Re: [Hampshire] ntpd vs. ptpd

Top Page

Reply to this message
Author: Bob Dunlop
Date:  
To: hampshire
Subject: Re: [Hampshire] ntpd vs. ptpd
Hi,

On Tue, Jan 24 at 06:15, James Courtier-Dutton wrote:
> Does anyone have experience of both ntp and ptp ?


Use NTP all the time and am planning to move to PTP for some deployments
in the near future. What are your actual requirements ? As we have to
keep telling our customers. "Tell us what you want to do, not how you
think it can be done."


NTP on a quiet LAN is good for ~100us maybe down into the tens of us. If
WAN links are involved then think milliseconds, might be better to just
sync each station to GPS.

Software PTP won't give you much better running on non realtime systems
such as linux. With software PTP just about any decent switch that
supports multi-cast will be fine.

Hardware PTP where the NIC does timestamping of the packets at capture or
transmission is better, you should easily be able to go sub microsecond but
have you thought about other factors ? PC processor clocks are rubbish,
you won't be able to maintain a stable reference without something like a
temperate compensated crystal. Also beyond two or three systems on a LAN
this is when you'll need proper PTP support in network switches etc, what
that really means is they run PTP internally and act a secondary clocks
for the attached devices. Back to the start, why would you need this
accuracy ?


Our customers are getting us to push the boundry on syncronisation of data
timestamps between stations. Next generation systems will have TXCO for
internal clocks, hardware PTP support in the NICs etc. Precision costs.

-- 
        Bob Dunlop


--
Please post to: Hampshire@???
Web Interface: https://mailman.lug.org.uk/mailman/listinfo/hampshire
LUG URL: http://www.hantslug.org.uk
--------------------------------------------------------------