In order to support desktop imaging requirements I need to correct the multicast design on an existing network, but am not well versed in this subject. The diagram below gives a basic example of the topology. The Juniper EX8200 is the hub in a simple hub and spoke design. It has 1Gb L3 interfaces directly connected to each of the other sites. OSPF is enabled on these interfaces to exchange routes with Cisco devices at the other sites. The hub and each site will have one or more VLANS trunked to closet switches. My understanding up to this point is that on each of the VLAN interfaces at each site, I should have both IGMP and IGMP-Snooping enabled. On the WAN connections, I should have PIM dense mode enabled on the Juniper side, and sparse-dense-mode enabled on the Cisco side. For the Juniper's directly connected VLANs, I'm unsure which mode to use, but since its the largest environment with the most clients, I would expect to simply use dense mode on all of them. So here are my questions:

1) Do I have to concern myself with assigning addressing anywhere here? 2) Are there any major flaws in this design?

Heres is a basic representation of the network topology

  • Really? Nothing yet? Surely someone is versed in this. So far I've been fairly validated in understanding to use PIM on router interfaces between sits and a combination of IGMP and IGMP snooping access switches. May 9 '11 at 14:29
  • It might help if you went into more detail about what you are trying to accomplish. I don't understand what multicasting has to do with "desktop imaging requirements"
    – user65237
    May 11 '11 at 14:19
  • Desktop imaging tools typically utilize multi-casting to increase efficiency. I have 1Gb connectivity between sites, but with a 20GB image and 500 remote machines multi-casting would mean the difference between transmitting that 20GB once, or transmitting it 500 times. May 11 '11 at 20:47

Do you have specific problems with your setup? It looks roughly okay, although I still cannot understand where your multicast sender is located.

You probably would not need PIM - your diagram looks like there always will be exactly one path from each host to another passing through the Juniper router. Introducing dynamic multicast routing here is probably overkill and just introducing additional complexity. And on interfaces where you do not expect to have further multicast routers (like your "directly-connected" VLANs) you surely would not need it at all - IGMP will take care of everything that's needed there.

IGMP snooping will help your L2 switches reduce broadcast traffic, but will not help functionality - multicasts are simply flooded to all ports (just as broadcasts) if it is not enabled, so for troubleshooting purposes you might want to disable that.

  • It seems to be working fine now, I was more looking for confirmation of design. Some of the site routers were converted from IOS to JunOS, when this happened the converter turned PIM on every VLAN, this caused lots of overhead, weird responses. I made adjustments as described, PIM site to site, IGMP on each VLAN, things appear to be working. I just wasn't versed in it and was looking for some validation. May 13 '11 at 12:50
  • I ended up having to enable PIM in sparse-dense mode in order for things to work between VLANs on the same switch. Jun 1 '11 at 21:01

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.