[ntp:questions] Detecting bufferbloat via ntp?
Rick Jones
rick.jones2 at hp.com
Thu Feb 17 22:42:52 UTC 2011
Dave Täht <d at taht.net> wrote:
> Rick Jones <rick.jones2 at hp.com> writes:
> > Or rather, that even after setting the tx queue lengths to 32
> > packets, a test between that system and one 7ms away still
> > resulted in 4MB socket buffers by the end of the test. Ie
> > confirming that the linux autotuning code was still willing to
> > grow the windows larger than necessary.
> Better said.
> The evidence of bloat is not conclusive. Did you give vegas
> a shot?
Nope. While my posting here suggests otherwise, I've been rather busy
with the "day job" :)
> >> A *little* is just fine. Bloated buffers - containing hundreds,
> >> thousands, tens of thousands of packets - which is what we are seeing
> >> today - is not.
> >
> > Well, the BDP of a 10GbE link might actually be measured in thousands
> > of packets or more... if my systems 7 ms apart were joined by a 10
> > GbE link, that would be a bit more then 5800, 1500 byte packets. I'm
> > thinking that while we may have to configure queues in terms of number
> > of packets, we shouldn't think of them that way, but as length of
> > time.
> I agree, the dynamic range of todays devices presents a problem.
Perhaps then, to keep the delays for NTP traffic small (trying to keep
on topic :) the user interface for configuring txqueues should take a
maximum expected RTT, which it can then combine with the known speed
of the interface to arrive at a number of packets for the queue. As
the default anyway. And if there is no information as to the maximum
expected RTT, perhaps use a non-trivial fraction of the IETF's
suggested minimum RTO. Then the "device" can say "I am running at N
bits per second and when I was created the RTO was 1000 milliseconds,
90% of which is 0.9 seconds, and my MTU is M so I should have a
default txqueue of 0.9N/8/M +1 packets. Now the default is automagic
no matter what the link speed.
rick jones
--
I don't interest myself in "why". I think more often in terms of
"when", sometimes "where"; always "how much." - Joubert
these opinions are mine, all mine; HP might not want them anyway... :)
feel free to post, OR email to rick.jones2 in hp.com but NOT BOTH...
More information about the questions
mailing list