Google Cloud for data and application teams

The services people mean when they say “data on Google Cloud” often include object storage, a warehouse such as BigQuery, and a container platform such as Google Kubernetes Engine. You still have to decide which system is the source of truth. A warehouse that every job writes “just this once” is not a source of truth.

Shared responsibility, and the extra phrase

Google’s architecture documentation distinguishes the shared responsibility model from an idea it calls shared fate. The shared responsibility model says the provider and the customer own different parts of security, and that the split moves with the kind of service. Shared fate is Google’s argument that responsibility alone does not help customers configure those services well, and that the provider should offer stronger defaults and guidance.

Read that as a vendor’s design philosophy, which the documentation states openly. It does not move your data classification, your IAM bindings, or your application bugs onto Google. A managed service can patch its own software and still run the query you should not have allowed.

Analysis

Choose Google Cloud for a workload when the data services and the operating model fit, not because a keynote used the word intelligence. Put the project structure and the IAM review in the first design, beside the pipeline drawing. The pipeline is the easy picture. The project is what you will still be cleaning up in a year.

What the documentation does not choose for you

It does not say whether your warehouse should be the system of record for a given product. That is an application decision. Shared fate will not attend the design review.