On Wednesday, January 08, 2014 06:12:44 PM Daniel Rohan wrote: > Would anyone running LDP + RSVP + QoS on their network > mind contacting me on or off list to share their > experiences? I'm contemplating a design that would > involve using LDP for L3VPNs and RSVP for some minor > traffic engineering + L2 services on a small (10 router) > ring topology. The reason I'd like to use LDP is because > one of the L3VPNs would have a presence on every single > router on the ring and setting up a full RSVP mesh seems > cumbersome, and the RSVP auto-mesh features seem poorly > documented on Juniper's end. But my major concerns are > with QoS; it would appear that RSVP lends itself to an > IntServ set up, but to keep consistent with the way we > run a few other networks, I'd like to keep a standard DS > environment on this backbone. My concern is that I can't > find resources that talk specifically on how to make > RSVP DSCP aware short of setting up a DS-TE RDM/MAM > environment, and I don't want to get into a situation > where one signaling protocol isn't aware of the traffic > on the links.
I would disassociate the inherent QoS designs of RSVP from the QoS paradigm existing in routers today, and rather, compliment them. Use regular DSCP to mark/schedule your IP traffic as it enters its VRF on your network. When it turns into MPLS for forwarding on to other participating PE routers, use EXP to mark/schedule that MPLS traffic throughout your core. The DSCP attributes you'll have used at the point of ingress will be re-used at the point of egress, provided you have a consistent PHB across your backbone. Mark.
signature.asc
Description: This is a digitally signed message part.
_______________________________________________ juniper-nsp mailing list juniper-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/juniper-nsp