Replies: 2 comments 1 reply
-
Adding a layer in your architecture will add an overhead, so you need to consider if the functionality of that layer is needed. You only need to set Do you have a specific use-case in mind that you are testing for ? Or a specific feature of a connection pool deployment ? |
Beta Was this translation helpful? Give feedback.
-
I was under the impression that creating new connections to PG has significant overheads. The above was a part of a benchmark I was working on comparing a few PG poolers and comparing their differences. I'm beginning to understand that the use case for poolers is in much more complex setups. |
Beta Was this translation helpful? Give feedback.
-
I've run the following against straight PG (TPS: 389), pgbouncer (TPS: 305) and pgagroal (TPS: 278).
I would expect a higher TPS from a pooler but straight PG beats pgagroal out. I set up the initial db with
pgbench -i intel
. Am I missing something?This is my pgagroal config:
Beta Was this translation helpful? Give feedback.
All reactions