[ntp:questions] Re: NTP clients not throttling back is this behaviour RFC compliant?

Maarten Wiltink maarten at kittensandcats.net
Wed Dec 3 23:20:39 UTC 2003


"Terje Mathisen" <terje.mathisen at hda.hydro.com> wrote in message
news:bql3gq$lj5$1 at osl016lin.hda.hydro.com...
[...]
> If I was worried about the capability to handle a lot of NTP request,
> (i.e. I have about 30K+ machines here inside the corporate LAN) I'd make
> such a fast path inside the FreeBSD kernel, and also timestamp all
> interrupts.

If I were ever to worry about such a thing, I would certainly not start
by hacking a single time server.

Thirty-thousand clients, let me think... three stratum 1 servers, thirty
peered stratum 2 servers, a thousand leaf nodes per S2 (maybe two or
three thousand when they poll more than one server). Polling once per
1024 seconds; no, I don't see the need for any fast path. Starting up
30,000 clients at once may induce additional considerations, though.

Broadcast mode doesn't seem to scale as well as it might have, if recent
posts are anything to go by. It would have made a nice alternative
otherwise.

Groetjes,
Maarten Wiltink





More information about the questions mailing list