[Pool] Virtualization

Andreas Krüger timekeeper at famsik.de
Sat Jul 18 14:53:20 UTC 2015

Hi, Florian,

for quite a few years now, I've been running one of the pool servers
on what I have, which is a KVM-based virtual root server.

My server provider has never mentioned anything that they have a
problem with my running ntpd.  Nor do I think a decent provider should
have a problem with it.  I pay for the bandwidth.  Whether I use it for
HTTP/TCP or NTP/UDP or whatever I consider my business, not theirs.

The NTP monitoring system rates my server at full 20 points, and the
offset displayed is within ± 5 ms mostly.

I use a combination of my own and several other pool servers for ntp
clients I run.  I occasionally look at what those clients think about
my server.  It puts up with other pool servers quite well.  It gets
elected as the main synchronisation source sometimes.  So I'm not
worried my service has any quality problem.

Here is a dmpeers output from an NTP client at home from right now,
with the server names replaced for anonymity:

     remote           local      st poll reach  delay   offset    disp
.A        3 1024  377 0.00209  0.000679 0.13852
*B        2 1024  377 0.03355  0.000705 0.13751
 C        3 1024  377 0.02814  0.002601 0.13800
.D        3 1024  177 0.03513  0.000968 0.12502
.E        2 1024  377 0.02754  0.000380 0.12416
.F        2 1024  377 0.02402  0.001057 0.13824

"A" is the my local DSL router, all the others are pool servers.
Which of those, would you guess, is my virtual server?

So, in my opinion, my virtual server gives good service to the many
clients it keeps providing time to.  I'm happy with that, and leave it
as it is.

Theoretical remarks of the "you shouldn't do this" sort I occasionally
answer with mails like this one, and otherwise ignore.

If someone gave me a test to run on client side that demonstrates that
my server actually does perform worse than other pool servers, I might



P.S.: For the record: Mine is not server "C" on the list.

Am 18.07.2015 um 12:36 schrieb Florian Weimer:
> The support.ntp.org site says this about virtualization:
> “NTP server was not designed to run inside of a virtual machine. It
> requires a high resolution system clock, with response times to clock
> interrupts that are serviced with a high level of accuracy.”
> <http://support.ntp.org/bin/view/Support/KnownOsIssues#Section_9.2.2.>
> However, following discussions on this list, there appear to be
> pool.ntp.org instances on Amazon EC2.
> Do the restrictions on virtualization still apply?  I'm pretty sure a
> properly configured KVM host and guest (for example) today has a much
> better hardware clock than most things that were available in the 90s,
> and people were running time servers even back then (even on PCs).
> _______________________________________________
> pool mailing list
> pool at lists.ntp.org
> http://lists.ntp.org/listinfo/pool

More information about the pool mailing list