[ntp:questions] Why does ntp keep changing my conf file?
unruh at wormhole.physics.ubc.ca
Mon Sep 20 19:34:57 UTC 2010
On 2010-09-20, Daniel Havey <dhavey at yahoo.com> wrote:
> Reject it? Why would I do that? I'm just makeing sure of stuff before I decide what to do. Besides, sometimes it is better to learn a little bit before I just blindly follow the advice of the old school ;^)
>> I would have guessed even 802.11a or g AP would be able to
> reach a couple thousand packets per second, which could
> result in sub-millisecond packet timing, and 802.11n AP
> perhaps few hundred thousand packets per second resulting
> in a few microseconds packet timing?
> Soooo, at 54Mbps sending a 1500 byte packet would take 1500 * 8 / 54M which gives me 0.2222 milliseconds per packet. Hmmmm, perhaps there is an argument for a PPS source here. If we are off by 1 ms then we could lose 4 or 5 packets to clock skew.
>> ...but latency? On my control plane? No way ;^) That lan is soooo
>> lightly loaded that any packet can get sent anywhere it wants at any
>> time on Gigabit ethernet.
>> With Gigabit you may get erratic latency due to buffering and interrupt coalescing.
> Just what do you mean by erratic? This is a modern networking device, it's going send the packet right away. Sure a few interrupts will coalese but it's not gonna get recieve livelock or anything like that. The GigE card is gonna get the packet out the door right away, I think any jitter will be small.
Experiment says you are wrong. I used to have my machines on regular
ethernet (110MB) and the time delay was 140usec almost withoug fail.
With a GB switch and GB cards, it is all o\ver the place, with a bimodal
( 140 and about 300usec roundtrip ) up to 10ms round trip. It is really
really horrible for timing purposes.
> --- On Fri, 9/17/10, John Hasler <jhasler at newsguy.com> wrote:
>> From: John Hasler <jhasler at newsguy.com>
>> Subject: Re: [ntp:questions] Why does ntp keep changing my conf file?
>> To: questions at lists.ntp.org
>> Date: Friday, September 17, 2010, 5:33 PM
>> Daniel Havey writes:
>> > ...but latency?? On my control plane?? No
>> way ;^) That lan is soooo
>> > lightly loaded that any packet can get sent anywhere
>> it wants at any
>> > time on Gigabit ethernet.
>> With Gigabit you may get erratic latency due to buffering
>> and interrupt
>> > That might be a little extreme, we don't have a PPS
>> source, or roof
>> > access for a GPS receiver.
>> Since all you want is synchronization among your machines
>> your PPS could
>> free run.? You should be able to build one.? It's
>> trivial: just a stable
>> one Hertz pulse generator.
>> > If we can get down to a millisecond then we are
>> Then you don't need a PPS source.
>> John Hasler
>> jhasler at newsguy.com
>> Dancing Horse Hill
>> Elmwood, WI USA
>> questions mailing list
>> questions at lists.ntp.org
More information about the questions