Designing a cloud solution infrastructure that meets business requirements

Brambleton Retail is redesigning its order service. Today the order API calls the warehouse system, the loyalty system, and the email service in sequence before it responds to the customer. When the warehouse system slows down, checkout requests time out and orders are lost. Each downstream system must still receive every order, and the teams want to add more consumers later without changing the order API.

Which design should you recommend?

1. AHave the order API publish an order event to a Pub/Sub topic, with a separate subscription for each downstream systemPub/Sub decouples the API from slow consumers, keeps each message until its subscription acknowledges it, and lets new consumers attach with a new subscription without changing the publisher.
2. BIncrease the order API timeout and add retries to each downstream callLonger timeouts and retries keep the API coupled to the slowest system, so checkout still slows down and retries can create duplicate calls.
3. CWrite orders to a Cloud Storage bucket and have each system poll the bucketPolling a bucket adds latency and leaves each consumer to track which files it has processed, which Pub/Sub subscriptions already handle.
4. DDeploy the order API and all downstream systems in one GKE cluster to reduce network latencyPlacing services together does not fix the synchronous chain; a slow warehouse system would still block checkout.

Pinecrest Mutual is writing the requirements document for a new claims portal. The business analyst collected five statements from stakeholders and asked the architect to separate the non-functional requirements, which will drive the infrastructure design, from the functional ones, which go to the product backlog.

Which two statements are non-functional requirements? Each correct answer presents a complete solution.

1. AThe portal must return 95 percent of page loads in under 300 milliseconds for users in North America.This sets a performance target that the design must meet, which makes it a non-functional requirement.
2. BAfter a regional outage, the portal must be serving again within 2 hours with no more than 10 minutes of lost claim data.This states a recovery time objective and a recovery point objective, which are reliability requirements and therefore non-functional.
3. CA policyholder can upload up to five photos when filing a claim.This describes something the system does for a user, so it is a functional requirement.
4. DAn adjuster receives an email when a claim is assigned to them.Sending a notification is a behavior of the system, which is a functional requirement.
5. EA supervisor can reassign a claim to a different adjuster.Reassignment is a user-facing feature, so it is a functional requirement.

Glenmora Bank is moving its mortgage application platform to Google Cloud. The executive sponsor approved the program for two stated reasons: release new features to customers faster, and reduce the infrastructure cost of processing each mortgage application. The architect must propose KPIs that the steering committee will review each quarter to judge whether the program is succeeding.

Which two KPIs best measure the program against its stated goals? Each correct answer presents part of the solution.

1. ANumber of production deployments per month and the lead time from code commit to productionDeployment frequency and lead time directly measure how fast features reach customers, which is the first stated goal.
2. BTotal number of virtual machines migrated to Google CloudA count of migrated VMs measures migration activity, not whether releases are faster or costs per application are lower.
3. CInfrastructure cost per processed mortgage applicationA unit cost ties cloud spend to business volume, so it shows whether cost per application falls even as volume changes.
4. DNumber of Google Cloud certifications earned by the engineering teamCertifications can support adoption, but they are an input to the program rather than a measure of its business outcomes.
5. ETotal monthly Google Cloud billThe total bill rises and falls with business volume, so on its own it cannot show whether the cost of each application went down.

Quillfeather Insurance runs its policy administration system on a 6 TB PostgreSQL database that it is moving to Google Cloud. All users are in one country, and the database will run in a single region. Actuaries run heavy analytical queries against live policy data every day, which today slow down transactions. The CTO wants these queries to run fast on current data, refuses to build and run a separate ETL pipeline to a data warehouse, and wants to keep PostgreSQL compatibility so that application code does not change.

Which database design best balances these requirements?

1. ASpanner with a multi-region configuration and the PostgreSQL interfaceSpanner gives global scale and strong consistency that a single-country system does not need, and its PostgreSQL interface is not fully compatible with existing PostgreSQL code, so the trade-off is poor.
2. BCloud SQL for PostgreSQL with Datastream replicating changes to BigQuery for analyticsThis is a sound analytics design, but it is exactly the separate replication pipeline the CTO refused to build and run.
3. CCloud SQL for PostgreSQL with a larger machine type for the primary instanceScaling up the primary gives more headroom but does not make row-oriented analytical scans fast, and the analytical load still competes with transactions.
4. DAlloyDB for PostgreSQL with the columnar engine enabled and a read pool instance for the actuariesAlloyDB is PostgreSQL-compatible, its columnar engine speeds analytical scans on live data, and a read pool keeps the analytical load away from the primary, all without a separate pipeline.