CGMP on 3500XL's

[gpingel]

I am trying to get CGMP working successfully on a small flat network consisting of 8 3548XL switches on the same VLAN. CGMP is running on all the switches, and I have a 2610 router (10Mbps interface) plugged in as the multicast router with the following commands entered

ip multicast routing

interface e0/0

ip address 172.20.0.5 255.255.255.0

ip pim sparse-dense mode

ip cgmp

The CGMP seems to work correctly where only the group members hear the multicast (Norton Ghost) conversation, but the ghosting of the images is very slow because the 2610 ethernet interface is pegged at 10Mbps during an image push to the client PC's from the ghost server. When we create the image from the client PC, and push the image back up to the server, it is very fast and the 2610 e0/0 interface is barely being used. Anyone know why this would be so slow one way and very fast the other way? Any help would be appreciated.

[milan.kulik]

Hi,

I spent a lot of time finding a Ghost multicast solution for my network two years ago. But my problem was only preventing multicast flooding to all my switches. We are using Ghost with a server and clients in the same VLAN, so no multicast routing involved.

You can find my solution description in this forum archive.

The difference between your and my router confing is that I'm not routing multicasts (no ip multicast routing in my config) and I use ip pim sparse mode.

But I'd guess that your Ghost server is in a different VLAN (or subnet) then your clients, so multicasts are really routed through your e0 interface when an image is being sent to the client.

And I'd also guess that in the case an image is being sent from client to the server, unicast destination address is used. So no multicast routing is involved.

I'd try to use unicasts in both directions. Latest Ghost version is able to do so when sending an image to a single PC - which I estimate is 98% of Ghost use.

Regards,

Milan

[gpingel]

Thanks Milan, actually the sender and receiver are on the same VLAN, so I guess i don't need the ip multicast routing statement.

So the next question would be should I force this to ip pim sparse mode? To do this I believe I need to advertise a RP?

ip pim send-rp-announce Ethernet0/0 scope 5

I looked in the archive and I think I found the link you were toaking about

http://www.cisco.com/warp/public/473/38.html#prereq

Do I need to specify a Join group with the Multicast address of the ghost server?

Thanks again Milan.

[milan.kulik]

Hi,

you don't need to specify anything.

ip pim sparse-mode

ip cgmp

are the only router config commands used on the proper interface if the Ghost server and clients are situated in one VLAN.

I've also configured

port storm-control broadcast action filter

port storm-control broadcast threshold rising 2500 falling 1500

port storm-control multicast action filter

port storm-control multicast threshold rising 3000 falling 2000

on the 3548 port to which the Ghost server is connected.

If your network contains any CatOS switches, you need to configure

set cgmp enable

on them.

On 2950s (or other IOS switches which don't support CGMP by default), you need to configure

ip igmp snooping vlan x mrouter learn cgmp

This should prevent Ghost multicasts flooding your network.

Regards,

Milan

[jrgarrigues]

Milan,

I'm interested in reading your solution, but cannot find it. I am facing a situation that sounds almost exactly like what you are describing. Would you mind posting a Topic Title and Date so that I can narrow down my search?

Also, how did you determine that the multicast was flooding on your network?

Thanks,

J R

[milan.kulik · Accepted solution]

John,

I've discussed this problem several times in this forum, you can find my solutin in the Forum Archive

(go to the Search field, fill "kulik CGMP ghost" and choose to search in Archive, not All Forums).

Basic description is

http://forums.cisco.com/eforum/servlet/NetProf?page=netprof&CommCmd=MB%3Fcmd%3Dpass_through%26location%3Doutline%40%5E1%40%40.ee93165/7#selected_message

e.g.

"how did you determine that the multicast was flooding on your network?"

Well, our company network was build on non-Cisco devices originally.

When my colleagues had tried to use Ghost the first time (I was not the company employee that time), the network was flooded with multicasts and went down - nobody was able to work.

When I joined the company, one of my first tasks was to find the way of routine Ghost usage. So I made some research and finally found the solution.

Regards,

Milan

[jrgarrigues]

Milan,

Got it, thanks.

John