Luyện thi GCP Data Engineer

BigQuery is a serverless warehouse for analytics. Cloud Storage stores objects, Pub/Sub handles messages, and DNS resolves names.

Cloud Storage stores objects such as files and backups. Databases, queues, and load balancers are different.

Pub/Sub provides asynchronous messaging. Warehouses, VMs, and DNS are different.

Dataflow processes batch and streaming data. Databases, object stores, and CDNs are different.

Dataproc runs managed Spark and Hadoop clusters. Warehouses, messaging, and DNS are different.

Bigtable is a low-latency NoSQL database for large workloads. Relational databases, object stores, and CDNs are different.

Cloud SQL provides managed relational databases. NoSQL, object storage, and messaging are different.

Firestore is a document NoSQL database. Relational databases, warehouses, and CDNs are different.

Spanner provides global relational consistency. Document databases, object stores, and queues are different.

BigQuery is a warehouse, Cloud Storage stores objects, Pub/Sub delivers messages, and Bigtable is low-latency NoSQL.

Dataflow builds ETL pipelines. DNS, CDN, and IAM are different.

Cloud Composer orchestrates workflows with Airflow. Warehouses, queues, and load balancers are different.

Ephemeral clusters exist only for a job to reduce cost. Always-running clusters, warehouses, and DNS zones are different.

Cloud Scheduler runs jobs on schedules. Databases, storage, and firewalls are different.

Cloud Functions run code in response to events. VMs, databases, and DNS records are different.

Push subscriptions send messages to endpoints. Pull subscriptions, databases, and DNS records are different.

Cloud Storage events such as object creation can trigger functions. DNS, VM reboots, and firewall changes are different.

Templates make Dataflow pipelines reusable. VMs, buckets, and DNS records are different.

Load jobs import data into BigQuery. Deletion, DNS, and backups are different.

Dataflow processes ETL, Composer orchestrates, Scheduler runs jobs, and Pub/Sub streams events.

BigQuery ML builds models with SQL. Notebooks, warehouses, and DNS are different.

Vertex AI provides end-to-end ML tools. Messaging, databases, and CDNs are different.

Looker Studio creates dashboards. Warehouses, queues, and load balancers are different.

Notebooks provide coding environments for ML. Databases, buckets, and DNS records are different.

Evaluation sets test generalization. Storage cost, encryption, and reports are different.

Batch prediction processes many records offline. Real-time prediction, storage, and DNS are different.

Feature stores centralize ML features. Transaction databases, DNS, and load balancers are different.

Model registries manage versions and metadata. Buckets, firewalls, and DNS zones are different.

Tuning searches hyperparameters. Deletion, encryption, and reports are different.

BigQuery ML uses SQL, Vertex AI is an ML platform, Looker visualizes, and notebooks provide code.

IAM controls access. Storage, CDN, and DNS are different.

KMS manages encryption keys. Databases, firewalls, and load balancers are different.

DLP discovers and protects sensitive data. Warehouses, queues, and DNS are different.

VPC-SC creates perimeters. Firewalls, DNS zones, and load balancers are different.

Customer-managed keys are controlled by the customer. Google-managed keys, DNS, and firewall rules are different.

Retention policies control data lifecycle. Firewall rules, DNS records, and load balancers are different.

Audit logs record admin and data access. Application logs, DNS records, and buckets are different.

Organization policies enforce constraints. Firewalls, DNS zones, and load balancers are different.

Private connectivity keeps traffic internal. Public exposure, higher latency, and disabled encryption are not security best practices.

IAM grants permissions, KMS manages keys, DLP protects sensitive data, and VPC-SC creates perimeters.

Cloud Monitoring collects metrics and alerts. Warehouses, DNS, and load balancers are different.

Cloud Logging stores logs. Databases, firewalls, and CDNs are different.

Quotas limit resource usage. Models, datasets, and DNS records are different.

Lifecycle rules manage object retention. Firewall rules, DNS records, and load balancers are different.

Budget alerts monitor spend. Firewall rules, DNS records, and load balancers are different.

Regions are geographic locations. Buckets, models, and DNS records are resources or data.

Retries handle transient failures. Deletion, disabled encryption, and latency are not retry purposes.

Idempotent operations are safe to repeat. Deletion, encryption, and DNS are different.

Backups support recovery. Increasing cost, disabling security, and slowing queries are not benefits.

Monitoring collects metrics, Logging stores logs, quotas limit usage, and lifecycle controls retention.

Batch processing runs on schedules. Real-time, DNS, and storage are different.

Streaming processes events as they arrive. Nightly, backups, and DNS are different.

Partitioning improves query efficiency. Deletion, encryption, and reports are different.

Clustering organizes related rows for better queries. Random splits, DNS, and firewalls are different.

Schema evolution handles changing fields. Deletion, disabled security, and reports are different.

Data lakes store raw data. Structured warehouses, DNS, and load balancers are different.

Star schemas use facts and dimensions. Backups, firewall rules, and DNS records are different.

Governance defines data policies. Buckets, firewalls, and load balancers are resources.

Data Catalog helps discover metadata. Databases, queues, and DNS zones are different.

Batch runs on schedules, streaming processes in real time, partitioning improves queries, and Data Catalog discovers metadata.