On Wed, Mar 18, 2020 at 06:52:49PM -0700, Tom Van Baak wrote: > "Facebook Switches to New Timekeeping Service" > https://spectrum.ieee.org/tech-talk/telecom/internet/facebook-new-time-keeping-service > > Unfortunately, I don't see a white paper, code, or technical details on > positive/negative smear duration or slew rate, so we don't know if at the ms > level it's the same as Google or Amazon or others on this bandwagon.
If I understand it correctly, they are using the time smoothing feature of chrony with 1ppb/s wander, following the example from the documentation. It's different from the linear smear implemented by Google/Amazon. It starts at midnight and it's a two-piece quadratic function. The duration is 2*sqrt(1e9) seconds (~17.57 hours). The last section of the following blog post has some graphs that show how the offset and frequency is changing during the leap smear. https://developers.redhat.com/blog/2015/06/01/five-different-ways-handle-leap-seconds-ntp/ -- Miroslav Lichvar _______________________________________________ LEAPSECS mailing list LEAPSECS@leapsecond.com https://pairlist6.pair.net/mailman/listinfo/leapsecs