PostgreSQL Performance
From Wiki
We cover external postgresql perf reports, as well as own run perf reports. The internal ones are both simple tests on a Intel Core2 Quad Machine w/ 4GB RAM as well as tests run on a NUMA Architecture, 16 Core machine (details to come)
Contents |
[edit]
Test methodology steps
- Vacuum full all databases on which we perform tests PostgreSQL_Configuration#PostgreSQL_vacuum
- Stop postgresql (pg_ctl stop -D .)
- Start postgesql (pg_ctl start -D .)
- Warm-up the disk cache using brute force and luck (we don't select the files, we just pass everything through the disk cache) Code_Snippets#Warm-up_PostgreSQL_before_querying.
- Run tests
[edit]
Their results
- Scaling performance of PostgreSQL vs. Database Size [1]
- Increasing load size they measure tx/sec (tps) and give the results.
- As expected we notice the 3-areas of performane: disk cache bound (very fast), disk io bound (quite slow) and disk seek bound (even slower)
[edit]
My results on SMP Architecture
[edit]
