Been using Comcast for a long time now, there was a news few years back that DOCSIS 3.1 will take care of the bufferbloat issue, i still get F on the bufferbloat tests, not sure what they really meant by DOCSIS 3.1 fixing the bufferbloat if it still gets F on the tests. I have Been running OpenWRT to get the A+ instead. Anyone knows what they really meant?
DOCSIS 3.1 mandates that modem/CPE have to implement a PIE traffic shaper for the egress/Internet-Upload direction. IIRC it also mandates that by default that shaper should be active. Which should massively reduce the bufferbloat in upload direction. But ISP are obviously free to disable that, and if run in DOCSIS < 3.1 mode even newer modems are not required by the standard to activate the shaper. As far as I can tell not all ISPs switched to 3.1 completely but often retain a considerable number of 3.0 channels, as they otherwise would need to replace all customer modems, so I guess there can be nominally DOCSIS 3.1 cable networks where the upstream still runs over 3.0 without the shaper activated.
Unfortunately, CableLabs did not mandate that such an upstream shaper should be under the control of the end user....
Hope that helps a bit...
that explains why the Upload spikes are in green, while download spikes are all over the place in Red on the DSLreports. not sure why they only fix just one thing and left the other one?
I believe this is because the downwload shaper already sits in the CMTS and there you have quite a lot of traffic and hence computations by the shaper become noticeable quickly and require more powerful/expensive hardware.
But it seems that low latency docsis might mandate PIE-style shapers at the ISPs side (it also does a bit more, some of it problematic IMHO, but I digress), now it is up to your ISP, but ISPs could have deployed competent downstream shapers if they so desired already, so I would not hold my breath...
1 Like