Cloud functions work OK if you don't use them too often
July 05, 2019

Cloud functions work OK if you don't use them too often

Anonymous | TrustRadius Reviewer
Score 5 out of 10
Vetted Review
Verified User

Overall Satisfaction with IBM Cloud Functions

We have several python functions that we had set up as API calls, but we have changed them to be cloud functions instead. We are using these on different applications for several different clients. These perform operations such as validating files, computing a value based off of a previously set up machine learning model and more complex report/database queries that take more than a second or two.
  • Great substitute for a simple API calls to run non-complicated code.
  • Easy way to run Python/Java/Javascript to get something done.
  • File validation.
  • They are not good if you are doing repetitive calls multiple times a minute.
  • They are not good for long processes.
  • They are not 100% reliable yet, they have been release for GA, but they don't standup to being beat up.
  • Implemented for a use-case that the documentation claimed it could support (100 calls a minute) and it failed. Couldn't handle more than 20-40.
  • Worked well for validating data, reduced API calls we needed to make and maintain significantly.
  • Good for cloud like cron jobs for sanity checks and smoke tests.
AWS Lambda is 100 times more robust than IBM cloud functions. They essentially do the same thing, but AWS works. AWS is stable. we have had epic failures with cloud functions. Support was horrible. We literally had an open ticket with them for 2 months and it literally went nowhere. They said it could do 100 calls a minute. We proved over and over that we couldn't get above 20 without getting failures. They had NO explanation whatsoever. The ticket got closed because we were tired of asking them questions and getting no understandable or usable response.
These functions are good if you are trying to diminish the overhead of API maintenance and support. Not too difficult to build and deploy. Good for doing model inference, or data validation. This is not good for doing tasks that take a long time. Not good for high compute and not good for things you are doing multiple times a minute.