The scenario:

I am replacing a router / switch setup in a central cluster ina hosting center with a RB 1100 AH (just 2 servers at the moment).

I will use a bridget to an internal switch group to manage the servers there, use the other switch group for now as internal backbone.

There is a 100mbit uplinnk that is normally utilized for 1-3 mbit down, although it may peak up when I do downloads (windows iso images from microsoft, service packs etc.). or when the vpnto the 2office we do runs something. I hardly ever sees 30mbit together , though, andthe bandwidth from the offices is about 25.

1mbit of the traffic is HIGHLY time critical. It goes to a syste syncing its clock to 1ms accuary. Same to trffic from internal to 2-3 other locations, but ALL This is very low bandiwdth.

ANY use for QOS here? Given the hugh bnadwidth reserves I fail to see it.Possibly (and it will go in= on the VPN links to the offices (have it in the office upward), but i dont see the need for a generic QOS queue tree on the links to the internet and the internal server switch. Anyone disagrees with a good reason for me to learn?

I don't see the need for QoS either. You could create some mangle rules and queues to help with some indication of what is going through those links... just for peace of mind?

Thanks. Thought the same. Simply too much real bandwidth to bother. Obviously different on those parts that DO have limitations (vpn connections), but the raw box just hits it all by haiving so much reserves.