2

I m partitioning my problem using a Domain-Driven Design approach. Several databases arise with low or null coupling.

In the long term they will be separate databases on separate hosts, horizontally scaled. However, currently I do not need/want such an expensive deployment. I just plan to run the whole thing on a single PostgreSQL cluster. The development environment is another clear situation where having a single cluster is desirable.

Authors as Chris Richardson recommend using schemas in the near term however, I have found this implies rewriting part of the SQL code when logical replication is used to solve the couplings. I imagine he is using the term schema with MySQL in mind. https://microservices.io/patterns/data/database-per-service.html

Using separate databases seem to be the perfect solution in terms of SQL code invariance but, does this imply a performance penalty on a sole cluster compared to a single database virtually segmented by schemas?

I have read this similar question https://stackoverflow.com/questions/1152405/is-it-better-to-use-multiple-databases-with-one-schema-each-or-one-database-wit/1157008#1157008 but mine is clearly focused on performance penalty. I am clearly biased to segment the problem using databases from a logical point of view.

Improve this question
2

2 Answers 2

Reset to default
3

If you don't need to join data from several databases, splitting the data into multiple databases is the cleaner solution, and there should be no performance disadvantages.

Improve this answer
2
  • Thank you Laurenz. I suppose that no extra PG cluster processes are launched for every database, right?
    – coterobarros
    Commented Jun 13, 2022 at 15:53
  • 1
    Right. They also share the same cache.
    – Laurenz Albe
    Commented Jun 13, 2022 at 19:51
1

Yes. The main problem is that you can't reuse connections. Which means that you will have more connection, which means that you will use more memory.

OTOH separate DBs means that each DB stats will be smaller.

Improve this answer
4
  • 1
    That need not be a problem. Each database will receive fewer requests, so you can use several smaller connection pools.
    – Laurenz Albe
    Commented Jun 13, 2022 at 15:44
  • Thank you. I'm ok with that (having more connections). I was expecting something more painful, as having a bunch of separate processes per database or something like that.
    – coterobarros
    Commented Jun 13, 2022 at 15:52
  • Sure. You will have separate processes per database. For each connection, as always.
    – eijeze
    Commented Jun 13, 2022 at 15:52
  • Yes, that is reasonable.
    – coterobarros
    Commented Jun 13, 2022 at 15:55

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.