Google Cloud Digital Leader Course Notes

This section connects cloud adoption to business outcomes. Elasticity lets organizations add or remove resources as demand changes instead of paying for fixed peak capacity. IaaS, PaaS, and SaaS shift different amounts of infrastructure and platform responsibility to the provider, so the right choice depends on the control, expertise, and operational effort the organization wants to retain. Cloud transformation also requires organizational readiness, including skills, collaboration, governance, and a practical adoption plan. Total cost of ownership should include acquisition, facilities, staffing, maintenance, operations, and lifecycle costs rather than only purchase price. Cloud consumption can shift spending from capital expenditure toward operating expenditure. Regional design improves resilience, while open standards can improve portability and reduce unnecessary lock-in.

• Confusing IaaS, PaaS, and SaaS responsibility boundaries.

• When a scenario describes unpredictable or seasonal demand, look for elasticity and dynamic scaling.

This section focuses on selecting data services according to workload. BigQuery is a fully managed, serverless data warehouse for large-scale SQL analytics. A data lake preserves raw or full-fidelity data for later analytics and AI use, while a data warehouse organizes curated data for analysis and reporting. Cloud SQL is a managed relational database for operational applications that need familiar engines such as MySQL, PostgreSQL, or SQL Server. Dataflow supports unified batch and streaming pipelines, and Spanner supports globally scalable relational transactions with strong consistency. Data governance establishes ownership, quality, security, access, and consistent definitions so data can be trusted and used responsibly. Cloud Storage classes—Standard, Nearline, Coldline, and Archive—should be selected according to access frequency, retention needs, and retrieval economics.

• Confusing BigQuery with Cloud SQL: BigQuery is optimized for analytics, while Cloud SQL supports operational applications.

• First decide whether the scenario is transactional, analytical, streaming, or archival.

This section focuses on choosing Google Cloud AI capabilities according to the business problem and the amount of customization required. Gemini Enterprise Agent Platform is Google Cloud’s managed platform for building, deploying, governing, and operating Gemini-powered agents and model-based solutions. AutoML reduces the expertise and manual configuration needed to train supported model types, while custom training gives teams control over code, frameworks, and model architecture. Pretrained APIs are appropriate for common capabilities such as sentiment analysis, image analysis, translation, speech recognition, and speech synthesis when custom model development is unnecessary. BigQuery ML lets data teams create and run supported models with SQL close to data already stored in BigQuery. Responsible AI requires attention to data quality, fairness, explainability, privacy, human oversight, and ongoing monitoring, especially for consequential decisions.

• Confusing BigQuery ML with Gemini Enterprise Agent Platform: BigQuery ML is SQL-based model work in BigQuery, while Agent Platform supports broader AI and agent lifecycles.

• Choose a pretrained API when the requested capability is common and no custom training is required.

This section focuses on matching modernization approaches and Google Cloud services to application requirements, deadlines, and operational capabilities. Cloud Run is a fully managed platform for stateless containers and request- or event-driven workloads that should scale without cluster administration. Rehosting moves a workload with minimal change and is often appropriate when speed is the primary constraint. Replatforming makes limited platform changes to gain managed-service benefits without a full redesign. Refactoring changes code or architecture to take greater advantage of cloud-native capabilities. GKE is appropriate when Kubernetes compatibility, cluster-level control, or advanced orchestration is required and the team can accept greater operational complexity. Apigee provides API management capabilities such as security policies, analytics, traffic management, and developer enablement for internal or partner ecosystems.

• Confusing GKE with Cloud Run: GKE exposes Kubernetes control, while Cloud Run abstracts cluster administration.

• Use rehosting when speed and minimal change dominate the scenario.

This section covers shared responsibility, core security principles, identity and access management, data location and legal control, security operations, and compliance. Google secures the cloud infrastructure, while customers remain responsible for areas such as their data, identities, applications, access policies, and configurations, with the exact boundary varying by service model. Confidentiality, integrity, and availability are foundational goals. Authentication verifies identity, authorization determines permitted actions, and auditing records activity for accountability. Data residency describes where data is stored, while data sovereignty concerns the laws and governmental authority that may apply to data. IAM policies bind principals to roles on resources, and inherited policies can grant access from higher levels of the resource hierarchy. Compliance reports provide evidence about provider controls but do not transfer the customer’s accountability for its workload.

• Confusing data residency with data sovereignty: residency is location, while sovereignty is legal authority and jurisdiction.

• Separate authentication, authorization, and auditing when evaluating an identity scenario.

This section focuses on financial governance, reliability, operational practices, support, resource hierarchy, and sustainability. Billing reports and cost attribution provide visibility into spending. Budgets and threshold rules send notifications but do not automatically stop usage or cap charges. Quotas constrain consumption of specific resources and can protect environments from uncontrolled growth, although quota behavior and adjustment depend on the service. Reliability planning uses redundancy, fault tolerance, high availability, tested disaster recovery, monitoring, incident response, capacity planning, and change management. DevOps improves delivery by aligning development and operations, while site reliability engineering applies software-engineering practices to operational reliability. Effective support cases include impact, timestamps, affected resources, troubleshooting evidence, and clear reproduction details. Carbon reporting provides environmental information that is distinct from financial cost reporting.

• Confusing quotas with budget alerts: quotas constrain resource consumption, while budget alerts report cost status.

• Choose billing reports for cost analysis and budgets for cost notifications.