Hi,
you have to main technology to control the amount of bandwidth allowed to each traffic flow: policing or shaping. Shaping can be used just in outbound direction (egress), policing can be configured both igress or egress. So the first thing is to determine which technology best fit your goal.
Usually shaping is preferred because it brings to a better use of bandwidth due to the use of buffers. Policing is usually used by ISP at their network boundaries to guarantee that customer can't use more bandwidth the agreed.
Inside a private network, shaping is generally preferred because burst are buffered and not just "cut" has policing do. On the other hand policing require less resource and does not introduce delay (because it does not use buffers) so it could be better for voice and other time sensitive traffic.
In case you decide for policing I suggest to configure on the WAN interface, it's useless to process packets that will be dropped on the LAN interface.
Finally let me say the, as you probably know, the right place to make this configuration is on the branch but, I suppose that you can't for some reason.
Bye,
enrico
PS rate if useful