One of the classic examples is leader failover: when is it safe for a replica to start serving? We only have to wait to be sure the original leader had realised it’s stopped getting heartbeats. So theoretically a tighter clock sync makes that wait time shorter. But...
-
-
Show this thread
-
my guess is that other factors, like scheduling delays, dominate in this case and contribute to potentially significant variance before a leader realizes what’s up. Putting the clock check on the critical path mitigates this at the expense of lots of clock checks.
Show this thread
End of conversation
New conversation -
-
-
henry is this a rhetorical question or...? do I need to read the paper right now?
-
I mean Heidi thinks she might be out of a job, so calibrate your threat level accordingly:https://twitter.com/heidiann360/status/1219629171590287360?s=20 …
- 7 more replies
New conversation -
-
-
It’s a very nice paper that advances the tightness that can be achieved with clock synchronization. Don’t think by itself it makes agreement and durability obsolete.
-
I agree in the strict sense, but I wonder how close we are to practically being able to appeal to probabilistic arguments about the chances of A and B having observed the same instant.
End of conversation
New conversation -
-
-
unroll plz
@threadreaderapp -
Bonjour, the unroll you asked for: Thread by
@HenryR: This paper that improves clock synchronization accuracy to tens of nanoseconds is amazing work. What… https://threadreaderapp.com/thread/1220166015838900224.html …. Talk to you soon.
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.