Posts - Vik FearingRSS

Published: June 9, 2021, 5:45 p.m.

Continuing my monthly PostgreSQL benchmarks series, these latest findings are aimed at helping developers improve PostgreSQL performance stability. I discovered the performance stability issue when I had to reset the AWS instance I used for the Daily 500 benchmarks and lost 20% performance. The only explanation I have is that the reset instance physically moved in the data center. From an ongoing benchmarking perspective, this is unacceptable.

Published: April 27, 2021, 11:09 a.m.

This document provides an introduction to tuning PostgreSQL and EDB Postgres Advanced Server (EPAS), versions 10 through 13. The system used is the RHEL family of linux distributions, version 8. These are only general guidelines and actual tuning details will vary by workload, but they should provide a good starting point for the majority of deployments.

Published: April 15, 2021, 11:21 a.m.

As part of my work on the CTO Team here at EDB, I do daily benchmarking of the PostgreSQL development branch with 500 concurrent users in order to make sure performance does not go down and to identify when in case it does. This also helps us determine which features improve performance for this TPROC-C workload.

Published: March 22, 2021, 5:03 p.m.

*/
/*-->*/

Published: Feb. 18, 2021, 12:46 p.m.

Usually, benchmarks are measured in transactions per second, but the TPC-C and TPROC-C, benchmarks are measured in new orders per minute (NOPM). What is a new order? It’s simply a predefined operation on the database, that even fails 1% of the time. This is a better metric than transactions because you can have transactions on the database that are not part of the benchmark. Autoanalyze uses a transaction, for example.

Published: Oct. 6, 2020, 9:05 a.m.

PostgreSQL 13. 0 was just released and included a number of performance improvements in various areas. In order to see what kind of difference these changes made, I ran a number of benchmarks using HammerDB with different numbers of virtual users, comparing PostgreSQL 12. 4 and 13. 0 using both an un-tuned configuration and a 'starter' tuning configuration, i. e. one that we would suggest that users use as a starting point for their own deployments.