[time] Accuracy goal of the pool

Tim Shoppa tshoppa
Tue Nov 14 12:52:53 UTC 2006


Arnold Schekkerman <ntp-list at mallos.xs4all.nl> wrote:
> On 13-11-06 13:18, Tim Shoppa wrote:
> > And making it even less useful, only the absolute
> > value of the measured (I say "measured" to distinguish
> > it from "true") offset is being stored and graphed.

> The usefulness of the monitor should be related to its primary goal:
drop
> "really bad" servers from the pool. It is doing that job well (I
guess).
> ...
> Let's first continue on evenly spreading the load(*) before taking
the road
> to a high precision pool again.

I am not trying to get on the high precision road again, I am trying
to get the server monitoring to not throw useful information away.

In fact I think it would be easier (one less function call to abs()
!) to track and plot the sign and magnitude of the offset rather than
just the magnitude.

Everywhere else in everybody's ntp tools they keep the
sign of the offset around rather than throwing it away,
so this is also a consistency thing.

Tim.



More information about the pool mailing list