On Fri, 28 Sep 2007, Patrick Domack wrote:

> I would venture to guess most users use ntpdate instead of ntpd so it
> would relookup every time.
>

I had a server in the ca pool set at the lowest "net speed". I had little 
spikes when my IP was returned by the pool DNS but nothing to worry 
about really, my IP took turns in the North-America pool as well.

The interesting point is that when I removed the server from the pool it 
took at least one week before I could see traffic slowly going down. More 
than a month later I still see 0,5 to 1,5 kbits/s trafic.

This tends to show that an important percentage of ntpd servers use the 
pool indeed. (or other caching client, like VOIP devices, home-routers 
might be). Regarding ntpdate, customers using it will obviously only hit 
you while your IP is returned by the pool DNS (unless they query some 
kind of broken caching-DNS that do not respect time-to-live, windows DNS 
helper maybe..? ;-).

As Ask (I think) mentioned, taking your server out of the pool is a nice 
experiment to figure out wheather you get hit by clients caching your IP 
or not.

I assume a server at the highest "net speed" will get more ntpdate 
style requests (non caching client) than a server with the lowest "net 
speed".

Overtime, I also assume a pool server with the lowest "net speed" settings 
will manage to build up quite a long list of ntpd style (caching) client 
customers over time. It will get less spikes from ntpdate style client 
although since its IP is returned less often by the pool DNS.

So as Nelson (I think) reported, it would be normal not to see a big 
difference when changing your "net speed" when you have been in the pool for a 
while. Especially if downgrading to a lowest speed, at first you would 
just have less spikes.

-ls


_______________________________________________
timekeepers mailing list
[email protected]
https://fortytwo.ch/mailman/cgi-bin/listinfo/timekeepers

Reply via email to