[ntp:questions] mtp srver load figures?

unruh unruh at wormhole.physics.ubc.ca
Thu Dec 3 20:03:48 UTC 2009

On 2009-12-03, Christoph Weber-Fahr <wefa2 at gmx.de> wrote:
> Hello,
> Unruh wrote:
>> Christoph Weber-Fahr <wefa2 at gmx.de> writes:
>>> I'm preparing to build a few ntp servers for a somewhat large population of
>>> NTP-misbehaving boxes (they top out at a 120 sec interval).
>> Why have you set them up that way?
> Basically, that's the vendor-specified behavior.

OK, I understand. Now very friendly behaviour and you might want to let
the vendor know that this kind of behaviour is considered very impolite
and disruptive by the public ntp community ( and makes me wonder as to
the quality of the software that they actually wrote), but I understand
you can do nothing about it. 
So it is crucial that these boxes use a private source for their time
signals, and not go out onto the net. Your private server should
certainly be able to handle a few thousand. Each request is only about
a few hundred bytes, flying back and forth. (something like 60 out and
60 back) so it should not be a problem for the net (unless you are using
telephone modems) or for the machine.

>>> So, to size this right - what kind of request load can I safely throw
>>> at ntpd on a current Xeon server?
>>> Is 200 requests per second a problem?
>>> What do you calculate?
>> That is not very many. But why? I am sure you do not have 20000
>> machines. 
> Actually that's exactly the design target.
> And that's not that big.  Somewhere further down on my priority list
> is doing the same for about a million boxes already out in the
> field (though these behave marginally nicer).

Ooo. I do hope they do not go out onto the net to get their time! That
would really really be very impolite ( to phrase it mildly). 

More information about the questions mailing list