Neo4j vs Mongo vs SQL

Which one would you choose for a high density scalable database?

Please comment with your ideas about speed and scalability.

I think you mean mysql

@RazorSh4rk Why? Is oracle's SQL server not even worth mentioning?

A graph db, document db and relational db ?, I mean wouldn't it depend more on the application ?

A lot more parameters if you ask me. What will the database contain? How many users? How many estimated queries per user? Location? Centralisation? To name just a few @ksdme

Take DevRant for example, if you created it what kinda database would you use.

I would say neo4j if you have enough numbers of relations or you want to find patterns in the data (neo4j user myself)

MongoDb if you dont know the model yet
Or the model you need to represent doesnt need special contraints and restrictions.

MySQL well, I would go with Postgres but if everything is well defined and you need integrity and contraints

Again, IMO

Any ideas on speed? Obviously in-memory databases are the best but for traditional databases which is the way to go for low latency queries?

Elasticsearch all the way!

Postgresql , I have already explained in a long comment why and bored to re-explain.. however take a look to it, performances are awesome!

Postgresql , I have already explained in a long comment why and bored to re-explain.. however take a look to it, performances are awesome!

@omomthings thanks! I'll find your post and read it. We're using postgresql now actually, I was just wondering if there's something even faster than that. Cos we don't really need sql

@vertti I didn't really worked with nosql, however I made some researchs and to make it short: if you have linked data go for sql, otherwise you can go on whatever you want.
However indexed sql data is faster than nosql but is in a fixed schema. Nosql lacks speed since it's not structured but it gives you more possibilities if you're data is not obeing to a schematical structure but a logical structure..

@omomthings ok got it! Looks like our current setup of postgres + mnesia as an in-memory db is perhaps the best alternative then. Plus postgres is very mature and has a huge set of functionality for business analytics purposes etc.

Like a few people have said it really depends on the type of application and requirements.

The structure of devRant lends really well to a graph database so that's why we use Neo4j.

@dfox yes I agree, each database is good for a different purpose. Your choice is certainly not a conventional one, it looks interesting!

I'd like to experiment with kdb+ next