18 votesMark commented
Per-WAN definition of available bandwidth.
Traffic shaping and WAN load-balancing, as currently defined in the XG don't allow us to take advantage of knowledge of the available connections.
In our typical use case, we have a secondary connection which it'd be great to load balance over, but we need to guarantee that we never use the last 10~20% of that connection as it's reserved for high-priority services that cannot sit behind the firewall.
In the SG this was easy; you defined an bandwidth limit per interface when setting up QoS; I'm not sure why someone was possessed to come up with an interface in which you manually sum the total available bandwidth in KBps (connection speed A + connection speed B)*125
Also, the new interface allows no method to define inbound/outbound bandwidth separately. Not fun in the land of asymmetric connections.