Thanks to everyone who replied. All helpful. I learned and have new ideas to work with.
> On Aug 11, 2020, at 10:42 AM, Adam Brusselback <adambrusselb...@gmail.com> > wrote: > > I mentioned this in another email thread yesterday about a similar topic, but > I'd highly suggest if you do go the UUID route, do not use the standard UUID > generation functions, they all suck for database use (v1 also sucks). > > I use: https://pgxn.org/dist/sequential_uuids/ > <https://pgxn.org/dist/sequential_uuids/> written by Thomas Vondara (writeup > here: https://2ndquadrant.com/en/blog/sequential-uuid-generators/ > <https://2ndquadrant.com/en/blog/sequential-uuid-generators/> ) > > I don't mind having a time component correlated with my UUID's because it's > simply not a threat model that matters for my use case, so I use the time > based variant. It helped me immensely with FPW and > write amplification when I switched from V4 UUIDs. It is still not as fast as > an int, but it is much much better than random UUIDs. > >
