Jeff Janes wrote
> You could periodically merge older partitions into larger tables, index
> those aggregated tables, then transactionally disinherit the old
> partitions
> and inherit the new aggregated one. This would keep the value of K down,
> at the expense of re-writing data multiple times (but all method write
> data
> multiple times, some just hide it from you).
Forgot to add:
I thought also that we could:
- use the RAM as tablespace for indexes, and move the indexes later (but
postgresql doesn't handle very well a machine crash in this case... it would
be cool to create an index as "recreate on crash"...)
- use unlogged tables and turn those to logged to speed up somehow the
insertion; I actually started to write a patch for it, but making it work
for replication was too hard (not that I'm using replication, but it
wouldn't be accepted for "wal_level = minimal"). But this wouldn't solve the
problem anyway.
--
View this message in context:
http://postgresql.1045698.n5.nabble.com/Fast-insertion-indexes-why-no-developments-tp5776227p5776418.html
Sent from the PostgreSQL - hackers mailing list archive at Nabble.com.