GAE Flexible vs Cloud Run on GKE for streaming response

[guz...@gmail.com]

Greetings,

I would like your opinion on this - I have some HTTP endpoints that return a large streaming response (a file that gets created at request time). Since I am limited by both response timeout and size in GAE standard, I am using GAE Flexible for this. I looked at Cloud Run and saw it has similar limitations - no streaming response and limits on response timeout and size, so if I were to switch to Cloud Run, I would need to use it with GKE. My app is pretty stateless - it uses Datastore and no in-memory state, so I thought I could save a few bucks by switching to Cloud Run and also get faster deployments. I would like to get your thoughts about GAE Flexible vs Cloud Run on GKE - from what I understand, the downside of GAE Flexible is slower deployment (takes me about 10 minutes) and higher costs because there is at least one instance running at all times, and the downside of Cloud Run on GKE is having to manage Kubernetes (there is no dedicated devops people on our team, so it would take developers' time and effort). Can you think of more concerns? How would you recommend I approach this decision?

[Aref Amiri (Cloud Platform Support)]

Hi,

If your main objective is to save on costs, then Cloud Run would be a good solution, since you only pay when you are processing requests and the billing granularity is 0.1 seconds, whereas on GAE Flex, as you mentioned, you are always running at least one instance at a time and billing granularity is 1 minute. There are other factors that need to be considered such as, infrastructure, portability and Autoscaling, as discussed here, to help you with your final decision.

You can also refer to this Stackoverflow thread, as this topic has already been discussed and you can gather additional useful information to help you with your final decision.