> -----Original Message-----
> From: [EMAIL PROTECTED] [mailto:timekeepers-
> [EMAIL PROTECTED] On Behalf Of Koos van den Hout
> Sent: Tuesday, March 27, 2007 10:34 PM
> To: [email protected]
> Subject: Re: [time] deleted my server from the dns pool for the time being
>
> 10000 ntp queries per minute sounds like a lot. But when I calculate it to
> pure bytes traffic:
> 
> 10000 queries/minute -> 167 per second.
> 
> given the ntp packets I see (76 bytes) that would mean 12692 bytes payload
> traffic per second, which will take somewhere near 126 kbit (including
> some
> overhead) of those 512 kbit up.

My numbers are not calculated, but seen with /proc/net/dev and "ntpdc -c
iostats" received bytes (mrtg).
The peaks are steady at around 5900 packets/minute and outgoing traffic with
last peak was 18,2 kB/sec. 18,2*1024=18636,8 bytes, 5900/60=98,33
packets/sec, 18636,8/98,33=189,53 bytes/packet

These numbers are not exact and are read from graphs, but I think I can
conclude something is wrong with these numbers. It could be that my linux
box (Clarkconnect distro) is calculating something wrong or can't keep up
but I trust the output of /proc/net/dev.

Anyway, I have stopped monitoring scripts, because I see load peaks around
3,5  while tcpdump runs on port 123 and heavy processor use of tcpdump with
top.
My last output from monitoring scripts is:

Estimated active ntp pool clients:      1567
Estimated abusive ntp pool clients:       21
Estimated inactive ntp pool clients:   86911
Total ntp pool clients being tracked:  88478

I think 88000+ is quite impressive and caused by the Turkish provider. Has
someone already contacted them?

Jos van de Ven


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

Reply via email to