It's all about the price. The database is largely irrelevant.

I have been involved in a dozen or so Hadoop deployments now and almost all of them are there not because it will allow them to do amazing new things but because the licenses for Teradata or Oracle were too expensive. So they have a giant "data lake" aka "dumping ground" on HDFS and keep just the mission critical data in the EDW.

Out of curiosity, did all of those clients use CDH (Cloudera Hadoop)? They're pretty much becoming a hegemony in their own right like DataStax and Cassandra

We run both CDH and Cassandra at my employer. DataStax made a play, but they were Oracle level expensive, so we stuck with OSS.

CDH provides benefits in the early days, but at this point it feels more constraining. Their release cadence is slow, so we're stuck writing long term workarounds for bugs that we're fixed two years ago in the upstream project. The stuff that they are working on is half-baked and doesn't support what would seem to be trivial use cases.

I wouldn't be surprised if their ecosystem calcifies now that they are public.

We had a similar experience and walked away very early on. Now run >1k node clusters without outside consulting.

Bit late to reply but it was a mix of the big two. Cloudera and Hortonworks.

And I wouldn't say Cloudera is dominating Hadoop.

Without proper planning, most data lakes end up as the digital equivalent of a super fund site.

We straight up call ours the "data swamp" because no upfront effort was put into governance.

Lots of RBDMS's have their issues or drawbacks, none like Oracle DB. The only RDBMS where you actually have to throw cash at it to solve a problem like scale or backups.

The Oracle RDBMS is absolutely terrible. You'd have to be absolutely mad to use it, even if it was free. In all my years I have never met anyone ever that chose it for it's technical merit, especially from people whom have knowledge of more than one RDBMS. Whatever it's got, another RDBMS's do it properly, and (usually) for a lot less money.

Pick a feature, any feature they advertise and will most likely be from a product they acquired, or half-arsed themselves.

I am not at all convinced any oracle engineer knows about fsync().

EDIT: PL/SQL is not a feature. It's a punishment.

Its behavior towards open-source software has been chaotic, sometimes birthing great projects (eg. btrfs, which wasn't even a great fit for their database systems), sometimes strangling them slowly (eg. OpenOffice). Overall, its relationship with external contributors has been awkward.

Not sure what that has to do with anything.

Companies aren't interested in how altruistic their vendor is. They care if their product works and is supported. Also many of Oracle's customers aren't exactly strangers to aggressive tactics in the marketplace.

"...shit mediocrity, inflict misery, lie our asses off, screw our customers and make a whole shitload of money..."[0]

0 - https://www.youtube.com/watch?v=-zRN7XLCRhc&t=2147

And also send letter to FCC to encourage repeal Net Neutrality, try to make use of API copyrightable, try to steal Android revenues from Google with debatable reasons, and so ones... That is for the company in general, but for the customers of their database, there is also much to say.

I meant the database, not the company.

By using the DB you have to deal with the company.

By giving a grant to a company that is (apparently) more ethical will signal what kind of behavior you encourage, hence choosing to work with MariaDB over OracleDB.

Well, OK, but e.g. REFRESH MATERIALIZED VIEW ON COMMIT might be more valuable to you than working with a more ethical company. YMMV, of course.