On Thu, Jun 11, 2015 at 2:05 PM, Josh Berkus <josh@agliodbs.com> wrote:
> * Faster sorts: sorts on large amounts of text data are up to 5X faster.
The same goes for numeric, BTW (perhaps more so), thanks to Andrew
Gierth's work on numeric abbreviation. Also, while a 3X - 5X
improvement is common, there are rarer cases that are significantly
better still. I've seen an independent report of a 12X improvement on
a large server with a big sort that spilled to disk. The actual
operation was a CREATE INDEX on text. This was reported by pgConf.US
speaker Corey Huinker, who happened to have this problematic case that
he was generally concerned about. We talked about this over lunch in
NYC, and I advised him to check out 9.5. Apparently he is looking
forward to being able to do this in production, because such CREATE
INDEX operations are common in his data warehousing app.
--
Regards,
Peter Geoghegan