[ntp:questions] What traffic from pool is normal ?

Rick Jones rick.jones2 at hp.com
Wed Jun 22 00:10:01 UTC 2011


E-Mail Sent to this address will be added to the BlackLists <Null at blacklist.anitech-systems.invalid> wrote:
> m m wrote:
> Rick Jones wrote:
> >> unless one has a non-trivial latency in their 10GbE LAN,
> >> 1MB or so should be more than sufficient for "link-rate" TCP.
> >
> > Not everyone is moving data around a LAN.
> > I've had to go way higher than 1MB even with gig-e NICs
> >  on the local machine.
> > If you're pushing lots of data over an OC-3 WAN link
> >  with more than 100ms latency, you need those TCP buffers
> >  to be cranked way up to achieve high throughput.

> Cue Dave Taht for a lecture on BufferBloat?

Perhaps, though at 100 ms RTT, which I would consider non-trivial
latency :) (*) one does indeed need something like 1.8 MB of TCP
window to get link-rate for an OC-3.  Montgommery Scott may have the
ability to appear to change the laws of physics, but I don't think
anyone else can :)

I've just seen two or three too many 10GbE performance escallations
where the initial, knee-jerk reaction was to blow the TCP settings out
to 16 MB rather than actually consider the conditions.  My dimm
wetware institutional memory thinks it saw that first in some (older)
"tuning guides" from one or more 10GbE NIC vendors.

And while it may not be "directly" NTP related, certainly issues of
bloated queues is one that affects the quality of what NTP can
deliver.

rick jones

(*) assuming, of course that the 100ms wasn't queing delays and was
indeed speed-of-photon.

-- 
It is not a question of half full or empty - the glass has a leak.
The real question is "Can it be patched?"
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