News for hackers. Stuff that matters.
Re:The Technology Behind Convore by foobarbazetc on Wednesday February 16, 2011 @08:28PM (#2229243)
Just FYI: we run a very large web service with purely PostgreSQL, and have no need for sharding or anything of the sort.

Scale up first. Use redis or memcached to store 'overview' stats. If you lose those stats, recalc from Pg.

Re:The Technology Behind Convore by ericflo on Wednesday February 16, 2011 @08:32PM (#2229256)
That's exactly the approach we're taking. We're not going to spend any time prematurely worrying about various scaling things until we can forecast that it's legitimately going to be a problem.
Re:The Technology Behind Convore by foobarbazetc on Wednesday February 16, 2011 @08:45PM (#2229295)
Great to hear. :)

While we're here, here are some PG tips:

1. Make sure your PG install is fully optimized. I've seen countless postgresql.conf's at major companies with no opts what so ever. Take a look at this book: http://www.2ndquadrant.com/books/postgresql-9-0-high-perform... .

2. Make sure you're using PG 9+.

3. Use pgbouncer for connection pooling and try to write your app so you can use 'pool_mode = transaction'.

4. Please don't run PG virtualized.

Re:The Technology Behind Convore by ericflo on Wednesday February 16, 2011 @09:07PM (#2229355)
We're good on all fronts except 4, and the transaction pool mode :)