Regions

Google Cloud Platform uses regions, subdivided into zones, to define the geographic location of physical computing resources. When you run a job on AI Platform, you specify the region that you want it to run in.

You should typically use the region closest to your physical location or the physical location of your intended users, but note the available regions for each service as listed below.

Available regions

AI Platform is available in the following regions:

Region considerations

Training with Accelerators

Accelerators are available on a region basis. Below is a table that lists all the available accelerators for each region:

If your training job uses multiple types of GPUs, they must all be available in a single zone in your region. For example, you cannot run a job in us-central1 with a master worker using NVIDIA Tesla V100 GPUs, parameter servers using NVIDIA Tesla K80 GPUs, and workers using NVIDIA Tesla P100 GPUs. While all of these GPUs are available for training jobs in us-central1, no single zone in that region provides all three types of GPU. To learn more about the zone availability of GPUs, see the comparison of GPUs for compute workloads.

Insufficient resources

Demand is high for GPUs and for compute resources in the us-central1 region. You may get an error message in your job logs that says: Resources are insufficient in region: <region>. Please try a different region..

To resolve this, try using a different region or try again later.

Cloud Storage

• You should run your AI Platform job in the same region as the Cloud Storage bucket that you're using to read and write data for the job.

• You should use the Standard Storage class for any Cloud Storage buckets that you're using to read and write data for your AI Platform job.

Online prediction

• When you deploy a model for online prediction, you specify the region that you want prediction to run in. Online predictions are always served from the default region specified for the model.

Batch prediction

• You cannot deploy a model or model version in us-west1, us-west2, europe-west4, europe-north1, asia-east1, or asia-southeast1, but you can perform batch prediction in these regions using a TensorFlow SavedModel stored in Cloud Storage.
• For best performance in batch prediction, you should run your prediction job and store your input and output data in the same region, especially for very large datasets.
• When you deploy a model for batch prediction, you specify the default region that you want prediction to run in. When you start a batch prediction job, you can specify a region to run the job in, overriding the default region.

What's next

• See how to work with Cloud Storage and AI Platform.
• Read the details of Google Compute Engine regions and zones.
• Explore the pricing per regional group (US, Europe, Asia Pacific).
• See how to use online prediction and batch prediction in AI Platform.