> Using 4 bytes and 100ms resolution your can fit in 13 years of
timestamps if you use the time you deploy the cassandra DB (aka 'now')
as epoch.
In our app we will be fine with this. 100ms is good enough.
we can probably do some garbage collection on timestamps like we do on
deletes. If timestamp is older then let say 14 days (read repair will
sync them) then during compaction it might be safe to turn it into
lowest value possible (like vacuum freeze in postgresql) and free time
space. If we will require let say one compaction per 2 years, we can fit
into 4 bytes much higher resolution then 100ms.
I also think that cassandra has too high row/column overhead. I hope
that vector clocks will never happen.
