DISQUS

Vitor Domingos · 1 year ago

Lesson to learn from this: you can actually be crappy on your software, database schemas/models and development. Hardware is cheap. That will be the *real problem* for some in a not so distant future.

I, for once, like to target clean code that can, by itself, scale.

Mike Kavis · 1 year ago

I agree with your database use case but only for relational databases. For massive datawarehouses in the billions or trillions of rows, a massively parallel and distributed database is required (ie. Map Reduce). These types of databases perform better on smaller nodes and many of them. Big iron just isn't as feasible for this type of problem. I have read several articles today that have said that databases are better off not being in the cloud. I think we should clarify that this is only true for traditional DBMS's only (which are gradually falling out of favor for non OLTP processing).

My 2 cents!

randybias · 1 year ago

@Vitor Domingos

Agreed.

@Mike Kavis

Yes, definitely only relevant for relational databases and agree with your assessment about a shift of some RDBMS use to distributed database systems. I think that just reinforces my point, though. Relational DBs will have a place for quite a while, regardless of certain loads moving to distributed database-like systems (Hadoop, etc.). That just means scale-up will take what pieces you need in your RDBMS even further.

Thanks!