Overcoming Performance Hurdles in Postgres Partitioned Tables

In this talk, you'll learn about 2 prevalent performance issues encountered with partitioned tables in Postgres. The learnings shared in this talk are based on my experience working with diverse customer workloads in Azure, but these issues are relevant for anyone using Postgres partitioned tables, particularly those of you who manage vast data volumes or high traffic.

High Query Planning Time on Large Partitioned Tables: High query planning time on tables with very large numbers of partitions (5000+), where the optimizer is not able to identify the partition key till the execution time leads to planning time greater than execution time.

LW Lock Manager Wait Events: In highly concurrent environments, queries that don't use the partition key can trigger LW Lock Manager wait events, leading to higher CPU utilization (sometimes as much as 100%) and lower TPS.