GRE over MPLS design question

[mmertens]

I have a customer who just migrated from point-to-point T-1s to an MPLS network. They have a hoot 'n' holler application that uses voice streams via multicast; therefore, I did GRE tunnels over MPLS to get multicast across the WAN. Right now, I have three remote site hoots that tunnel back to a Central Site hoot location, although all hoot (speaker phones) need to be able to hear all other sites. At the C/S, I have all three remote GRE tunnels terminating on the same loopback interface.

The customer is reporting intermittent choppiness on multiple hoots. Besides the QoS requirement, I'm wondering if my GRE design may be playing into this.......would I be better off meshing my GRE network so every site with a hoot has a tunnel to every other site? (There's only four sites total and probably won't grow beyond that). Or would there be any merit in terminating each tunnel to a seperate loopback interface at the C/S? Note: multicast interface command is only on the tunnel interfaces and not on the loopbacks.

Anyone have any suggestions as far as design goes with GRE?

Thanks!

[Mohamed Sobair]

Hi,

You need to make sure of two things:

1- MTU size is adjusted on the tunnel besides TCP-MSS adjusted on the LAN.

2- End to End QoS config to pritorize your voice traffic.

Do you have any bandwidth availability issues? Have you adjusted the (IP MTU ) of the tunnel interface?

HTH

Mohamed

[mmertens]

Mohamed,

   Thanks for the input- I did adjust the IP MTU on the tunnel interface but the problem continues...... The voice packets' size are going to be small and not close to the MTU size. The problem appears to be consistent, though. So I don't think it is a QoS issues, and I am marking them for DSCP EF.

   Any other thoughts?

Thanks!

Mike.