When choosing how to implement this requirement, you need to answer several questions:
Where to run the compute?
For this service, you have a long-running process, so it is not suitable for Cloud Functions. You will be using Java and Spring Boot in a container, so the startup time will be several seconds when using Cloud Run. This means it will not be able to cope well with on-demand requests within the required 500 ms. However, the autoscaling of Cloud Run will still be able to cope with unexpected demand and to be highly available. You will see how to make use of Cloud Run in a slightly different way to get the best of both worlds.
What type of database to use?
This service requires the use of a traditional relational database, or at least a service that appears as such to the application code. You will use Cloud SQL, as it is the simplest solution. Specifically, use Cloud SQL for PostgreSQL, as PostgreSQL is a de facto standard supported by several other database services on Google Cloud and other clouds. It gives you the option to switch to more powerful options later that also support PostgreSQL, if you need to. When architecting applications, keeping your options open like this is a good idea.
How to connect to the database?
You’ll want to follow the 12-factor principles of externalizing configuration and storing any secrets securely. You will make use of Cloud Secret Manager for storing secrets and explore secure options for connecting to the chosen database.
How to identify a user?
You want to be able to authenticate users and link them uniquely to the facts they provide. Storing sensitive information like email addresses and passwords, creating a registration form, and changing and resetting passwords is a big headache you want to avoid if at all possible. You will use Identity Platform, a service that does the authentication and user management for you.
Summary of Services
Here is a summary of the Google Cloud services you will be using in this solution.
Spring Boot with Spring Cloud GCP
Spring is a popular Java framework for building web applications. It is a mature framework that has been around for a long time and has a large community. With the release of Spring Boot in 2012, it became even easier to get started with Spring, and it is now the most popular way to build Spring applications. I started using Spring in the mid-2000s, and I found it a very steep learning curve on par with the difficulty I had when learning Kubernetes for the first time. Spring Boot was a godsend in making Spring easy to adopt and be productive with.
One of the key features of Spring Boot is that it is opinionated. It makes a lot of decisions for you and provides a lot of defaults. However, it also provides abstractions over specific technologies, keeping options open, which is a great architectural principle that enables us to remain open to change. For example, Spring Boot provides abstractions over databases, messaging systems, caching, and security. Switching between different implementations of these technologies is often as simple as changing the configuration. A bit later you will see how this has been extended to the services provided by Google Cloud with Spring Cloud GCP.