Veem
You should also be familiar with Aurora Postgres's storage architecture, which is very different from regular Postgres (see https://docs.aws.amazon.com/AmazonRDS/latest/AuroraUserGuide/Aurora.Overview.html)
Aurora is remote storage, which means if your read workload can't fit into the PG's shared buffers, it will have a very different performance than if the storage is a local SSD.
On write, it writes six copies to three different availability zones for high durability and availablity. So having enough network bandwidth is a factor as well.
Ken
On Tue, Dec 26, 2023 at 11:30 PM Kirk Wolak <wolakk@xxxxxxxxx> wrote:
On Thu, Dec 21, 2023 at 8:31 AM veem v <veema0000@xxxxxxxxx> wrote:Can someone please guide me, if any standard scripting is available for doing such read/write performance test? Or point me to any available docs?
...Veem, first things first... "Top Posting" is when you reply at the top of the email... Notice how I replied at the bottom (and I deleted context, clearly).
This is the style we prefer here.Second, since you are new to postgreSQL... Let me recommend some reading. Cybertec has articles on performance (Tom Kyte style).
Also, read the "Don't Do That" wiki, and finally, have a look at pgbench and psql documentation. And specifically look at GENERATE_SERIES(),
but the Cybertec articles will touch on that. Regardless... Reading the docs is insightful.
Links:
https://www.cybertec-postgresql.com/en/postgresql-hash-index-performance/
https://wiki.postgresql.org/wiki/Don't_Do_This
https://www.postgresql.org/docs/current/app-psql.html
https://www.postgresql.org/docs/current/pgbench.html
https://www.postgresql.org/docs/16/functions-srf.html
HTH,
Kirk Out!
