You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on Mar 21, 2019. It is now read-only.
Basically within each node, monotonic time is possible because of the OS managed clock. But between nodes there's no guarantee. And also I find it difficult to actually localise monotonic time to unixtime. Maybe this is application dependent.
There will always be some level of desynchronisation, possibly even with PTP. We cannot really rely on PTP due to hardware requirements. Instead, applications must eventually use logical clocks.
This is important for log synchronisation, causality determination, and coordination-free lockless distributed algorithms.
The text was updated successfully, but these errors were encountered:
Time management is often combined with decentralised k-ordered ids that allow it to be possible to create unique IDs decentralised without any coordination (avoids centralised locks).
Sign up for freeto subscribe to this conversation on GitHub.
Already have an account?
Sign in.
Basically within each node, monotonic time is possible because of the OS managed clock. But between nodes there's no guarantee. And also I find it difficult to actually localise monotonic time to unixtime. Maybe this is application dependent.
There will always be some level of desynchronisation, possibly even with PTP. We cannot really rely on PTP due to hardware requirements. Instead, applications must eventually use logical clocks.
This is important for log synchronisation, causality determination, and coordination-free lockless distributed algorithms.
The text was updated successfully, but these errors were encountered: