Google Cloud Digital Leader cheat sheet

TCO must include migration, integration, egress, training and dual-running costs.

Choose migration approaches per workload after checking dependencies, support, operating burden and business value.

An SLA defines measurable levels and remedies such as service credits, not a guarantee.

Exit strategy and data portability should be agreed before, not after, signing.

Provider selection weighs certification, data location, roadmap and support alongside price.

Select VM, container and managed execution models by workload and retained responsibility.

The certification guide’s product names are versioned study references, not a guarantee of current packaging.

APIs, networking and load distribution still need access, data and operating controls.

Hybrid and multicloud claims require evidence about policy, portability and failure behavior.

Security of the cloud belongs to the provider; security in the cloud belongs to the consumer.

The responsibility boundary shifts with IaaS, PaaS and SaaS.

ISO/IEC 27017 covers cloud-specific controls; ISO/IEC 27018 covers PII in public clouds.

Verify actual legal roles, data locations, contractual duties and the scope of assurance evidence.

A provider's certification does not make the consumer's configuration compliant.

RTO is the maximum acceptable outage; RPO is the maximum acceptable data loss in time.

RPO drives backup and replication frequency; RTO drives standby and failover design.

Showback reports consumption to teams; chargeback actually bills it to them.

Incident management restores service; problem management removes the root cause.

Monitor end-to-end service levels and spend, not only infrastructure health.

Start with the business question and the required quality and freshness.

Operational storage, object storage and analytical workloads have different access patterns.

Check keys, join multiplicity, lineage and reconciled totals.

Managed services do not remove ownership and access responsibilities.

Define the task, error costs, users and evaluation before selecting a model.

Prevent leakage and separate training, validation and testing.

Precision and recall answer different questions; accuracy can hide imbalance.

Product names and platform features need version-aware verification.