Fall of NOSQL?

Is your shop still using it? Most NoSQL systems I've seen in the wild are pretty much just used as a caching store .. fancy Reddis really. All the big shops I've worked at still use Postgres for persistant storage. If they need speed; they'll add Elastic Search or Spark and have indexing services to make sure they stay up to date.

I think there are probably quite a lot of production Mongo databases out there from companies founded around the peak of the (bootcamp-driven) NoSQL hype train.

No one's bragging about it, probably everyone regrets it, and most companies will never practically be able to get off it (because, when your data is an unstructured mess it's a huge project to try and figure out how to get to structure, not to mention maintaining consistency during the migration. And you have that home-grown "relational" layer someone wrote for Mongo when they realised that they really did have relational data after all, and which is now coupled to everything...)

(If you thought I was describing the primary database at my day job, well...)

Most companies that adopted MongoDB at peak NoSQL either failed or were forced to switch to a better option (NoSQL or otherwise): that was a steaming pile of crap that couldn’t be trusted with data. The people behind it were concerned with marketing and going viral and succeeded at just that, it’s only recently that the money and time has been put into making the product itself actually any good.

Then perhaps we really are an outlier. That'd be nice: fewer people than I thought toiling under the yoke of a crappy database.

The post-IPO success of Mongodb Inc, and Amazon's recent roll out of their own compatible hosted solution makes me suspect not, though.

Perhaps I was a little too optimistic ;)

It’s not sexy but we use both Azure table storage and Amazon Redshift and they are nosql in my book.

Redshift is a columnar MPP fork of postgres, definitely not NoSQL. Do you mean DynamoDB?

I feel like many experienced developers I know have gone back to defaulting to SQL databases again, because it turns out the trade-offs of NoSQL (schema-free, horizontally-scaled-but-eventually-consistent etc) weren't worth the trade-offs.