Hacker News Reader

Comment by carabiner | Hacker News Reader

I did AWS training at the Amazon offices in Seattle for data science. I was blown away by the configuration... I have recompiled linux kernels and configured iptables as a teenager, and this was an entire galaxy of more complexity. It took us 6 hours to the point where some of us had a Jupyter Notebook running. Many people didn't make it though.

View on HN

Were you doing things the hard way on purpose? It's literally one click to spin up a Jupyter notebook on GCP, arguably less if you use Colab (instant access to shared instances), and I'm sure AWS has a similar service.

https://cloud.google.com/vertex-ai-workbench

So, if I remember right, the endgame was to have an automated ML model running as a service on an AWS instance with SageMaker. Part of that involved having a Jupyter setup installed. Don't remember much past that, though.

Why?

That doesn't make any sense to me. What problems were you running into? Did you need to configure a bunch of super-non-standard stuff or something?

Artificial complexity to justify lock-in and artificial salaries.

The microsoft model of "if sysadmins are spending much more time debugging windows, then they put windows first on their CV and gradually management forgets linux exists"

Obviously! Just like devops as a whole. Why do we have this current mess instead of just pushing stuff to heroku?

Because heroku has limitations.

Oracle DBA’s would like to have a word with you in the back alley

Reply on news.ycombinator.com