Back from FOSDEM - the NoSQL trend

This week end, I went to FOSDEM. Even if it was a bit too geeky for my taste, it was certainly worth going. I attended quite a few talks and met the JBoss team, the play! framework team and JBUG.be people.

I was actually quite surprised by the number of people attending the NoSQL tracks. While I understand the enthusiasm this new trend brings, it seems I am not sharing much of it with the people I talk to.

What is NoSQL anyway? NoSQL promotes the use of very simple interfaces such as associative arrays or Key-Value pairs. It is also possible to largely distribute them and is highly scalable.

What really interested me about the NoSQL movement is the fact that solutions can be schema free. How is that relevant for web applications? I think it is a crucial point when you think about web applications, and the social aspects that web applications bring to users nowadays. Humans are not really structured and it is getting more and more complicated to enforce a structure that pretends to link users, services and technologies together.

However, to my surprise no one was interested by this aspect: it was all about performance! Let me tell you that this is the last aspect of this technology that I am interested by. I don’t think that any of the applications we develop require nanosecond response times and scaling to 100 million users. Besides, when you are talking about serving billions of terabytes of content, it is a completely different infrastructure that you are talking about anyway. So the performance aspect was quite irrelevant to me.

So, the bottom line is: should we really care about it?

I don’t see myself advising ditching relational databases just because of performance reasons or because the latest trend tells us that NoSQL is the future. Database performance is not yet a problem for us. What I was interested in, being schema or data structure agnostic, is not the main concern of the people who are developing these NoSQL solutions. However, the tooling that SQL databases have, and the relationships between entities are not really quite there, even if there is more and more development in those areas.

However, because this NoSQL movement is getting more and more popular, we might see object databases coming back as first-class citizens. Then I might get the best of both worlds: tooling, relational queries and a rather schema agnostic persistent storage.

In conclusion, I think this latest hype might bring new solutions that can be used for small projects helping us to develop quicker solutions, even if we are not quite there yet.