@steren Hey! Congrats on the launch! I was wondering if the containers go to sleep? If they do, do you have any statistics on how fast apps take to cold boot?
@nahtnam there is no guarantee about how long we keep containers warm, it depends on many factors. In practice, at least minutes. I have seen a container of mine running for days. (Of course you do not pay when no request reaches it).
Cold start depends mostly entirely about what your code does at startup (not on the size of the image).
@steren Great! And do you have any information on what the overhead is for the cold start ups? For example, if my server takes 50ms to boot, how much overhead does Cloud Run add? 10ms or 100ms, etc?
@nahtnam not sure how much our system adds by itself. I will write an article to give more details. Most of the time comes from what files are loaded in memory and any initialisation that the code does before listening for requests.
Hi, I am new to google's ecosystem and I am not sure what to use, currently I am using GKE and cloud sql and it would be very interesting to compare this to GKE and GAE in terms of pricing and manual maintenance required.
This looks great! Congrats to @steren and team.
I've been a fan of FaaS but with reservations - until now: Cloud Run appears to address my concerns about flexibility and vendor lock-in... as I understand it, it is an open implementation of knative serving spec, with events based on cloudevents.io (in the node case at least).
Now I feel much more comfortable committing to Google Cloud Functions, knowing I have the option to break out and have completely flexibility with what and where I run.
Would be great if the upcoming v8 runtime (for GAS) were to find it's way to Cloud Run...
@webforspeed the pricing and scaling capabilities of Cloud Run are closer to Google Cloud Functions. Flex runs on VMs, so does not scale to zero and takes longer to scale up.
Google Cloud Platform