Taking a step back from internal QoS designs, one thing that I have worked on for a large portion of both my professional and consulting career is border queuing. To take a walk down memory lane and help explain where my perspective is rooted, I wrote a custom OpenBSD based distribution back in the early 2000s that used OSU flow tools flow data and some poorly written perl to jam multi dwelling unit abusers in large apartment campuses into QoS groups dynamically, built and deployed some early L7filter based gateways in Linux for the same application, wrote L7 Linux based scripts for DD-WRT and generally played with any open source or commercial QoS devices within a reasonable budget.
Nowadays, I spend my consulting time tuning pfsense installs for HSFC based QoS and professionally have extensively tested A10 devices, IOS micro flow based rate limiting, Juniper MX based rate limiting, Palo Alto Networks devices, Procera applicances, Sonicwall devices and a number of other boxes capable of doing >1Gbps deep packet and L7 based QoS at network borders.
In my experience, I have yet to find one that I'm completely satisfied with. I'll admit that I'm a tad idealistic. Some are very, very close, but I am curious as to what others are using? How do others ensure user experience on a potentially free-for-all network? How is user experience ensured in this BYOD, bandwidth hungry world?