Ask HN: Which is your go to Database to use in a project?
Which is the database that you will use when thing to any new projects or recommend when working for a client
14 comments
Depends entirely on the type of project and the data.
Data that models easily via sql? Percona Xtradb Cluster works pretty well.
Data that models better as a hierarchy of objects? OpenLDAP is pretty amazing.
But a project using either of those is likely to also use eg redis+qless for processing out of band tasks.
If it’s a utility that needs to store some data it might use SQLite.
My point is that understanding how various options can be used is likely to give a better result than just having a “goto” square peg and hammering it into every hole you find, regardless of shape.
Data that models easily via sql? Percona Xtradb Cluster works pretty well.
Data that models better as a hierarchy of objects? OpenLDAP is pretty amazing.
But a project using either of those is likely to also use eg redis+qless for processing out of band tasks.
If it’s a utility that needs to store some data it might use SQLite.
My point is that understanding how various options can be used is likely to give a better result than just having a “goto” square peg and hammering it into every hole you find, regardless of shape.
SQL. Doesn’t really matter what form, but it’s just easy to work with. We tend to mainly operate MSSQL, but we’re not so in bed with T-SQL that its hard for us to operate the few MySQL and Postgres databases we have.
SQL is easy to build on and maintain. Systems operators know how to run it both locally, on rented iron and in the cloud, and developers know how to utilise it. At least around here.
It frankly does everything we need as well, from being a database to being a queue and so on, and while there are other ways to do much of that, we’re not Netflix. Having the perfect tech stack is more expensive for us than not having it, because it requires our staff to know more things, and you can buy a loooooot of hours in Azure for the price of an additional developer/operations engineer.
SQL is easy to build on and maintain. Systems operators know how to run it both locally, on rented iron and in the cloud, and developers know how to utilise it. At least around here.
It frankly does everything we need as well, from being a database to being a queue and so on, and while there are other ways to do much of that, we’re not Netflix. Having the perfect tech stack is more expensive for us than not having it, because it requires our staff to know more things, and you can buy a loooooot of hours in Azure for the price of an additional developer/operations engineer.
Do you self host it or use a hosted service? Also why not any other NoSQL or SQL based DB’s
Mysql/Mariadb because it's fast, easy to setup and featured enough for all my needs. Also because I'm used to it and storage isn't my core business, I don't need to change.
Postgres. It does everything I need on most projects, it's fast, I can whack it in a Docker container to run it, and it's very well supported.
Postgres managed on RDS for projects that make money
I avoid a db or use dynamodb for all other toy projects
I avoid a db or use dynamodb for all other toy projects
PostgreSQL with Hasura answer 200% of our needs.
Hasura exposes a GraphQL API and translates GraphQL queries to SQL queries (that's how it solves the N+1 resolver problem).
It handles authorization and permissions using the PostgreSQL user system.
It manages triggers and so much more. You can extend the functionalities with Webhooks and custom remote GraphQL schema.
In shorter terms: Everything we need and so much more.
Hasura exposes a GraphQL API and translates GraphQL queries to SQL queries (that's how it solves the N+1 resolver problem).
It handles authorization and permissions using the PostgreSQL user system.
It manages triggers and so much more. You can extend the functionalities with Webhooks and custom remote GraphQL schema.
In shorter terms: Everything we need and so much more.
Postgres
Postgres and redis
PostgreSQL
Sqlite.