No More Hustleporn: What is a Data Warehouse


Tweet by justin g

https://twitter.com/itunpredictable

streetwear analyst / latex salesman @retool, lowest ftp on the block. this is a parody account


If you saw Snowflake's S-1 and you're wondering what the hell they do, i've got a thread for you 🎶

Snowflake is a managed data warehouse. Data teams pay them to take care of storing analytics data so they don't have to deal with managing infrastructure

If you're wondering what a data warehouse is, i wrote about it here.

What’s a data warehouse?
Where warehouse workers means something else

it's a place to put your data that you want to query for analytics - e.g. longer running, bigger queries with more joins than say your normal app would require

A typical Snowflake use case (my old team at DigitalOcean used it for this)

- we had data in our production database

- i.e. the tables powering our app

- and we wanted to query them for analytics.

So we took snapshots of them and moved them over to Snowflake

Snowflake is relational, which means that you query it via SQL. They've also added some really nice SQL ergonomics for analytics, like the ability to pivot data (a notoriously missing SQL feature). But you can also query Snowflake via Python, R, Java, and lots of other fun stuff

Snowflake charges their customers for the amount of data they store in it, which makes it PaaS and not SaaS in the traditional sense. They're probably negotiating fixed cost deals with larger customers, which is where most of their revenue will come from longer term

Snowflake is also (mostly) enterprise focused. And I know this because of my patented Website Test (TM) - if you go to their site, read the headline copy, and have no f*cking clue what the product does, they're targeting enterprises

"mobilize your data" ok yea sure