Google Cloud products, services, and solutions practice questions

Cloud SQL

Why wrong: Cloud SQL is a managed relational database (MySQL, PostgreSQL, SQL Server) designed for transactional workloads (OLTP). It's not designed for petabyte-scale analytics — it's limited in storage and compute scale.

Cloud Bigtable

Why wrong: Cloud Bigtable is a NoSQL database optimized for single-row reads/writes at high throughput (IoT, time-series). It doesn't support SQL analytics or ad-hoc query patterns.

Cloud Spanner

Why wrong: Cloud Spanner is a globally distributed relational database for OLTP workloads (transactions). While it supports SQL, it's not optimized for analytical query patterns over petabytes of data.

Google Kubernetes Engine (GKE)

Why wrong: GKE runs containers using Kubernetes but requires cluster configuration and management. It doesn't scale to zero by default and requires more operational expertise than Cloud Run.

Compute Engine

Why wrong: Compute Engine provides VMs, not managed container platforms. It requires significant configuration and management and doesn't automatically scale to zero.

App Engine Standard

Why wrong: App Engine Standard also scales to zero and is managed, but it's a code deployment platform (not container-native). Cloud Run accepts any container, not just App Engine-compatible runtimes.

BigQuery ML

Why wrong: BigQuery ML allows training ML models using SQL queries within BigQuery — useful for analysts but limited in the types of models and features. Vertex AI is a complete MLOps platform for more complex model training.

Cloud AI Platform Notebooks (now Vertex AI Workbench)

Why wrong: Vertex AI Workbench provides managed Jupyter notebook environments for ML development — it's an IDE for data scientists, not the full managed training and serving pipeline.

Cloud Dataflow

Why wrong: Dataflow is a managed stream and batch data processing service (Apache Beam), not an ML platform. It's used for ETL and data pipeline work, not model training.

Persistent Disk

Why wrong: Persistent Disk provides block storage for Compute Engine VMs — it functions like a hard drive attached to a server. It's not an object store and doesn't provide global web accessibility.

Cloud Filestore

Why wrong: Cloud Filestore provides managed NFS file system storage for applications needing POSIX file access from VMs. It's a file system, not an object store.

Cloud Spanner

Why wrong: Cloud Spanner is a globally distributed relational database for structured transactional data. It cannot store arbitrary unstructured objects like images or videos.

Vertex AI Workbench

Why wrong: Vertex AI Workbench is a managed Jupyter notebook environment for data scientists writing ML/Python code. It's not a BI dashboarding tool for non-technical stakeholders.

Cloud Dataprep by Trifacta

Why wrong: Cloud Dataprep is a visual data preparation and cleaning tool — used to transform data before analysis. It's not a dashboarding or report-sharing tool.

BigQuery Studio

Why wrong: BigQuery Studio provides an integrated SQL and notebook environment for data analysts in GCP — not a shareable dashboard tool for non-technical stakeholders.

Cloud Vision API

Why wrong: Cloud Vision API analyzes images (label detection, OCR, face detection). It does not process natural language text or enable conversational applications.

BigQuery ML

Why wrong: BigQuery ML builds ML models (regression, classification) using SQL — it's a predictive modeling tool, not a conversational AI platform.

Cloud Translation API

Why wrong: The Translation API converts text between languages. While related to natural language, it doesn't understand intent or enable conversational interactions.

Cloud Tasks

Why wrong: Cloud Tasks manages task queues for asynchronous, guaranteed delivery to a single worker. It doesn't support the fan-out pattern (one message → multiple independent subscribers).

Cloud Scheduler

Why wrong: Cloud Scheduler creates cron-based scheduled jobs. It doesn't provide a messaging queue or pub-sub pattern for service-to-service communication.

Eventarc

Why wrong: Eventarc routes events from GCP services to Cloud Run and other targets — it's an event routing service built on Pub/Sub. For direct service-to-service messaging patterns, Pub/Sub is the foundational service.

Cloud SQL (PostgreSQL)

Why wrong: Cloud SQL is a regional database — it doesn't automatically replicate globally. Cross-region failover is supported but not automatic global distribution with consistent transactions.

Cloud Bigtable

Why wrong: Bigtable is a NoSQL database (key-value) optimized for high-throughput non-transactional workloads. It doesn't support ACID transactions or SQL queries.

Firestore

Why wrong: Firestore supports ACID transactions and global distribution but is a NoSQL document database — it lacks full SQL query capabilities required for financial reporting and complex joins.

Cloud Dataproc

Why wrong: Cloud Dataproc is a managed Hadoop/Spark service for batch processing and analytics. While Spark Streaming exists, Dataflow (Beam) is the purpose-built GCP stream processing service.

Cloud Composer

Why wrong: Cloud Composer is a managed Apache Airflow service for orchestrating batch data workflows. It schedules and orchestrates pipelines but isn't itself a stream processing engine.

BigQuery Streaming Insert

Why wrong: BigQuery streaming inserts write data to BigQuery in near-real-time, but they don't provide stream processing transformations. Data must be transformed before streaming insert.

BigQuery ML for both use cases — it trains vision models on image data stored in BigQuery.

Why wrong: BigQuery ML builds regression, classification, and forecasting models from tabular data. It cannot process images or perform computer vision tasks.

Vertex AI AutoML Vision — train a custom model on your own images.

Why wrong: AutoML Vision trains custom models on your labeled image data. The question asks for pre-trained APIs that work without training — Vision API serves this need.

Cloud Natural Language API for text extraction from images.

Why wrong: The Natural Language API analyzes text content (sentiment, entities, syntax) but does not process images or perform OCR. Vision API provides OCR.

Cloud Run

Why wrong: Cloud Run runs individual containerized services serverlessly without requiring Kubernetes knowledge. For orchestrating many interdependent microservices with custom networking, storage, and scheduling, GKE provides more control.

Cloud Composer

Why wrong: Cloud Composer is a managed Apache Airflow service for orchestrating data workflow pipelines (DAGs), not for container orchestration.

Compute Engine with managed instance groups

Why wrong: MIGs manage groups of identical VMs, not containers. Container orchestration features (pod scheduling, Kubernetes networking, service discovery) require GKE.

Cloud SQL with read replicas to distribute updates to clients.

Why wrong: Cloud SQL is a relational database for traditional applications. It doesn't provide native real-time client synchronization — clients would need to poll for updates.

BigQuery streaming inserts for real-time data delivery to mobile clients.

Why wrong: BigQuery streaming inserts write analytical data to BigQuery — it's not designed for real-time client-facing data synchronization in mobile apps.

Cloud Pub/Sub subscriptions on the mobile clients.

Why wrong: While Pub/Sub delivers messages, mobile clients don't directly subscribe to Pub/Sub topics. Firebase SDKs abstract the real-time synchronization mechanism for mobile/web clients.

Cloud Firewall (VPC firewall rules)

Why wrong: VPC firewall rules operate at the network level (IP/port filtering) inside the VPC. They don't inspect HTTP request content for application-layer attacks like SQL injection.

Cloud VPN

Why wrong: Cloud VPN creates encrypted tunnels between networks for private connectivity. It's not a security inspection service for public internet traffic.

Cloud Identity-Aware Proxy (IAP)

Why wrong: IAP controls authenticated user access to applications (identity gate). It doesn't protect against DDoS attacks or inspect traffic for SQL injection patterns.

Cloud VPN with split tunneling for internal application access.

Why wrong: VPN grants network-level access — once connected, users can reach all resources on that network. IAP provides per-application access control based on identity, which is more granular and aligns with zero trust.

Cloud Armor IP allowlist to restrict access to corporate office IP ranges.

Why wrong: IP allowlists restrict by network location — the opposite of zero trust. Remote employees' home IPs can't be statically allowlisted. IAP uses identity, not IP, for access decisions.

Cloud Load Balancing with SSL termination.

Why wrong: SSL termination encrypts traffic but doesn't authenticate users or make authorization decisions. It's infrastructure, not an access control mechanism.

Cloud Interconnect — it connects on-premises to Google Cloud.

Why wrong: Cloud Interconnect provides network connectivity between on-premises and GCP. It doesn't manage workload deployment or provide unified container management.

Google Distributed Cloud — runs Google Cloud services inside the customer's data center.

Why wrong: Google Distributed Cloud (GDC) brings Google Cloud hardware and services on-premises. It's an extreme option for air-gapped environments. Anthos is the broader hybrid management platform for most use cases.

Cloud Deployment Manager — deploys resources via infrastructure-as-code templates.

Why wrong: Cloud Deployment Manager is GCP's native IaC service (similar to CloudFormation). It doesn't manage cross-environment container workloads.

Use Gmail to manually send all transactional emails.

Why wrong: Gmail is for human communication, not programmatic transactional email at application scale. Gmail has strict sending limits unsuitable for application-generated emails.

Use BigQuery to store and send emails directly to customers.

Why wrong: BigQuery is a data warehouse for analytics — it cannot send emails. Email requires dedicated email infrastructure.

Deploy an SMTP server on Compute Engine and send emails directly from GCP IP addresses.

Why wrong: GCP blocks outbound SMTP port 25 by default, and GCP IP ranges are often blocklisted by email receivers for spam prevention. Third-party email services are the correct solution.

Cloud Scheduler — cron-based job scheduling.

Why wrong: Cloud Scheduler triggers jobs on a cron schedule but doesn't manage multi-step dependencies, retry logic, or complex DAG workflows between different services.

Cloud Dataflow — stream and batch processing.

Why wrong: Dataflow processes data within a single pipeline (transformation). Cloud Composer orchestrates the sequence of multiple services and pipelines — one of which could be a Dataflow job.

Cloud Functions — event-driven function execution.

Why wrong: Cloud Functions executes individual functions on events. For complex multi-step workflows with dependencies across many services, Composer (Airflow) is the appropriate orchestration tool.

Cloud CDN — it manages domain names for cached content.

Why wrong: Cloud CDN caches content at edge locations. DNS record management for custom domains is handled by Cloud DNS, not CDN.

Cloud Load Balancing — it automatically assigns domain names.

Why wrong: Cloud Load Balancing provides IP addresses for load balancers, not domain name management. DNS records in Cloud DNS map custom domain names to those IP addresses.

Cloud Armor — it routes traffic based on domain names.

Why wrong: Cloud Armor is a security service (WAF/DDoS). DNS routing requires Cloud DNS configuration.

BigQuery ML — train a custom language model on company policy documents.

Why wrong: BigQuery ML builds traditional ML models (regression, classification) from tabular data. It's not designed for large language model training or conversational AI from documents.

Cloud Natural Language API — it reads and summarizes documents automatically.

Why wrong: The Natural Language API analyzes text (sentiment, entity extraction, syntax). It doesn't build conversational AI assistants or perform RAG-based question answering.

Cloud Translation API — it translates policy documents into the user's language.

Why wrong: Cloud Translation converts text between languages. Building a Q&A assistant over documents requires LLM + retrieval infrastructure, not translation.

Cloud Dataflow — it runs all Spark and Hadoop jobs natively.

Why wrong: Dataflow runs Apache Beam pipelines, not native Spark or Hadoop jobs. Migrating to Dataflow requires rewriting jobs in the Beam programming model.

BigQuery — it runs Spark SQL queries via BigQuery Spark stored procedures.

Why wrong: BigQuery supports Spark stored procedures but is primarily a SQL data warehouse. It's not a full Hadoop/Spark replacement for arbitrary batch processing jobs.

Google Kubernetes Engine — deploy Spark clusters on Kubernetes.

Why wrong: Spark can run on GKE (Spark on Kubernetes) but this requires managing Kubernetes and Spark configuration — more complex than Dataproc's managed service.