Ask HN: Google Cloud Functions – Do you use it in production? What do you think?
How do Google Cloud Functions stack up against AWS Lambda?
5 comments
Pros:
* When it works, it works. We're processing ~100 msg/s in prod right now, for pennies
Cons:
* Only US-Central region support (no cross region)
* Your function runs as Admin (root) in it's project
* Your function must be deployed to the same project as other assets (e.g. can't listen to GCS or PubSub events cross project)
* No way to pass configs/params to a function (it's all gotta be in the code, including secrets)
* Very limited retry semantics (brute force, retry forever)
* No integration with Cloud Endpoints or APIgee, so every HTTP event driven function gets new unique URL
* Building on above, all HTTP endpoints will be public, no auth
* Slow deployments - e.g. took me 45min today to deploy 8 functions
* When it works, it works. We're processing ~100 msg/s in prod right now, for pennies
Cons:
* Only US-Central region support (no cross region)
* Your function runs as Admin (root) in it's project
* Your function must be deployed to the same project as other assets (e.g. can't listen to GCS or PubSub events cross project)
* No way to pass configs/params to a function (it's all gotta be in the code, including secrets)
* Very limited retry semantics (brute force, retry forever)
* No integration with Cloud Endpoints or APIgee, so every HTTP event driven function gets new unique URL
* Building on above, all HTTP endpoints will be public, no auth
* Slow deployments - e.g. took me 45min today to deploy 8 functions
I've noticed the speed of deployment as well. Any idea why it's that slow?
I would really like to use Google Cloud Functions as the integrate super easy with the rest of the Google Cloud platform, unfortunately, they are still only in the USA and if you have you DB somewhere else then it is not really optimal.
There are other pitfalls as it can connect to your Google Cloud SQL without opening it up to the public as it is not in the same network and there is a limit of a 1000 functions deployed.
But the developer experience is in top compared to how it is to deploy a http reachable endpoint on Amazon Lambda!
There are other pitfalls as it can connect to your Google Cloud SQL without opening it up to the public as it is not in the same network and there is a limit of a 1000 functions deployed.
But the developer experience is in top compared to how it is to deploy a http reachable endpoint on Amazon Lambda!
That's great feedback. I had no idea about the public access around GCF -> Google Cloud SQL or the US-only aspect
I think the main differentiator is Firebase integration. Especially registering with Analytics events such as Auth. But yeah obvs you trade convenience for vendor lockin ;)
Here is a Codelab to get you started and I believe you'll see how it can reduce your development time dramatically...
CodeLab: Cloud Functions for Firebase
https://codelabs.developers.google.com/codelabs/firebase-clo...
Here is a Codelab to get you started and I believe you'll see how it can reduce your development time dramatically...
CodeLab: Cloud Functions for Firebase
https://codelabs.developers.google.com/codelabs/firebase-clo...