Title says most of it. I have a 2/2 Gig pipe and expect to cap out the actual bandwidth needs at around 1 Gig. I use an ASUS PN42 N100 as my router.
I was excited to set up SQM (still learning about it all), but discovered that my bufferbloat score is already A+ with latency of around 4-5 ms. With SQM enabled, I take a bit of a hit on top speed and latency goes down to 3-4 ms, if that.
My question is: assuming I will always be comfortably under the 2 Gig speed cap, is there any benefit at all to enabling SQM?
Use case: I want my son's school/work videoconferencing traffic to take absolute priority over everything else. I'm happy to have lower latency on that if it means sacrificing max speed a bit. So I guess to phrase my question more specifically: is it possible to have bufferbloat (theoretically? what about realistically?) in my conditions?
Can, in theory, some TCP packets get stuck in the buffer on my router, despite the 2 Gig bandwidth, thus delaying my son's video UDP? In that case, I suppose it'd still make sense to have a queue manager that bumps these UDP packets to the front of the queue. I just don't know if such a scenario is even realistic. Thank you!!
As more subscribers join the top speed in your surroundings it may need review later, but for now sqm does not improve or solve any problem.
There is more to udp, like hrrp/3 to grandparents favourite news site. I doubt its easy to differentiate that from 8k youtube.
SQM is a neat package containing:
a) a traffic scheduler
b) an active que management
c) a traffic shaper
c) is only needed so a) and b) get control over the true bottleneck queue...
That out of the way, a) and b) can help whenever queues build up, and that happens typically at places where capacity reductions happen. For many folks such a reduction step is going from Gigabit LAN to <<1 Gbps WAN, in your case I would guess a similar step to happen in the WAN to LAN direction, unless your LAN is already at 2G5 Ethernet of faster... (and in that case the reduction step is in the traditional LAN-to-WAN transition). But whether that theoretical observation comes to bear depends on whether you ever saturate the WAN link... If you never saturate that link no, SQM is not going to help. However if you expect even occasional saturation then SQM might be a nice back-stop against the sideeffects.