0

I've been trying to log some benchmark for Postgres XL using pgbench. I've used pgbench in a 78MB database with a json column. And to my knowledge, I cannot judge if my output result is okay to proceed with. I mean with what respect should I consider the results to be good or bad.

The structure of my main database table is as follows:

id: integer,
data: json,
created_at: timestamp without time zone,
updated_at: timestamp without time zone,
user_id: integer

Indexes: 

"index_datasets_on_user_id" btree (user_id)"

I've been receiving around 47 tps for this database table with scale factor 1, initial clients 5 and it decreases to 21 tps while the client number reaches up to 200 and scale factor remains constant. The pgbench test runs for 10 seconds each.

I have a system with 8GB RAM and Intel Core i5 CPU with 4 cores.

So it would be great to know if I can improve or if there's anything that'd let me know that I'm doing it the right way will be appreciated too.

Thanks!

Pragun
  • 1,123
  • 8
  • 20

1 Answers1

1

Check this url

http://pgtune.leopard.in.ua/

It can be a good start in order to performance your postgres database configuration. This only works fine in dedicated servers

In my own experience a conservative configuration will decresase the tps of your database despite you enlarge the shared_buffers or effective_cache.

Pablo Luna
  • 369
  • 3
  • 5
  • Be carefull with shmmax and shmall recomendations. Maibe your system update this automatically and it doesn't need to be changed. – Pablo Luna Nov 25 '16 at 11:51