- When we increased NUM_BUFFER_PARTITIONS to 1024, this problem is disappeared for 8 core machines and come back with 16 core machines on Amazon EC2. Would it be related with PostgreSQL locking mechanism?
- I tried this test with 4 core machines including my personel computer and some other instances on Amazon EC2, I didn't see this problem with 4 core machines. I started to see this problem in PostgreSQL when core count is 8 or more.
- Here are the results of "vmstat 1" while running 8 parallel select count(*). Normally I would expect zero idle time.
- When I run 8 parallel wc command or other scripts, they scale out as expected and they utilize all cpu. This leads me to think that problem is related with PostgreSQL instead of OS.