[Sauce1971]

The fear of SQL, stored procedures, proper database constraints, the love of ORM and the embrace of nosql exists because a large portion of influental programmers just don't bother to understand relational databases and SQL. It's not in fashion, so it becomes a necesseary evil the primadonna rock star will try to avoid at any cost. Just like a real rock star might drop to learn to read notes or sing for that matter. Obviously DB theory should be regarded just as important as understanding the latest greatest language feature and framework.

[InclinedPlane]

I don't think that's true at all. I suspect that the most ardent noSQL advocates are actually considerably more versed in rdbms specifics than the average dev.

[gaius]

Not really. The NoSQL movement is based on the assumption that the limitations of MySQL apply to all databases. Their ideas of what an RDBMS is are 15-20 years out of date - really.

[jbooth]

It's more based on the assumption that the limitations of MySQL should apply to all databases.

I've worked 2 places with Oracle stored procedures. At both places, they were a giant nonperformant mess of spaghetti that could never be cleaned up because changing things == breaking things. And since all of the application logic had been written as transactional SQL with lots of joins over normalized tables, it was gridlock heaven inside the DB and adding more cores to the single machine barely even helped.

[gaius]

You can write bad code in any language!

[jbooth]

Yeah but that's a total nonpoint.

Are you arguing we should all use COBOL and every other stupid-ass approach, as well?

[InclinedPlane]

I think he's saying that anecdotes are not necessarily the most valid form of data for making decisions.

[InclinedPlane]

So you're saying that, for example, google uses bigtable because they don't understand rdbmses?

[FooBarWidget]

Part of NoSQL is about scalability using commodity hardware and without spending tons of money on licenses. Sure Oracle can scale, but people want those features for free.

[gaius]

Paraphrasing Zawinski, it depends what your time is worth. Remember in business, the concepts of "cheap" and "expensive" simply don't exist. A thing is "worth the money" or "not".

[greyfade]

I suspect that most "noSQL" advocates have a poor understanding of the difference between relational/structured data and unstructured/sparse data. Virtually all of the "noSQL" software is incapable of managing complex relational data, as far as I can tell. Why no one points this out is baffling to me.

"noSQL" is useful for for application layer - where caching and short-term storage are important; but most "noSQL" advocates I've seen seem to think that it should replace an RDBMS, when that couldn't be further from the truth. "noSQL" is useless as a structured backing store where RDBMS are used today, especially for large systems.

[FooBarWidget]

The problem I have with that statement is that you're dismissing everybody who have a legit reason to use NoSQL databases.

Show me an autosharding solution for an open source SQL database that doesn't involve paying tons of money. Until then I'm not convinced I should ditch MongoDB.

[tluyben2]

Yes, came here to say that as well. I hear people saying 'Oracle can scale as well'; that could be true, or not. I don't know: I don't have a lot of experience with Oracle on large scale projects (I did a lot of J2EE projects using Oracle, but those really could've just as well worked on SQLite, and I mean that... Complete overkill.). I know my database theory and I don't understand how Oracle can scale like a NoSQL database while preserving their RDBMS nature; as I understand that's not possible right?

So how is there no room for NoSQL databases? I understand that most sites running NoSQL at the moment could just as well (or even better) run on Postgresql, even with 'biggish data', but is for 'enormous' data and near real time analysis of it, NoSQL not a proper solution over 'traditional' RDBMSs. Not trolling; really curious what people have to say.

Do I understand it correctly that Oracle 'Exadata' just uses fiber cables and such to threat the entire rack as 'one node' or does it work differently? How does Oracle scale? Any information about that (tech information, not opinions/interpretations of reality as OP posted)?

[babebridou]

Why ditch anything?

Just go with both. MongoDB can't make your data live & evolve the way an RDBMS can. You can't inject intelligence in your noSQL database, but you can sure extract data, a lot of it, a ton of it, with as many concurrent clients you wish. If you can't find a use for a relational model in any project, then I'm not sure what to say. But if you can, maybe try a postgres & Mongo combo for the best of both worlds?