PostgreSQL Scalability Analysis Deconstructed

[DrQ]

Latest blog post examining how to apply the USL scalability model to Postgres performance measurements.

" if you are going to use it, then it is really in your own interest to learn to use it correctly. It's not just a matter of throwing performance data at the USL model and seeing if the plots look like what you were expecting. If the USL always did what we expected, there would be little point in doing it at all. No new insights would be gleaned."

"In case you think that's a completely crazy projection coming from the USL, then check out what actually happened to the newer version of Postgres with the so-called fast locking modification. Those PG 9.2 throughput measurements, shown as green joined data points, reached an Xroof that is 5x higher than expected from the PG 9.1 data. "