Re: PostgreSQL, OLAP, and Large Clusters

[Chris Travers <chris.travers@xxxxxxxxx>]

Obviously OLAP performance depends to some extent on what you want to do with it.   My recommendation is to follow Scott's advice first (keep things simple as long as you can), and then look at supplemental approaches when you reach the limit there.

The big limit IMO is the lack of intraquery parallelism.  This has implications when trying to run OLAP queries on terabytes worth of data and so at some point you may reach the limits of what a single instance can do.  At that point you can go with a proprietary vendor like Greenplum or you can look at PostgresXC (which is open source).  Keep in mind that solving this issue has significant complexity costs and so you are usually better off with tuning your existing instance first.

Best Wishes,

Chris Travers