2 ms·
> But in 2010's, internet datasets were growing much faster - and computing power wasn't. So Relational DBs weren't "fast enough" for these cases... hence "NoSQ
by aidenn0 6y ago
> But in 2010's, internet datasets were growing much faster - and computing power wasn't. So Relational DBs weren't "fast enough" for these cases... hence "NoSQL".
I joke that the only thing a NoSQL database can do faster than an RDBMS is give you the wrong answer to a query.
There are two different features that people typically think of when they think of NoSQL as compared to a traditional RDBMS:
1. Document database (i.e. unstructured, or at least weakly structured data)
2. Eventual consistency
You could have an ACID document database, or an eventually consistent relational database, so the two are actually orthogonal. It's definitely easier to get something up and running quickly if you aren't going to implement all of SQL though.
If you make money serving up ads on lists of documents, then getting any answer quickly is going to be better than getting the "right" answer slowly, so for google at least, this made sense, as google's databases actually were that big.
IMO the rest of it was people cargo-culting, not understanding the hard-won knowledge of the 70s, and also not realizing that, if your working set was 100GB, you could just buy a server with 256GB of ram and have zero performance issues. Prototypes went up fast, benchmarks were great, and then somewhere down the line, someone wanted to run a query on the data, and discovered that fast writes come at a significant cost.
- claytongulick 6y agoI don't know. After hating it a lot for a long time, I've come to really appreciate MongoDB's aggregation pipelines. They're very explicit and easy to control / optimize / understand. Yes, it sort of feels like you're writing an EXPLAIN plan, but I'm starting to actually enjoy that now that I've gotten deeper into the guts of it.
- fctorial 6y agoWith m.2 nvme, you don't even need a ton of ram for that. They're fast enough.
- strken 6y agoIf your database keeps growing, you'll eventually exceed the write capacity of a single server, and have to write to more than one place. You're correct that most of the NoSQL hype was cargo-culting, but I don't think it came solely from Google trying to serve ads quicker, I think it also came from companies having too much data to scale vertically.
- f6v 6y agoEvery time I heard “why not mongo?! NoSQL is fast!”, my pg instance barely used 50 gb and was nowhere near the scaling issues. Also, I’ve seen so many developers who never even heard term “eventual consistency” in my time interviewing, it’s terrifying.