CBWFQ or other policy / Congestion

I have a remote site that has a point to point T1 to our main site. The remote site has an application that streams data at a fairly constant rate of 800Kb to the main site (udp based). This stream is not 24x7 as it kicks off at various intervals depending on when the data is needed.

We also have a second application that has a data stream of 200Kb which is also not constant. This stream when it is being sent is more important than the 800Kb stream.

Every once in a while a user will kick off some file transfer and the pipe will get congested and cause issues with starving the two data streams. My goal is in times of congestion make sure that both the

800Kb stream and 200Kb stream are not interrupted.

I was thinking of implementing CBWFQ for this, but I am not sure if this is the right policy.

My questions is relating to what will happen in times of congestion?

[1] Between the two classes (critical and priority-data) the guaranteed bandwidth is 1125Kbit. This leaves roughly around 400Kbit for the default class. Now if the 800Kbit stream is NOT in use will the default class be able to tap into that bandwidth? Or will it always be limited to the remaining bandwidth left from the other classes? [2] If the bandwidth for the priority-data class exceeded 225Kb would it be tail dropped even in times where congestion was not present (bandwidth was available for it to be used)?

Config: class-map match-all critical match dscp af41

class-map match-all priority-data match dscp af31 ! policy-map qos-out-policy class critical bandwidth 225 class priority-data bandwidth 900 class class-default fair-queue

Thanks Amy