GraphQL Database design patterns

Question

I am researching on the technologies available to build a social network website. The front-end portion has been decided, but I am still unsure about the back-end. The back-end will also be handled by an server running NodeJS. I was planning on using something like MongoDB, or even a combination of MongoDB and a relational database, but I do not know what the best practices are.

I stumbled upon an article that talked about using GraphQL with different types of database structures. The writer mentioned SQL, Redis and MongoDB for their user data, but I did not quite understand their sturcture, or if it would work for mine.

Basically I would like to have users to sign up to my page and add descriptions about themselves, have friends, join groups, upload media (pictures and videos), etc. A pretty basic social network setup, but ultimately the question narrows down to:

"What database structures are best for what?"

Should the SQL handle: userID, email, password
MongoDB: about, description, posts
Redis: cache for something like posts feed

And how would you handle media uploads? SQL, MongoDB? I have not found any useful articles that talk about using different database structures and most of the other StackOverflow questions feel outdated. They either favor one or another, but not a combination.

Answer 1

TL;DR : There aren't really any particular databases you "should use for GraphQL." Just use whatever is best suited for your data and requirements.

I'll give you a rough idea how we use it at work. We have an app that lets users upload songs and videos for us to distribute to digital streaming platforms, eg Spotify and such.

We use Postgres to handle all of the app's "transactional data." This is user account info, their song/video metadata, where to send the money, other stuff for the app, etc. All these data are a mere few-hundred MB.

Multimedia uploads go to AWS S3, and the location of each file is stored in Postgres.

The "analytical data," regarding how well each song did on each platform, eg number of streams and money made in each region, is stored in Google BigQuery. There is much, much more of these data and Google BigQuery is more equipped for this size.

Of course, the question of which data store one should use is highly case-dependent. GraphQL didn't really influence any of these decisions. The nice thing about GraphQL is that the resolvers let you pull data from anywhere and conglomerate it all into a single API response. By using GraphQL you should feel liberated to use a variety of databases, each well tailored to the type of data it handles.