Designing so a solution can adopt new cloud capabilities, follow changing business needs and move toward managed, cloud-first services.
Study it
Designing for change: cloud-first and managed services
Lesson coming
Sample questions
Try each one before opening the answer. Every option is explained, with the Google Cloud documentation page that proves it.
Question 1 · choose 1
A manufacturer's web storefront reads directly from its order database. Over the next two years the business plans to add a mobile app, dealer integrations and partner marketplaces, and each channel will change on its own schedule. Which design choice best prepares the backend for these changes?
AGive each new channel its own database user with read and write access to the orders
BExpose the order functions as versioned APIs in Apigee for every channel to call
CCopy the storefront backend for each new channel and deploy each copy separately
DMove the order database to Spanner so that every new channel can query it directly
Show the answer and why
AGive each new channel its own database user with read and write access to the orders
Incorrect
Direct database access couples every channel to the schema, so any change to it can break all of them.
BExpose the order functions as versioned APIs in Apigee for every channel to call
Correct
A stable, versioned API layer hides the database from the channels, so each channel and the backend can evolve independently behind a consistent interface.
CCopy the storefront backend for each new channel and deploy each copy separately
Incorrect
Copies drift apart, multiplying the code to maintain and making business rules inconsistent across channels.
DMove the order database to Spanner so that every new channel can query it directly
Incorrect
Spanner solves scale and availability, not coupling; the channels would still depend on the schema.
Designing for future change means loose coupling through clear interfaces. Putting versioned APIs in front of the backend lets new channels arrive and old ones change without touching each other.
An order service writes each new order to a database and then calls a fraud service directly. Over the next year the business expects to add loyalty, analytics and shipping consumers, and it does not want to change the order service for each one. What should the architect introduce now?
AA direct HTTP call from the order service to every new consumer service
BA Pub/Sub topic for order events, with a subscription per consumer
CA nightly export of the orders table for each consumer to read
DA shared Memorystore list that every consumer polls for orders
Show the answer and why
AA direct HTTP call from the order service to every new consumer service
Incorrect
Every new consumer would mean a code change and a new runtime dependency in the order service.
BA Pub/Sub topic for order events, with a subscription per consumer
Correct
Each consumer gets its own subscription to the same events, so new consumers are added without touching the publisher.
CA nightly export of the orders table for each consumer to read
Incorrect
A nightly batch delays every consumer by up to a day and ties them to the table's schema.
DA shared Memorystore list that every consumer polls for orders
Incorrect
A hand-built queue in a cache lacks per-consumer delivery, retention and retries.
Publish-subscribe decouples producers from consumers. Designing it in now lets the business evolve the set of consumers without reworking the core service.
A support assistant calls a specific Gemini model on Agent Platform. Google publishes a lifecycle table for these models, with announced retirement dates and recommended replacements, and newer models keep being released. How should the architect plan the assistant's future model upgrades?
APlan each migration to the listed replacement ahead of its retirement date
BSwitch the assistant to each new Gemini model on the day it is released
CPin the region and quota of the current model so that it stays available
DSelf-deploy the current Gemini model from Model Garden into the project
Show the answer and why
APlan each migration to the listed replacement ahead of its retirement date
Correct
Retirement dates are posted in advance and are not moved earlier, so migrations to the recommended upgrade can be planned and tested in time.
BSwitch the assistant to each new Gemini model on the day it is released
Incorrect
A release does not retire the current model, and switching at once skips the planning and testing that a migration needs.
CPin the region and quota of the current model so that it stays available
Incorrect
Region and quota settings do not change a model's retirement date, which applies to the model itself.
DSelf-deploy the current Gemini model from Model Garden into the project
Incorrect
Model Garden self-deploys open models; Gemini models are served by Google and follow the published lifecycle.
Future improvements include planned model upgrades. The model lifecycle page lists each model's retirement date and recommended upgrade, which lets the team schedule a migration and test the critical features before the date.
A company's order management monolith on Compute Engine is hard to change, and an earlier attempt to rewrite it from scratch was abandoned after two years. The business still needs new features every quarter and wants the application to become cloud native over time. Which approach should the architect recommend?
AFreeze feature work and rewrite the whole application in one new release
BGrow new services around the monolith and move its functions out gradually
CMove the monolith to a larger VM so that new features fit inside it more easily
DBuild a new user interface over the monolith and leave its internals unchanged
Show the answer and why
AFreeze feature work and rewrite the whole application in one new release
Incorrect
A big-bang rewrite repeats the approach that already failed and stops feature delivery for years.
BGrow new services around the monolith and move its functions out gradually
Correct
The strangler fig pattern grows new cloud-native services around the monolith and shifts functionality to them gradually.
CMove the monolith to a larger VM so that new features fit inside it more easily
Incorrect
A bigger VM adds capacity, but the code stays as hard to change as before.
DBuild a new user interface over the monolith and leave its internals unchanged
Incorrect
A new front end changes what users see, but the features still have to be built in the same hard-to-change code behind it.
Envisioning future improvements means evolving a system rather than replacing it at once. The strangler fig pattern adds new functionality as modern services that talk to the monolith and gradually moves the old functions, so the business keeps getting features during the change.
An organization's cloud estate has grown to hundreds of projects. The architect wants a continuous, low-effort way to find idle resources, oversized VMs and overly broad IAM grants as the environment keeps changing. What should the architect put in place?
AA yearly manual audit of every project by the central cloud platform team
BCloud Asset Inventory exports of the resources to a spreadsheet
CBudget alerts at 50%, 90% and 100% of each project's monthly budget
DActive Assist recommendations, reviewed and applied on a regular cadence
Show the answer and why
AA yearly manual audit of every project by the central cloud platform team
Incorrect
A yearly manual audit is neither continuous nor low effort at this scale.
BCloud Asset Inventory exports of the resources to a spreadsheet
Incorrect
An inventory lists what exists but does not say what is idle, oversized or overprivileged.
CBudget alerts at 50%, 90% and 100% of each project's monthly budget
Incorrect
Budget alerts track spending against a number; they do not point to specific waste or risky grants.
DActive Assist recommendations, reviewed and applied on a regular cadence
Correct
Active Assist recommenders analyze usage across projects and suggest cost, performance and security improvements such as idle resources, right-sizing and role changes.
Continuous improvement works best when the platform surfaces opportunities for you. Active Assist provides recommendations across cost, security, performance and more.