Google Professional Data Engineer on Google Cloud Platform Exam Dumps and Practice Test Questions Set 12 Q166-180
Visit here for our full Google Professional Data Engineer exam dumps and practice test questions.
Question 166
A global logistics company wants to monitor the status of thousands of delivery trucks in real time. Telemetry data includes GPS coordinates, speed, and engine parameters. The system must trigger alerts for route deviations or abnormal engine behavior. Which architecture is most appropriate?
A) Cloud Pub/Sub → Dataflow → BigQuery → Looker
B) Cloud SQL → Cloud Functions → Dataproc → Looker
C) Cloud Storage → Cloud Run → BigQuery ML
D) Bigtable → App Engine → Data Studio
Answer: A
Explanation:
Real-time monitoring of a global fleet requires ingesting telemetry from thousands of trucks simultaneously. Data is continuously generated from sensors measuring GPS location, speed, fuel consumption, engine temperature, and other critical metrics. The ingestion system must handle high-throughput streaming data and guarantee reliability even during network fluctuations or bursts of traffic. Cloud Pub/Sub is an ideal choice because it provides a fully managed, globally distributed messaging system that can reliably collect millions of events per second. Pub/Sub ensures that messages are durable, supports at-least-once delivery, and decouples producers and consumers, enabling trucks to send data without being blocked by downstream processing workloads.
Dataflow serves as the processing engine for this stream of telemetry data. It can normalize incoming events, enrich them with metadata such as driver details, routes, or maintenance history, and apply windowed computations to detect anomalies. For example, if a truck deviates from its planned route or engine parameters exceed safe thresholds, Dataflow can trigger alerts for operational teams. Its stateful processing capabilities allow continuous monitoring over time windows, such as detecting sustained over-speeding or fuel inefficiency trends. Managed scaling ensures that the pipeline can adjust automatically to peak traffic without manual cluster management, providing seamless operation across global regions.
BigQuery acts as the analytical warehouse for both raw and enriched data. It supports time-partitioned tables, which allow analysts to efficiently query specific periods without scanning all records. Historical queries enable predictive maintenance, route optimization analysis, and fuel efficiency studies. Data stored in BigQuery can also be used to train machine learning models to detect potential failures or optimize dispatching and routing strategies. Its serverless architecture allows the organization to scale queries automatically without infrastructure management, making it ideal for large datasets generated by global fleets.
Looker integrates with BigQuery to provide visualization dashboards and alerting interfaces. Fleet managers can monitor real-time operational metrics, receive notifications for anomalies, and analyze long-term trends. By combining interactive dashboards with automated alerts, operations teams can take timely corrective action, reduce fuel costs, prevent delays, and enhance overall fleet safety.
Alternative solutions are less suitable. Cloud SQL combined with Cloud Functions and Dataproc introduces latency and operational complexity, as manual scaling is required and stream processing capabilities are limited. Cloud Storage with Cloud Run and BigQuery ML is better suited for batch processing workflows and does not provide real-time alerting. Bigtable, combined with App Engine and Data Studio,o can handle time-series storage, but lacks integrated streaming analytics and automated anomaly detection for operational workflows. Therefore, Pub/Sub, Dataflow, BigQuery, and Looker together form the most appropriate end-to-end solution for real-time fleet monitoring, anomaly detection, and operational analytics.
Question 167
A fintech company wants to implement a global, real-time fraud detection system. Transactions must be evaluated in milliseconds, and risk scores must be available instantly for authorization decisions. Which storage solution should be used?
A) Memorystore Redis
B) Cloud SQL
C) BigQuery
D) Cloud Storage
Answer: A
Explanation:
Real-time fraud detection in financial systems demands ultra-low latency to ensure that every transaction can be approved or blocked in milliseconds. Memorystore Redis is a fully managed in-memory key-value store that provides sub-millisecond read and write operations, making it ideal for storing operational data such as account histories, transaction velocity metrics, and fraud risk scores. Its low-latency access ensures rapid decisioning for high-volume transactions, preventing fraud while maintaining a smooth user experience.
Redis supports high concurrency and can scale to handle global traffic spikes during peak periods, such as holiday shopping or promotional events. Advanced data structures like hashes, sorted sets, and bitmaps enable efficient implementation of rules, thresholds, and velocity checks required for evaluating potential fraudulent behavior. Memorystore provides managed replication and high availability, ensuring that the system remains operational even during node failures or regional disruptions. Managed operations reduce overhead, eliminating the need for administrators to manually manage clusters, scaling, or failover procedures.
Alternative solutions are not suitable for real-time fraud detection. Cloud SQL provides relational ACID transactions, but its latency and scaling limitations make it unsuitable for sub-millisecond operational lookups. BigQuery is designed for large-scale analytics, not low-latency retrieval, and would be too slow to support transaction approvals. Cloud Storage is object-based and intended for archival or batch processing, making it unsuitable for rapid access to operational risk data.
By using Memorystore Redis, the fintech company ensures that fraud evaluation occurs instantly across all regions, with high reliability and scalability. This solution provides the low-latency operational database necessary to evaluate transactions in real time, maintain user trust, and prevent financial losses due to fraudulent activity. Redis’s speed, advanced data structures, and global scalability make it the industry-standard choice for operational fraud detection.
Question 168
A healthcare provider wants to predict patient readmission risk using electronic health records and real-time telemetry from wearable devices. The models must update continuously as new data arrives to remain accurate. Which architecture is most appropriate?
A) Pub/Sub → Dataflow → BigQuery → Vertex AI
B) Cloud Storage → App Engine → BigQuery ML
C) Dataproc → Cloud Storage → Cloud Functions
D) Bigtable → Cloud Run → AutoML Tables
Answer: A
Explanation:
Healthcare predictive analytics requires integrating structured electronic health record (EHR) data with continuous telemetry streams from wearable devices. Data ingestion must be reliable, scalable, and capable of handling high-frequency events such as heart rate, blood pressure, activity levels, and sleep patterns. Pub/Sub provides a fully managed, globally distributed messaging system for ingesting this continuous data. It guarantees durable delivery, supports at-least-once message semantics, and decouples data producers from downstream processing, allowing reliable real-time ingestion from thousands of patients and devices.
Dataflow processes these streams in near real time, performing normalization, enrichment, aggregation, and feature extraction. For example, it can combine telemetry with historical EHR data, compute rolling averages, detect anomalies, and generate features for predictive models. Dataflow supports stateful processing and windowed computations, which are essential for time-dependent metrics such as trends in heart rate variability or abnormal activity patterns. Its managed service ensures exactly-once processing, high availability, and automatic scaling, reducing operational complexity while maintaining data integrity.
BigQuery acts as the analytical warehouse, storing both raw and processed data for large-scale querying. Analysts can explore trends, perform feature engineering, and build cohorts for research or model training. BigQuery’s columnar storage, serverless execution, and partitioned tables enable efficient queries across billions of rows of data without manual infrastructure management. This allows healthcare teams to analyze historical patterns and support machine learning workflows seamlessly.
Vertex AI integrates with BigQuery to train, evaluate, and deploy predictive models for patient readmission risk. Continuous retraining pipelines ensure that models are updated as new telemetry and EHR data arrive, maintaining accuracy over time. Vertex AI also provides monitoring for model drift, experiment tracking, and low-latency prediction endpoints for clinical decision support systems. Alternative architectures are less effective. Cloud Storage with App Engine and BigQuery ML cannot efficiently handle continuous streaming data or provide fully managed retraining pipelines. Dataproc with Cloud Storage and Cloud Functions adds operational complexity and lacks integration for streaming-to-ML workflows. Bigtable with Cloud Run and AutoML Tables can store telemetry data, but it is less suited for combining structured EHR with real-time streaming and automated retraining.
Therefore, Pub/Sub, Dataflow, BigQuery, and Vertex AI provide the most appropriate end-to-end architecture for predictive healthcare analytics, continuous model updates, and patient readmission risk prediction.
Question 169
A global e-commerce company wants to provide real-time personalized recommendations to users. Clickstream events must be ingested continuously, enriched with user profiles, and used for analytics and machine learning pipelines. Which architecture is most appropriate?
A) Cloud Pub/Sub → Dataflow → BigQuery → Vertex AI
B) Cloud SQL → Cloud Functions → BigQuery ML
C) Cloud Storage → Dataproc → BigQuery
D) Bigtable → App Engine → AutoML Tables
Answer: A
Explanation:
Real-time personalized recommendations rely on continuous ingestion of high-volume clickstream events, such as page views, product clicks, and search queries. Cloud Pub/Sub serves as the ingestion layer, capable of handling millions of events per second globally. Pub/Sub ensures reliable message delivery, supports automatic scaling, and decouples producers from downstream consumers, allowing event sources to continue sending data without being blocked. It guarantees durability and at least once delivery, making it ideal for mission-critical e-commerce systems where no user interaction should be lost.
Dataflow processes incoming streams and batch data, performing normalization, enrichment, filtering, and aggregation. Clickstream events are enriched with user profile data, purchase history, and session information, enabling meaningful features for machine learning models. Dataflow supports windowed computations and stateful processing, allowing metrics such as session duration, click frequency, and rolling conversion rates to be calculated in real time. Its managed service ensures exactly-once processing semantics, automatic scaling, and high availability, reducing operational complexity while maintaining accuracy for analytics and ML pipelines.
BigQuery serves as the analytical warehouse, storing raw and enriched events. Analysts can query billions of rows efficiently to identify trends, evaluate campaign effectiveness, and generate features for machine learning models. BigQuery’s serverless architecture allows queries to scale automatically, making it ideal for global e-commerce datasets. Time-partitioned tables enable efficient historical analysis while controlling costs, which is critical for long-term personalization and recommendation strategies.
Vertex AI enables the creation, training, and deployment of predictive models for product recommendations. Continuous retraining pipelines allow models to incorporate the latest clickstream and behavioral data, maintaining relevance and personalization accuracy. Vertex AI supports experiment tracking, drift monitoring, and low-latency prediction endpoints, enabling real-time recommendations on web and mobile platforms.
Alternative architectures are less suitable. Cloud SQL with Cloud Functions and BigQuery ML cannot scale to handle millions of streaming events per second and may introduce latency in processing. Cloud Storage with Dataproc and BigQuery is better suited for batch processing rather than real-time streaming analytics. Bigtable with App Engine and AutoML Tables can store data, but lacks native, fully managed integration for real-time enrichment, analytics, and continuous ML pipelines. Therefore, Pub/Sub, Dataflow, BigQuery, and Vertex AI together form the optimal end-to-end architecture for real-time personalized recommendations in a global e-commerce environment.
Question 170
A fintech company wants to evaluate transactions in real time to prevent fraud. Each transaction must be scored within milliseconds, and the system must handle global traffic spikes reliably. Which storage solution is most appropriate?
A) Memorystore Redis
B) Cloud SQL
C) BigQuery
D) Cloud Storage
Answer: A
Explanation:
Fraud detection requires evaluating financial transactions at extremely low latency. Every transaction must be scored within milliseconds to prevent fraud without affecting legitimate user experience. Memorystore Redis is a fully managed, in-memory key-value store capable of sub-millisecond read and write operations, making it ideal for storing operational data such as account histories, velocity metrics, blacklists, and precomputed risk scores. Its in-memory architecture ensures instant access to critical information required for decision-making at transaction time.
Redis supports extremely high concurrency, allowing the system to handle global traffic spikes during peak times, such as holidays or promotions. Advanced data structures, including hashes, sorted sets, and bitmaps, enable efficient implementation of rules, thresholds, and aggregation of multiple risk signals. Memorystore provides managed replication and high availability, ensuring continuous operation even if nodes or regions experience failures. Its fully managed service also reduces operational overhead, eliminating the need for administrators to manage scaling, replication, or failover.
Alternative solutions are less suitable. Cloud SQL provides ACID compliance, but cannot deliver the required sub-millisecond performance at scale, especially under high concurrency and global traffic. BigQuery is optimized for large-scale analytics but is unsuitable for low-latency transaction scoring. Cloud Storage is object-based, designed for archival or batch workflows, and cannot support real-time operational lookups.
By using Memorystore Redis, fintech companies achieve a globally scalable, highly available, low-latency solution for operational fraud detection. Redis ensures rapid access to transaction histories and risk scores, enabling real-time fraud prevention while maintaining performance and user experience during high-volume transaction periods. Its speed, reliability, and advanced data structures make it the ideal choice for operational risk scoring.
Question 171
A healthcare provider wants to predict patient readmission risk using structured EHR data and continuous telemetry from wearable devices. The models must update continuously as new data arrives. Which architecture is most appropriate?
A) Pub/Sub → Dataflow → BigQuery → Vertex AI
B) Cloud Storage → App Engine → BigQuery ML
C) Dataproc → Cloud Storage → Cloud Functions
D) Bigtable → Cloud Run → AutoML Tables
Answer: A
Explanation:
Predictive healthcare analytics requires combining structured electronic health record (EHR) data with high-frequency telemetry from wearable devices. Data ingestion must be reliable, scalable, and capable of handling continuous streams of events such as heart rate, blood pressure, activity level, and sleep patterns. Pub/Sub provides a fully managed, globally distributed messaging system that ensures durable delivery, decouples data producers from consumers, and supports at-least-once message semantics. It allows reliable real-time ingestion from thousands of patients and devices globally.
Dataflow processes this streaming data in near real time. It can normalize heterogeneous telemetry data, join streams with EHR records, aggregate metrics, and compute derived features necessary for predictive modeling. Dataflow supports stateful processing and windowed computations, which are essential for calculating rolling averages, detecting anomalies, or capturing time-dependent health patterns. Its managed service guarantees exactly-once processing semantics, automatic scaling, and high availability, reducing operational complexity while maintaining data integrity.
BigQuery acts as the analytical warehouse, storing both raw and processed data. Analysts and data scientists can perform large-scale queries for cohort analysis, feature engineering, and historical trend evaluation. Partitioned and clustered tables allow cost-efficient queries on time-series data without sacrificing performance. BigQuery provides a reliable platform to support predictive modeling and integration with downstream ML workflows.
Vertex AI enables the creation, training, and deployment of predictive models for readmission risk. Continuous retraining pipelines ensure models are updated as new data arrives, maintaining predictive accuracy. Vertex AI supports low-latency prediction endpoints for clinical decision support, experiment tracking, and drift monitoring. Alternative architectures are less suitable. Cloud Storage with App Engine and BigQuery ML cannot efficiently handle real-time streams or continuous retraining. Dataproc with Cloud Storage and Cloud Functions adds operational overhead and lacks integrated streaming-to-ML capabilities. Bigtable with Cloud Run and AutoML Tables may store telemetry data, but it is less suited for combining structured EHR with real-time predictive workflows.
Therefore, Pub/Sub, Dataflow, BigQuery, and Vertex AI provide the optimal end-to-end architecture for continuous predictive analytics in healthcare, enabling accurate readmission risk assessment and timely interventions.
Question 172
A global ride-sharing company wants to optimize driver allocation and surge pricing in real time. The system must ingest millions of location and trip events per second and feed analytics and machine learning pipelines. Which architecture is most appropriate?
A) Cloud Pub/Sub → Dataflow → BigQuery → Vertex AI
B) Cloud SQL → Cloud Functions → BigQuery ML
C) Cloud Storage → Dataproc → BigQuery
D) Bigtable → App Engine → AutoML Tables
Answer: A
Explanation:
Optimizing ride-sharing operations requires real-time processing of millions of location and trip events, such as GPS coordinates, ride requests, cancellations, and trip completions. Cloud Pub/Sub is the ideal ingestion layer for these high-volume streams, providing a managed, globally distributed messaging system capable of handling millions of events per second. Pub/Sub ensures at-least-once delivery, durability, and automatic scaling, decoupling data producers from consumers. This allows ride-sharing apps to send real-time events without delays, even during traffic spikes or peak demand hours.
Dataflow processes these events in real time, performing transformations, enrichments, and aggregations. For example, Dataflow can join event streams with driver availability, historical trip patterns, and surge pricing rules to compute dynamic metrics. Windowed computations and stateful processing allow detection of demand spikes, real-time surge calculations, and optimization of driver allocation. Dataflow’s managed, serverless architecture provides automatic scaling and exactly-once processing semantics, minimizing operational overhead while ensuring data integrity.
BigQuery serves as the analytical warehouse for storing raw and processed events. Analysts can query billions of rows efficiently to uncover patterns, evaluate pricing models, and generate features for machine learning pipelines. Partitioned and clustered tables reduce query costs while maintaining fast performance, enabling timely business insights. Historical data also supports trend analysis, forecasting, and model evaluation, ensuring operational decisions are data-driven.
Vertex AI integrates with BigQuery to build, train, and deploy predictive models for demand forecasting, surge pricing, and driver allocation optimization. Continuous retraining pipelines ensure models remain up to date with the latest data. Low-latency endpoints provide predictions in real time, allowing dynamic adjustment of prices and driver recommendations. Vertex AI also supports experiment tracking, monitoring, and model drift detection, ensuring predictive models remain accurate and reliable.
Alternative architectures are less suitable. Cloud SQL with Cloud Functions and BigQuery ML cannot handle millions of streaming events per second and may introduce latency. Cloud Storage with Dataproc and BigQuery is better suited for batch processing rather than continuous streaming. Bigtable with App Engine and AutoML Tables can store large datasets, but lacks native integration with streaming pipelines and continuous ML retraining. Therefore, Pub/Sub, Dataflow, BigQuery, and Vertex AI provide a scalable, low-latency, end-to-end architecture for global ride-sharing optimization.
Question 173
A fintech company needs to prevent fraudulent transactions in real time. Each transaction must be evaluated in milliseconds, and the system must maintain high availability across multiple regions. Which storage solution is most appropriate?
A) Memorystore Redis
B) Cloud SQL
C) BigQuery
D) Cloud Storage
Answer: A
Explanation:
Real-time fraud detection requires evaluating each transaction in milliseconds to prevent financial loss while ensuring legitimate transactions proceed without delay. Memorystore Redis is a fully managed, in-memory key-value store providing sub-millisecond read and write operations, making it ideal for operational data such as transaction histories, user velocity metrics, blacklists, and precomputed risk scores. Its in-memory design ensures rapid access to critical information required for decision-making at the time of each transaction.
Redis supports extreme concurrency, allowing the system to scale globally and handle traffic spikes during peak financial periods. Its advanced data structures, including hashes, sorted sets, and bitmaps, enable efficient aggregation of risk factors and implementation of velocity rules, thresholds, and anomaly detection logic. Memorystore provides managed replication and high availability, ensuring continuous operation even in the event of node or regional failures. Its fully managed nature reduces operational complexity by eliminating the need to manage scaling, failover, or clustering manually.
Alternative storage solutions are less suitable. Cloud SQL provides relational ACID compliance, but cannot reliably provide sub-millisecond latency at a global scale. BigQuery is designed for large-scale analytics and cannot support the operational, real-time lookup requirements of transaction scoring. Cloud Storage is object-based, intended for batch processing and archival, and cannot deliver the rapid access needed for fraud detection.
By leveraging Memorystore Redis, fintech companies achieve a highly available, low-latency, globally scalable operational database for fraud prevention. Redis ensures rapid access to account histories and risk scores, enabling real-time decision-making for millions of transactions per second while maintaining a high level of reliability, security, and performance. Its speed and advanced data structures make it the industry-standard choice for operational fraud scoring.
Question 174
A healthcare organization wants to predict patient readmission risk using electronic health records and continuous wearable telemetry. Predictive models must update frequently to remain accurate. Which architecture is most appropriate?
A) Pub/Sub → Dataflow → BigQuery → Vertex AI
B) Cloud Storage → App Engine → BigQuery ML
C) Dataproc → Cloud Storage → Cloud Functions
D) Bigtable → Cloud Run → AutoML Tables
Answer: A
Explanation:
Predictive healthcare analytics combines structured electronic health record data with continuous telemetry streams from wearable devices. The ingestion system must handle high-frequency events from multiple devices, such as heart rate, blood pressure, activity levels, and sleep patterns. Pub/Sub provides a fully managed, globally distributed messaging system capable of ingesting millions of events per second. It guarantees at-least-once delivery, durability, and decouples producers from downstream consumers, allowing continuous real-time data ingestion from thousands of patients globally.
Dataflow serves as the processing layer, transforming, enriching, and aggregating incoming streams. It can normalize telemetry, merge it with historical EHR records, compute derived features, and perform windowed and stateful computations. This allows detection of anomalies, calculation of rolling averages, and generation of metrics critical for predictive modeling. Dataflow’s managed service provides automatic scaling, exactly-once processing, and high availability, reducing operational complexity while ensuring data accuracy and consistency.
BigQuery acts as the analytical warehouse, storing raw and processed data for large-scale analysis. Analysts and data scientists can explore patient cohorts, perform feature engineering, and run historical trend analysis. Partitioned and clustered tables allow efficient queries over time-series data, enabling cost-effective and scalable analytics for millions of patients. BigQuery provides a platform for training predictive models, integrating seamlessly with downstream ML pipelines.
Vertex AI allows the creation, training, and deployment of predictive models for readmission risk. Continuous retraining pipelines ensure models are updated as new telemetry and EHR data arrive, maintaining high predictive accuracy. Low-latency endpoints provide real-time risk scores to clinical decision support systems. Vertex AI also supports experiment tracking, model monitoring, and drift detection, ensuring reliable performance and compliance. Alternative architectures are less suitable. Cloud Storage with App Engine and BigQuery ML cannot handle continuous streaming data or provide automated retraining pipelines. Dataproc with Cloud Storage and Cloud Functions adds operational complexity and lacks tight integration for streaming ML workflows. Bigtable with Cloud Run and AutoML Tables may store telemetry but is less suited for structured EHR integration and continuous predictive analytics.
Therefore, Pub/Sub, Dataflow, BigQuery, and Vertex AI provide the most appropriate end-to-end solution for predictive healthcare analytics, enabling accurate, continuous, and scalable readmission risk prediction.
Question 175
A global logistics company wants to monitor real-time shipment conditions, including temperature, humidity, and location. Alerts must be triggered if conditions deviate from predefined thresholds. Which architecture is most appropriate?
A) Cloud Pub/Sub → Dataflow → BigQuery → Looker
B) Cloud SQL → Cloud Functions → Dataproc → Looker
C) Cloud Storage → Cloud Run → BigQuery ML
D) Bigtable → App Engine → Data Studio
Answer: A
Explanation:
Monitoring global shipments in real time requires ingesting telemetry data from thousands of containers, trucks, or warehouses. Each shipment sends continuous data streams, such as temperature, humidity, GPS location, and vibration metrics. Cloud Pub/Sub provides a managed messaging system capable of ingesting millions of events per second globally. Its reliability ensures messages are durable and delivered at least once, preventing loss even during spikes or network issues. Pub/Sub decouples producers from consumers, allowing devices to send data without depending on downstream processing capacity.
Dataflow processes this high-volume data in real time. It can normalize incoming events, enrich them with metadata such as shipment details or location context, and detect anomalies like temperature breaches or route deviations. Windowed computations allow for aggregating metrics over defined intervals, enabling detection of trends such as prolonged exposure to unsafe conditions. Stateful processing ensures continuity and accuracy across multiple related events, while managed scaling ensures the pipeline can handle sudden traffic increases without manual intervention. Dataflow’s exactly-once processing semantics eliminate duplication, maintaining high-quality data for analytics and alerting.
BigQuery stores both raw and processed data for analytics and machine learning purposes. It allows analysts to query large datasets efficiently, performing historical analyses, trend detection, and operational reporting. Time-partitioned tables reduce query costs while providing quick access to specific time intervals. Data stored in BigQuery can also be used for predictive maintenance or shipment risk modeling.
Looker integrates with BigQuery to provide dashboards, real-time alerts, and visualizations of shipment conditions. Operations teams can monitor critical metrics, receive notifications for anomalies, and analyze trends for continuous improvement. Alternative architectures are less suitable. Cloud SQL with Cloud Functions and Dataproc introduces latency and requires manual scaling. Cloud Storage with Cloud Run and BigQuery ML is suitable for batch processing but not for real-time streaming. Bigtable with App Engine and Data Studio can store time-series data but lacks integrated streaming analytics and automated alerting. Therefore, Pub/Sub, Dataflow, BigQuery, and Looker provide the most appropriate architecture for global real-time shipment monitoring and anomaly detection.
Question 176
A fintech company wants to prevent fraudulent transactions in real time. Each transaction must be evaluated within milliseconds, and the system must remain highly available globally. Which storage solution is most appropriate?
A) Memorystore Redis
B) Cloud SQL
C) BigQuery
D) Cloud Storage
Answer: A
Explanation:
Real-time fraud detection requires that every financial transaction be evaluated almost instantaneously to prevent loss while allowing legitimate transactions to proceed without interruption. Memorystore Redis, a fully managed in-memory key-value store, provides sub-millisecond read and write operations, making it ideal for storing operational data such as user transaction histories, velocity metrics, blacklists, and precomputed risk scores. Its in-memory design ensures rapid access to critical data required for real-time decision-making.
Redis supports high concurrency and can scale horizontally to handle global traffic peaks during holidays or promotional events. Advanced data structures, such as hashes, sorted sets, and bitmaps, allow implementation of aggregation rules, thresholds, and anomaly detection logic efficiently. Memorystore Redis also provides managed replication and high availability, ensuring the system remains operational even if nodes or regions fail. Its fully managed service eliminates operational burdens such as scaling, failover, and cluster maintenance.
Alternative solutions are less suitable. Cloud SQL provides relational storage with ACID compliance but cannot meet sub-millisecond latency at global scale, especially during traffic surges. BigQuery is optimized for analytics but is unsuitable for operational, real-time transaction lookups. Cloud Storage is object-based and designed for batch or archival storage rather than rapid, low-latency retrieval.
Using Memorystore Redis ensures that fintech companies achieve a low-latency, highly available, and globally scalable system for operational fraud detection. Redis provides instant access to account histories and risk scores, enabling real-time evaluation of millions of transactions per second while maintaining reliability, performance, and operational simplicity. This makes Redis the preferred choice for preventing fraud in high-volume, time-sensitive environments.
Question 177
A healthcare organization wants to predict patient readmission risk using structured electronic health records and continuous wearable telemetry. Predictive models must update frequently to remain accurate. Which architecture is most appropriate?
A) Pub/Sub → Dataflow → BigQuery → Vertex AI
B) Cloud Storage → App Engine → BigQuery ML
C) Dataproc → Cloud Storage → Cloud Functions
D) Bigtable → Cloud Run → AutoML Tables
Answer: A
Explanation:
Healthcare predictive analytics involves combining structured electronic health records with high-frequency telemetry from wearable devices. Data ingestion must be reliable, scalable, and capable of handling continuous streams such as heart rate, blood pressure, activity levels, and sleep patterns. Pub/Sub provides a managed, globally distributed messaging system capable of ingesting millions of events per second, guaranteeing at-least-once delivery, durability, and decoupling of producers from consumers. This ensures continuous real-time data ingestion from thousands of patients worldwide.
Dataflow processes these streams in near real time. It normalizes telemetry data, merges it with historical EHR records, aggregates metrics, and computes derived features for predictive modeling. Windowed and stateful processing allow calculation of rolling averages, detection of anomalies, and tracking of temporal health trends. Dataflow’s managed service ensures exactly-once processing semantics, high availability, and automatic scaling, reducing operational complexity while maintaining data accuracy and integrity.
BigQuery serves as the analytical warehouse, storing both raw and processed data for large-scale querying. Analysts and data scientists can perform cohort analysis, feature engineering, and historical trend exploration. Partitioned and clustered tables allow cost-efficient queries over time-series data without compromising performance. BigQuery supports integration with downstream machine learning workflows, providing features for model training and evaluation.
Vertex AI enables creation, training, and deployment of predictive models for readmission risk. Continuous retraining pipelines ensure models remain accurate as new telemetry and EHR data arrive. Vertex AI also supports low-latency prediction endpoints for clinical decision support, drift monitoring, and experiment tracking, ensuring reliability. Alternative architectures are less effective. Cloud Storage with App Engine and BigQuery ML cannot handle continuous streaming or automated retraining efficiently. Dataproc with Cloud Storage and Cloud Functions adds operational overhead and lacks integration for streaming-to-ML workflows. Bigtable with Cloud Run and AutoML Tables can store telemetry but is less suited for combining structured EHR data with real-time predictive modeling.
Therefore, Pub/Sub, Dataflow, BigQuery, and Vertex AI provide the optimal end-to-end architecture for predictive healthcare analytics, continuous retraining, and accurate readmission risk prediction.
Question 178
A global food delivery company wants to track real-time orders, delivery status, and driver locations. The system must generate alerts if delivery times exceed thresholds and provide analytics for operational improvement. Which architecture is most appropriate?
A) Cloud Pub/Sub → Dataflow → BigQuery → Looker
B) Cloud SQL → Cloud Functions → Dataproc → Looker
C) Cloud Storage → Cloud Run → BigQuery ML
D) Bigtable → App Engine → Data Studio
Answer: A
Explanation:
Real-time tracking for food delivery involves capturing continuous streams of data from multiple sources: order placement events, driver GPS telemetry, restaurant preparation status, and customer interactions. The ingestion system must handle millions of events per second, maintain durability, and scale automatically as demand fluctuates globally. Cloud Pub/Sub provides a fully managed, globally distributed messaging service designed for this type of high-throughput real-time ingestion. It guarantees at-least-once delivery, ensuring no critical event is lost, while decoupling producers from consumers, allowing devices and services to continue publishing data even during downstream processing delays or failures.
Dataflow serves as the real-time processing layer. It normalizes and enriches incoming events, merges driver telemetry with order information, and applies business logic such as delivery time calculations, route deviations, and anomaly detection. Windowed computations allow aggregating metrics over intervals, such as average delivery time per city or region, while stateful processing enables the system to detect persistent delays or operational patterns that require attention. Dataflow’s managed service automatically scales to handle sudden surges in order volume, and its exactly-once processing guarantees data integrity without duplications, which is critical for accurate analytics and alerting.
BigQuery acts as the analytics and storage layer, capturing both raw and processed streams. Analysts can query historical and real-time data to identify trends, optimize routing, evaluate restaurant or driver performance, and improve customer satisfaction. Partitioned tables allow efficient time-based queries, reducing cost while providing fast access to large datasets. BigQuery also integrates seamlessly with machine learning pipelines, enabling predictive modeling for factors such as estimated delivery times or dynamic resource allocation.
Looker provides visualization, alerting, and operational insights. Operations teams can monitor delivery performance in real time, identify bottlenecks, and set thresholds for automated alerts when deliveries exceed acceptable times. Dashboards enable managers to understand regional trends, optimize driver allocation, and improve service reliability.
Alternative architectures are less effective. Cloud SQL with Cloud Functions and Dataproc introduces latency, is harder to scale, and is unsuitable for high-throughput real-time ingestion. Cloud Storage with Cloud Run and BigQuery ML is better suited for batch analytics and predictive modeling but cannot handle live streams or real-time alerting efficiently. Bigtable with App Engine and Data Studio can store high-volume time-series data but lacks integrated streaming processing and real-time alert capabilities. Therefore, the combination of Pub/Sub, Dataflow, BigQuery, and Looker provides the most scalable, low-latency, end-to-end solution for global food delivery monitoring and analytics.
Question 179
A global fintech company must evaluate transactions in milliseconds to prevent fraud. The system must remain highly available across multiple regions and handle sudden traffic spikes. Which storage solution is most appropriate?
A) Memorystore Redis
B) Cloud SQL
C) BigQuery
D) Cloud Storage
Answer: A
Explanation:
Real-time fraud detection is a mission-critical operational workflow where every transaction must be evaluated instantly. Sub-millisecond latency is required to approve legitimate transactions without delay while preventing fraudulent activity. Memorystore Redis is a fully managed, in-memory key-value database designed for such use cases. It allows rapid access to account histories, transaction velocity metrics, precomputed risk scores, and other operational data, providing the speed necessary for instant decision-making.
Redis supports extreme concurrency and can scale globally to handle millions of transactions per second. Its advanced data structures, including sorted sets, hashes, and bitmaps, allow efficient calculation of thresholds, aggregation of multiple risk factors, and execution of complex rules needed for fraud evaluation. Memorystore provides managed replication and high availability, ensuring continuous operation even if nodes or entire regions fail. Fully managed operations reduce administrative overhead, eliminating the need to manually handle scaling, failover, and cluster maintenance.
Alternative solutions are unsuitable. Cloud SQL provides ACID transactions but cannot deliver the required sub-millisecond response under high concurrent global load. BigQuery is optimized for analytical queries and cannot provide real-time operational lookups at transaction speed. Cloud Storage is object-based, intended for batch workloads and archival, making it unsuitable for operational fraud evaluation.
Using Memorystore Redis allows fintech companies to achieve a highly available, low-latency, globally scalable operational database for real-time fraud detection. Redis ensures rapid access to critical operational data, enabling the evaluation of millions of transactions per second while maintaining performance, reliability, and operational simplicity. This architecture provides the speed, scalability, and durability required for modern financial systems to prevent fraud effectively.
Question 180
A healthcare organization wants to predict patient readmission risk using structured EHR data and continuous wearable telemetry. The predictive models must update frequently as new data arrives. Which architecture is most appropriate?
A) Pub/Sub → Dataflow → BigQuery → Vertex AI
B) Cloud Storage → App Engine → BigQuery ML
C) Dataproc → Cloud Storage → Cloud Functions
D) Bigtable → Cloud Run → AutoML Tables
Answer: A
Explanation:
Predictive healthcare analytics involves integrating structured electronic health record data with continuous telemetry streams from wearable devices. Telemetry data can include heart rate, blood pressure, activity, and sleep patterns. The ingestion layer must handle high-frequency events reliably and at scale. Pub/Sub provides a managed, globally distributed messaging system capable of ingesting millions of events per second. It guarantees at least once delivery, durability, and decouples producers from downstream consumers, allowing devices to continuously send data without depending on processing availability.
Dataflow processes the continuous streams in near real time. It normalizes telemetry, merges it with historical EHR records, calculates derived features, and performs windowed and stateful computations. These capabilities allow the system to detect anomalies, compute rolling averages, and capture trends critical for predictive modeling. Dataflow’s managed service provides automatic scaling, exactly-once processing, and high availability, reducing operational complexity while ensuring data accuracy and reliability.
BigQuery is a fully managed, serverless data warehouse designed to handle large-scale analytics efficiently, making it an ideal platform for storing both raw and processed datasets. In the context of data-driven workflows, BigQuery enables organizations to centralize their data, allowing analysts and data scientists to perform a wide range of operations such as cohort analysis, feature engineering, and historical trend evaluation. By storing raw data, organizations maintain a complete record of events or transactions, which can be useful for auditing, retrospective studies, or feeding machine learning models. Processed data, on the other hand, supports immediate analysis and reporting, providing a clean and structured view of information for decision-making.
One of BigQuery’s key strengths is its ability to handle time-series data efficiently. Partitioned tables allow data to be organized by date or other relevant criteria, enabling queries to scan only the relevant partitions rather than the entire dataset, reducing both latency and cost. Clustered tables further optimize query performance by physically organizing data based on frequently queried columns, allowing analytics tasks such as aggregations, filtering, and joins to execute more efficiently. This is particularly useful for analyzing large datasets that span months or years, as it ensures queries remain cost-effective even when dealing with high volumes of historical data.
BigQuery also integrates seamlessly with machine learning workflows, enabling predictive modeling directly on the warehouse data using BigQuery ML. Analysts can train and evaluate models on structured datasets without moving the data to a separate platform, streamlining the pipeline from data ingestion to predictive insights. This integration simplifies the workflow for tasks like customer behavior prediction, recommendation systems, or risk assessment, providing both analytical flexibility and operational efficiency. By combining storage, querying, and machine learning capabilities, BigQuery serves as a central hub for end-to-end data analytics, supporting both historical analysis and advanced predictive modeling in a scalable and cost-effective manner.
Vertex AI allows the creation, training, and deployment of predictive models for readmission risk. Continuous retraining pipelines ensure models remain accurate as new telemetry and EHR data arrive. Low-latency endpoints provide predictions to clinical decision support systems. Vertex AI supports monitoring, drift detection, and experiment tracking. Alternative architectures are less suitable. Cloud Storage with App Engine and BigQuery ML cannot efficiently process continuous streams or handle automated retraining. Dataproc with Cloud Storage and Cloud Functions adds operational overhead and lacks streaming-to-ML integration. Bigtable with Cloud Run and AutoML Tables may store telemetry, but it’s less suited for combining structured EHR data with continuous predictive modeling.
Therefore, Pub/Sub, Dataflow, BigQuery, and Vertex AI provide the optimal end-to-end solution for predictive healthcare analytics, continuous model retraining, and accurate readmission risk prediction.