said by DaDawgs:
If you expect the customer to shape his own traffic then that setup is going to be more efficient than traffic shaping.
In this specific instance, I am the customer. I have a router in my upstream provider's datacenter, and I set the queue for the speed I am purchasing from them. I know they check my graph from time to time to make sure I'm not using more than I'm paying for, but it's never been an issue since I use the queue to limit myself to what I'm supposed to be using.
So far, the queue has worked quite well. As I approach or reach my limit, the queue kicks in and ping times start creeping up, but traffic still flows quite nice. If I were to set a hard limit, I'm assuming it would be less CPU intensive, but would I also see every packet over the limit be dropped, rather than queued? It seems that a slight increase in ping times is better than just getting packet loss, right?