Bufferbloat related censorship at Virgin Media

Dave Taht dave.taht at gmail.com
Mon Mar 2 00:48:07 UTC 2015


http://burntchrome.blogspot.com/2014/05/disabling-shaping-in-one-direction-with.html

On Sun, Mar 1, 2015 at 4:39 PM, Jack Bates <jbates at paradoxnetworks.net> wrote:
> On 3/1/2015 6:14 PM, Dave Taht wrote:
>>
>> It is 100% possible to fix excessive downstream buffering from some
>> misconfigured device with a shaper on the download  *on the CPE or
>> home router*.
>>
>>
> From OP: "However I've recently noticed periods of 500-800ms latency to the
> CMTS gateway when only using 15-20 of the 60Mbps total (and little to none
> upstream utilisation)."

Might be. Again, all I did on that thread was provide a few pointers
to bufferbloat related resources, and pointed at the downlink being a
real problem quite often with links to stuff like this

http://burntchrome.blogspot.com/2014/05/disabling-shaping-in-one-direction-with.html

and they yanked me. Certainly I could have tried again, from another
IP, but ya know, some sundays are more fun than others.

>
> I agree with you that it is better to run a shaper that insures your shaper
> hits saturation and handles queue policies before the upstream does. That is
> great if it is your pipe (and only its queue) that is saturating. I don't
> think this problem qualifies.

Might not. That said, it was hardly an accurate measurement. It is
also perfectly feasible for the upstream device or the downstream
device to be measuring these problems and deal with them
appropriately. It gets progressively easier cpu-wise, as the effective
bandwidth goes down.

It is unfortunately nearly impossible for the next device in line to
do (although we have some tools measuring interpacket "smoothness"
that can provide a hint now, they are not baked yet)

It was my hope, in working on the DOCSIS 3.1 standard that all the
possible downstream problems would be addressed. They weren't.

> I find it difficult to believe that he's hitting a buffer bloat issue on a
> single (not shared with others) queue using 1/3rd of the total bandwidth
> available to him at those speeds and with that latency value. His problem is
> more likely lower down (unable to obtain max speed resulting in saturation)
> or a shared queue where others are saturating it and him applying a shaper
> will not keep others from doing so.
>
> Jack



-- 
Dave Täht
Let's make wifi fast, less jittery and reliable again!

https://plus.google.com/u/0/107942175615993706558/posts/TVX3o84jjmb


More information about the NANOG mailing list