Just started to look at domos whitepaper for quality attenuation:
"Typically, a traffic stream of randomly-sized packets with an overall rate of
16kbps is sufficient to measure the ∆Q along a network path."
Let's see how applicable this is on a metered link:
(16*1000/8)*(60*60*24)/(1024^2) = 164.8 MB/day
(16*1000/8)*(60*60*24*30)/(1024^2) = 4.9 GB/month
this is a lot of traffic for say anything below 10GB/month plan, no?
Yet this is still less than running a ping stream at 50Hz which comes out at
roughly twice the measurement traffic volume.
But I digress...
Essentially for endusers the only thing reasonable/possible seems to measure
the final aggregate static and variable delay/loss components... anything more
involved seems to require participating friendly measurement nodes along a path
that ISPs inside their own network have access to , end-users however somewhat
less.
Just try a longer running mtr trace to your ISPs DNS servers, and note
the progression of delay and loss along the path; due to rate-limiting and
de-prioritization often intermediate hops show increased packet loss and delay
that is not reflective to delay and loss reported by the end node. The upshot
of which is these are useless for assessing "quality attenuation" for non-ICMP
traffic which will not be subject to rate-limiting and de-prioritization. So
with my end-user hat on, I am not sure what to do with this, as I can not
verify these numbers...
_______________________________________________
Bloat mailing list
Bloat@lists.bufferbloat.net
https://lists.bufferbloat.net/listinfo/bloat
