaces so adding one extra command for a static RP won't add much to the work that is needed. There's also the added advantage of running less protocols in your routers which means less protocols that can have bugs/exploits.

In the end they are both viable options but I prefer to set a static RP.

Daniel Dib
CCIE #37149
CCDE #20160011

Please rate helpful posts.

Daniel,

What is your view on static vs auto RP with regards to faster convergence for Multicast ?

I am aware of other features such as MoFRR and PIM or RPF timers tuning, bit just want to hear if there are any advantages for using static RP.

With regards to bugs, there is one bug affecting some XR platforms for auto RP...so static RP definitely comes in handy here.

I would argue that convergence with a static RP would be faster. With a static RP, there is no need to disseminate information. If your currently closest RP goes down, packets will be rerouted to the next RP available based on your routing protocols converging. This can be very fast if properly setup and tuned.

Also, keep in mind that in PIM ASM, the RP is only really needed when the source register and/or when a new receiver asks for the multicast. If that flow is already established, then traffic will flow on the (S,G) just fine so only things added after that point would actually break.

If you run auto RP or any control plane protocol, then first auto RP must learn of the failure before it can send updated information which would require the current mapping to time out. I'm not fully aware of the details and timers of auto RP but this could take significant time.

The good thing with multicast is that it is based on the unicast routing topology which means that if you have a properly tuned IGP and routing, then you can have very fast convergence because PIM Joins are triggered when the RPF interface changes which makes convergence very fast.

In closing, I believe static RP would be much faster unless the MSDP SA gets delayed for some reason and that is only a problem if the source and receiver are not routing towards the same RP. With auto RP you are more dependant on a control plane protocol to disseminate the failure of the RP which takes more time.

Daniel Dib
CCIE #37149
CCDE #20160011

Please rate helpful posts.

Thanks Daniel, This is exactly what I also concluded.

cedar_lee
Level 1
Level 1

Totally agreed.

Thanks Daniel for your comment.

Cedar

Getting Started

Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community: