Azure ML与Azure功能，用于运行R-Script

Question

The stats team has developed an R-Script that takes an array of variables and returns an array of integers (calculations). I want to build this as a stand alone function that I can call with HTTP requests from various apps. A lot of the stuff I am reading is out dated. Should I use Azure ML or Azure functions for this?

Note: The R-Script does not contain any machine learning. The R-Script contains a package that is not listed on azure.

Answer 1

Azure ML appears to be all about predictive models so I have not tried it. I tried Azure function under "consumption" pricing tier and I do not recommend it. Currently R site extensions are removed from the Azure function site's extension list so the work around is to copy the R folder to your Azure function and use R.Net library in C# to initiate R commands.

The issue with Azure function under "consumption" plan is that every 5 minutes of inactivity your function is erased from server memory so when the first user hit the app it initiates a "cold start". In layman's terms, during a cold start Azure finds a server that is available to run your function and copies all the files to that server and then executes the script. Copying the R folder takes very long and is not ideal for any client facing requests.

I'm sure using a pricing plan would solve the cold start issue but pricing tier means I'm hosting it on my VM which is not a "serverless" function and defeats the purpose of having a function. I might as well build an API web app. I am going to try Lambda AWS to see if it has the same cold start issues.