GPUs on Cloud Run

Google Cloud Serverless Applications
1

I’m running processes on Cloud Run. I tried spinning up a VM that was called from a Flask app on Cloud Run to get access to GPUs, but found the integration technically challenging, and perhaps the cost was not going to be optimal.

I was very excited, therefore, to hear that GPUs are now available on Cloud Run in us-central1: https://cloud.google.com/blog/products/application-development/run-your-ai-inference-applications-on-cloud-run-with-nvidia-gpus/

Although it looks like other regions will be added over time, my region, australia-southeast1, will probably be towards the last added.

Will latency be a significant issue to use a region well outside my own?

Any insights greatly appreciated.

Cheers

2

Hi Paul, if you’re using a custom domain, using a Cloud Load Balancer in front of your Cloud Run service for the custom domain serving will give you better latency than using Cloud Run’s built-in custom domains.

We hope to be adding a few regions soon (though not Australia yet), so hopefully the latency won’t be as bad as if you were using us-central1 today.