Slow Azure Machine Learning

Playing with Azure Machine Learning (AML) and it is really annoying:

1. It’s damn slow — you submit an experiment (at 18:30), and your compute cluster begins to scale only at 18:34 and finishes at 18:37

2. UI is bloated — same old MSFT, as it was 20 years ago

3. but completely uncaring at some places — you select to delete the selected resource group, and you are the only one who knows what you’ve just asked for: nothing on the screen (f*ng big and overloaded with all the names and colors and icons) resembles any progress or name of the operation – like nothing happened. Until you refresh the screen, then you’ll see a light orange toast.

4. UI is inconsistent:

   a. You deply a model, then you stare at the unhealthy status on the endpoint details page (and you go-visit other resources/statuses and come back and see same result) for 30 minutes, wrapping your head where to find any logs. Then notification appears saying that deployment timed out and you can see more on the details page — but there will be nothing for you, sucker!

   b. You just worship the cult of Refresh button — be my guest, hit it everywhere you see it, if you’re waiting for something from MSFT. (There should be a browser plugin with customizable intervals per every refresh button to be pressed on that portal — just kidding, there should be no custom intervals, that extension should just hit-that-button automatically or it would be from MSFT too). UPDATE: I’ve made it myself — check Azure Refresh Button Worshiper, Browser Extension.

   c. List of notifications is evolving the way they want: one time you see A, B and C messages, then in 3 min event D appears and message C is gone, who cares!

5. You’re welcomed into the community who puts everything in the name, because UI does not help with knowing what it is, like: name-of-thing-at-location-and-creation-date-and-thing-type-and-probably-better-to-put-the-author-in-the-name-too