CDL question patterns

Rehost (lift and shift) to Compute Engine.

Replatform and refactor need more redesign and time.

Open source and open standards, with multicloud-capable products such as GKE Enterprise and BigQuery Omni.

Openness and interoperability are Google Cloud differentiators and keep data and code portable.

Serverless: Cloud Run, Cloud Run functions, or BigQuery for analytics.

The provider handles infrastructure and scaling, including down to zero.

A prebuilt agent or pre-trained API.

Ready-made options are fastest and need the least skill. Custom models are for differentiation.

Cloud elasticity: autoscaling and load balancing.

Capacity grows and shrinks with demand, so you don't buy for the peak.

Hybrid cloud.

Hybrid joins on-premises or private infrastructure with public cloud.

Multicloud.

Multicloud means more than one public cloud.

Deploy in a nearer region and use edge caching (Cloud CDN).

Latency is distance and hops. Regions and edge locations shorten the path.

Spread across multiple zones.

Zones are isolated failure domains within a region.

Deploy across multiple regions.

A multi-zone design stays in one region.

PaaS.

PaaS hides infrastructure and runtime; IaaS leaves the OS to you.

IaaS (Compute Engine).

IaaS leaves OS and software to the customer.

SaaS.

The provider runs everything.

BigQuery.

Serverless data warehouse for analytics.

Cloud Storage.

Object storage for unstructured data and lakes.

Cloud SQL.

Managed relational database for typical workloads.

Spanner.

Horizontally scalable relational with global availability.

Bigtable.

Wide-column NoSQL for huge throughput.

Firestore.

Document database with real-time and offline support.

AlloyDB.

PostgreSQL-compatible with enterprise performance.

Archive class (or Coldline if read quarterly).

Colder classes cost less to store and more to read.

Autoclass.

Moves objects between classes automatically by access.

Looker.

Semantic layer gives consistent metrics and queries live data.

Pub/Sub.

Scalable messaging that decouples producers and consumers.

Dataflow.

Apache Beam based, autoscaling.

Managed Service for Apache Spark.

Managed open source engines (formerly Dataproc).

Database Migration Service.

Continuously replicates changes to the target.

BigQuery ML.

Builds and runs models with SQL where the data lives.

Cloud Translation API or Speech-to-Text API.

Pre-trained APIs need no data or ML expertise.

Gemini through Agent Platform.

Gemini is multimodal.

AutoML on Agent Platform.

No-code training on your own data.

Tune and ground a foundation model in Agent Studio.

Tuning and grounding customise without building from scratch.

AI Hypercomputer: TPUs, GPUs, open software, flexible consumption.

Integrated hardware and software built for AI.

Ground it in trusted data and keep human oversight.

Grounding reduces hallucination.

Fix data quality first.

Completeness, uniqueness, timeliness, validity, accuracy and consistency drive AI quality.

GKE.

Managed Kubernetes.

Cloud Run.

Serverless containers scale to zero.

Cloud Run functions.

Event-driven functions.

Spot VMs.

Heavily discounted but can be reclaimed.

GKE Enterprise.

Fleet management across environments.

BigQuery Omni.

Runs BigQuery analytics across clouds.

Apigee.

API management with security, analytics and monetization.

Zero trust (BeyondCorp) with Identity-Aware Proxy.

Every request is authenticated and authorized regardless of network.

Confidential Computing.

Protects data in use.

Cloud Armor.

WAF and DDoS protection.

Security Command Center.

Posture and risk platform.

Google Security Operations.

Unified SIEM and SOAR with Gemini.

Model Armor.

Screens prompts and outputs.

Sensitive Data Protection.

Google's data loss prevention service.

Cloud Interconnect.

VPN is encrypted over the internet; Interconnect is private.

Data residency, sovereignty controls and Assured Workloads.

Choose where data lives and who operates it.

CapEx to OpEx shift.

Cloud is pay as you go.

Set it on the folder.

Policies inherit downward through the hierarchy.

Billing budget with threshold alerts.

Budgets notify and don't cap.

Resource quotas.

Quotas limit consumption.

Cloud Trace.

Distributed tracing shows latency per hop.

Cloud Profiler.

Continuous profiling.

SLI, SLO and SLA.

Measurement, internal goal, customer contract.