Prepare and test your skills
Prepare and test your skills
Stakeholder management is the practice of aligning technical and non-technical owners around cloud decisions through shared facts, clear communication, and facilitated agreement. Product owners, developers, administrators, and financial analysts weigh cost, speed, risk, and performance differently, so the architect must make tradeoffs explicit. Consensus holds when everyone works from the same trusted data about cost and business value.
Building consensus with shared cost facts means giving every role the same view of what cloud resources cost and what business value they create. Individual choices by any developer or administrator can change total spend, so teams need projected costs at design time, real-time usage costs during operation, and costs mapped to key performance indicators (KPIs). Mapping spend to KPIs lets the group ask whether higher spending produces enough business value to justify slower delivery or higher reliability costs. Organization-wide visibility is supported by one consistent calculation method, standardized tools such as the Google Cloud Billing dashboard, cost allocation to teams, and open discussion of cost during design.
Exam tip: When raw cost data cannot be shared, share aggregated metrics such as total cost per environment or relative metrics such as average cost per transaction or user.
Defining platform roles and communication flows means naming who owns strategy, delivery, security, and assurance in a distributed platform such as a data mesh (opens in a new tab). The Data platform product owner owns strategy and roadmap, the Data platform engineer owns reusable delivery, the Platform and security engineer owns enterprise fit for networking and security, and the Enterprise architect owns alignment to enterprise strategy through design authority. Influence flows from leadership and domain owners into the product owner for roadmap decisions, from the architect outward for assurance, and from central engineers to producers and consumers through shared templates and services. Tailored updates match each role: roadmap and satisfaction data for owners, templates and uptime for builders, component health for platform engineers, and design decisions for governance.
| Stakeholder | Core interest | Communication focus |
|---|---|---|
| Data platform product owner | Ecosystem of successful data products | Strategy, roadmap, time to production, satisfaction |
| Data platform engineer | Reusable self-service infrastructure | Templates, releases, uptime, support responsiveness |
| Platform and security engineer | Healthy core technology | Underlying uptime and support responsiveness |
| Enterprise architect | Enterprise alignment | Design authority and best-practice assurance |
Shared signals from central services keep these messages consistent, including data sharing status, pipeline orchestration, logging and monitoring, governance controls, continuous integration and continuous deployment (CI/CD) guardrails, security reporting, and FinOps reporting.
Leading workshops within technical limits means facilitating sessions where business needs and technical reality are agreed together. The Data platform product owner brings leadership, domain owners, and governance requirements into the room, while engineers and architects translate them into reusable services for ingestion, storage, processing, sharing, logging, monitoring, security, and cost reporting. Central capabilities agreed in the workshop become self-service templates for Terraform, pipelines, containers, and orchestration, so guardrails flow directly into what teams can deploy. Designs are bounded by shared responsibility and location rules: the provider always owns the underlying network and infrastructure, the customer always owns access policies and data, and rules such as data residency can require data to stay in specific Google Cloud regions.
Continue with this study guide: Stakeholder management (e.g., influencing and facilitation)
Change management is the system of culture, pipelines, and controls that lets teams release often without causing long outages. Site Reliability Engineering (SRE) balances feature velocity with availability, latency, performance, and capacity, while DevOps connects development and operations through shared tooling. Together they allow necessary change while limiting blast radius and recovery time.
Balancing speed and stability with SRE and DevOps means breaking down silos and testing before every release. Development teams want fast features and operations teams want stability, so both sides use the same deployment pipeline, automation, runbooks, and reviews. Applications are designed for testability so unit tests carry most of the load, with integration and system tests checking larger interactions, and infrastructure as code creates disposable test environments with tools such as Terraform. Gradual deployment (opens in a new tab) then exposes a new version to larger audiences step by step, with a defined rollback if behavior is wrong. Failures are treated as normal in large systems, so structured incident roles, communication paths, playbooks, training, and blameless postmortems reduce impact and prevent repeat events.
Releasing changes gradually with validation and rollback means only verified artifacts reach production in small steps. Source code flows from Cloud Storage, Cloud Source Repositories, GitHub, or Bitbucket into Cloud Build for build and test, into Artifact Registry for versioned storage, and into Cloud Deploy for ordered promotion across development, test, and production. Automated unit, integration, end-to-end, performance, and load tests act as gates between stages, so higher coverage means more changes can ship before reliability drops. Migration plans keep rollback ready by dual-writing artifacts to source and Artifact Registry, shifting a small subset first, then cutting over only after validation, with a maximum allowed time and manual backup procedure for each step.
| Service | Role in the release flow | What it controls |
|---|---|---|
| Cloud Build | Builds from source | Build execution and test validation |
| Artifact Registry | Stores versioned artifacts | Lifecycle, access, scanning, and rollback history |
| Cloud Deploy | Delivers to environments | Ordered promotion across targets |
Security is embedded in the same pipeline: Artifact Analysis scans images and monitors for new vulnerabilities, Web Security Scanner checks web apps, and Binary Authorization blocks deployment to Google Kubernetes Engine (GKE) or Cloud Run unless images carry required attestations for trusted build, vulnerability policy, and promotion.
Controlling risk with policy and audit trails means preventing risky changes before they happen and recording what changed afterward. Identity and Access Management (IAM) enforces separation of duties and least privilege, such as separating Cloud Key Management Service (Cloud KMS) use from administration and restricting service accounts and API keys. Organization Policy adds centralized guardrails that block default networks, restrict Cloud Run ingress and Cloud SQL authorized networks, and constrain other risky settings. Infrastructure changes roll out progressively in order:
Exam tip: Treat a change to a global Virtual Private Cloud (VPC) network or global load balancer forwarding rule as high-risk requiring extra approval, while a zonal virtual machine change can roll out progressively.
Cloud Audit Logs record who did what, where, and when, with log metric alerts for governance changes such as IAM, firewall, route, storage, and database edits. The Google Cloud Service Health Dashboard and Personalized Service Health show whether disruption came from Google Cloud or from a tracked customer change.
Continue with this study guide: Change management
Team assessment and skills readiness is the process of defining needed cloud roles, measuring current ability, and closing gaps before go-live and day-2 operations. The Workforce focus area covers roles, skills, learning programs, team structure, and hiring, and it maps to organization, folders, and projects so ownership matches billing and permission boundaries. Readiness is sustained after migration through learning and retrospectives.
Defining roles and closing skill gaps means comparing required capabilities against what staff can do today in Google Cloud. Key roles typically include Cloud Architects, Data Engineers, and Security Specialists, plus operators for deployment, observability, and incident response. The assessment checks whether staff can design and test workloads, generate operating system and container images, store them in Artifact Registry, scan and sign code, deploy to Google Cloud targets, and inject runtime configuration such as secrets and keys. Gaps also appear in IAM, encryption, vulnerability handling, compliance, and operating across regions with residency limits. Teams close gaps with training, proofs of concept on real workload patterns, dedicated refactoring of deployment processes, and hiring for specialized needs based on time, budget, people, and transfer bandwidth.
Securing the artifact supply chain means ensuring only trusted, scanned code moves from source packages to production images. Artifact Registry manages container images and language packages, Artifact Analysis provides vulnerability scanning and Software Bill of Materials (SBOM) generation, and Assured Open Source Software (Assured OSS) provides validated open-source packages with provenance. Ready teams enforce integrity controls such as cleanup policies for stale images, virtual repositories that prefer internal sources to prevent dependency confusion, immutable tags or deployment by SHA-256 digest, package managers pointed only at internal repositories, and Cloud Build quality gates that verify signatures and provenance. Software vulnerabilities (opens in a new tab) move through intake, assignment, remediation, and verification with severity tracking and periodic scans after major changes.
Exam tip: Deploy by immutable SHA-256 digest such as @sha256:... rather than by mutable tag alone, because a tag can be moved while the digest always identifies one exact image.
Enforcing trusted deployment and learning culture means combining a final deployment gate with continuous skill growth. Binary Authorization verifies container integrity at the cluster level and allows or denies deployment based on attestations, with progressive gates toward production, continuous validation of running GKE Pods, and monitored breakglass bypass for emergencies. Culture sustains these controls through psychological safety, blameless learning, conferences and training such as Google Cloud Skills Boost, and regular retrospectives. The Start-Stop-Continue model keeps reviews actionable: Start names new practices to adopt, Stop names ineffective practices to remove, and Continue names effective practices to reinforce. Culture builds skills in the Workforce focus area, while retrospectives change processes in the Processes and governance focus areas.
Continue with this study guide: Team assessment/skills readiness
Decision-making processes connect architecture choices to measured risk, cost, performance, and governance requirements. Migration Center and fit assessment surface dependencies and modernization fit, while Cloud Monitoring, BigQuery, and Cloud Billing turn operations and spend into KPIs, dashboards, and cost-benefit analysis. Decisions hold when business drivers, technical drivers, and controls are weighed together.
Assessing risk and defining success means finding what could go wrong and agreeing how success will be measured. Migration risk covers outage impact on operations, team experience with public cloud, and legal limits such as data sovereignty, while technical difficulty covers application size and age, dependency count, and third-party license or operating system limits. Teams reduce risk by starting with a non-critical but typical workload with few dependencies to create a blueprint for later waves. Success is defined with KPIs for overall performance, metrics for specific data points such as CPU or latency, service-level agreements (SLAs) across hybrid components, and dashboards that centralize CPU, memory, network, and latency. Capacity planning uses historical use, growth projections, SLA targets, and allowances for spikes and failover between primary and disaster recovery systems.
Turning spend data into architecture choices means converting raw usage into contextual decisions about total cost of ownership. Billing data flows from resources into invoices and BigQuery exports, is structured by Cloud Billing reports and FinOps Hub, is enriched with labels and hierarchy, and is then used by product teams to forecast and optimize. The data-information-knowledge-wisdom sequence moves from raw hours and cost to organized spend by service, to business context such as team and purpose, to a value decision such as whether a pipeline should be re-architected. Cloud Billing budgets and alerts notify on overruns, while committed use discounts (CUDs), sustained use discounts, and Spot VMs lower rates, and Active Assist finds idle resources. Business drivers such as differentiation, savings, resilience, and faster releases are balanced against technical drivers such as analytics, automation, serverless scale, and global reach, since single-cloud can reduce complexity while multicloud can add management and integration cost.
Exam tip: Metrics are the raw data points like CPU or latency, while a KPI is the quantifiable measurement of overall performance calculated from those metrics.
Governing decisions and managing tradeoffs means enforcing who decides, who implements, and what evidence proves compliance. The customer defines workload controls for compliance and risk, configures Organization Policy Service (opens in a new tab) constraints, IAM, hierarchy, and logging, and keeps regulated data in compliant folders, while Google provides default encryption, infrastructure controls, and policy enforcement. Responsibility shifts by service model: infrastructure as a service (IaaS) such as Compute Engine leaves most security with the customer, platform as a service (PaaS) such as App Engine and GKE shares application controls, software as a service (SaaS) such as Google Workspace leaves most with Google, and function as a service (FaaS) such as Cloud Run functions leaves code behavior with the customer. Availability tradeoffs follow math: chained availability multiplies tier values, so four 99.9% tiers yield about 99.6%, while parallel availability uses 1 minus the product of failure rates, so two 99% services in parallel yield 99.99%. High availability uses synchronous replication across zones in one region, disaster recovery typically uses asynchronous replication across regions, and managed services plus infrastructure as code reduce long-term debt and lock-in risk.
Continue with this study guide: Decision-making processes
Customer success management is the practice of proving that Google Cloud delivers business value and stays healthy through adoption, governance, and continuous improvement. Cloud Monitoring collects real-time metrics, BigQuery stores and analyzes them over time, and dashboards expose barriers such as bottlenecks, idle capacity, or missed targets. Acting on those signals reduces churn and guides cost and performance fixes.
Tracking adoption and observability means measuring system health and business outcomes from the same data pipeline. Data flows from workloads to Cloud Monitoring and Cloud Logging, then to BigQuery and dashboards for trends and decisions, with logs routable through sinks to Cloud Storage, Pub/Sub, or BigQuery for retention. Adoption tracking combines resource signals such as CPU, memory, network, disk, and response time, experience signals such as latency, error rate, throughput, and satisfaction, business signals such as revenue and engagement, and delivery signals such as deployment frequency, lead time, change failure rate, and time to restore. Error budgets define acceptable disruption, alerts notify on anomalies without causing fatigue, and traffic and load (opens in a new tab) analysis such as round-trip time reveals bottlenecks. Health review follows Measure, Analyze, Optimize, and Measure again, comparing service level indicators (SLIs) against service level objectives (SLOs).
Sustaining governance and partnership means preventing sprawl while keeping strategy aligned to changing needs. Without governance, resources and costs accumulate without control, so naming, labeling, access, cost, and service-level standards are enforced consistently. Three mechanisms connect policy to enforcement: security marks for Security Command Center insights, labels for billing and logging context, and tags for firewall sources and targets. Measurement uses KPIs for overall performance and metrics for contributing data points, visualized in dashboards of CPU, memory, network, and latency. Retrospectives after milestones or incidents use Start-Stop-Continue to produce action steps, fed by surveys, interviews, feedback forms, and team meetings. Continuous learning through training, certifications, and knowledge sharing keeps skills current and supports innovation.
Exam tip: Cloud Monitoring provides real-time collection and alerting, while BigQuery provides storage and long-term analysis; Personalized Service Health shows incidents relevant to your projects, while the public status dashboard shows all Google Cloud incidents.
Defining value and acting on gaps means turning requirements into measurable targets and closing the distance to them. A requirement states a need such as lower cost or faster performance, while goals, KPIs, metrics, SLIs, and SLOs make it trackable, such as latency below 100 milliseconds or 95% faster scale-up. Proofs of concept validate scope and business impact before broad rollout. Operational data flows from resources to Cloud Monitoring to BigQuery to dashboards, extended by logging exports, Recommender insights, Security Command Center findings, and billing analysis. Evaluation compares results to SLOs, best practices, technical-debt goals, and cost feasibility, then follows Confirm attainment, Assess feasibility, Proceed or refine goals, and Measure again. Rightsizing, autoscaling with managed instance groups, GKE, and Cloud Run, plus budgets, quotas, least privilege, CUDs, Spot VMs, and lifecycle management, convert findings into lower cost and stronger adoption.
Continue with this study guide: Customer success management
Cost optimization is the shift from fixed Capital Expenditure (CapEx) for owned hardware to variable Operational Expenditure (OpEx) for pay-as-you-go cloud use. Every virtual machine, container, dataset, and network flow creates usage data that flows into Cloud Billing, where labels, budgets, quotas, and recommendations connect spend to owners and actions. Control comes from seeing who spent what and preventing uncontrolled growth.
Creating visibility and accountability means attributing every cost to an owner and enforcing limits. Labels and tags such as team=banking_appdev travel with usage into billing and BigQuery exports for chargeback and audits, supported by the hierarchy of organization, folders, projects, and billing accounts plus Terraform enforcement and regular reviews. Usage data moves from collection in billing and monitoring, to organization by service and region, to enrichment with business context, to forecasting and rightsizing with cloud resource management (opens in a new tab). Three controls act at different points:
| Control | What it governs | How it enforces |
|---|---|---|
| Cloud Billing budgets and alerts | Actual versus planned spend | Sends notifications; does not block use |
| Quotas | Quantity of shared resources | Blocks provisioning beyond the limit |
| Organization Policy Service | How resources may be used | Restricts products or configurations |
Least-privilege IAM, reviewed infrastructure as code, and drift detection strengthen prevention before spend occurs.
Exam tip: Cloud Billing budgets and alerts only notify stakeholders and are subject to billing-data delay; quotas and Organization Policy constraints actually block or restrict provisioning.
Comparing CapEx to OpEx total cost means looking beyond instance price to full ownership over time. Traditional CapEx buys servers, network gear, power, cooling, space, and staff upfront and overprovisions for peaks, paying for idle capacity. Google Cloud OpEx scales with use through elasticity and autoscaling, but network use is often billed by utilization and egress, so data-heavy designs must control transfer costs. Total cost of ownership (TCO) includes licenses, power, maintenance, operations salaries, migration cost per virtual machine, storage, compute, network, and assumptions such as hardware life, storage price, and redundancy. Migration Center and the Google Cloud Pricing Calculator build consistent comparisons across summary, technical, and pricing views. Financial agility comes from tracking utilization, rightsizing with Recommender, autoscaling with Compute Engine managed instance groups, GKE, and Cloud Run, and choosing CUDs for steady use, sustained use discounts for continuous use, and Spot VMs for flexible work.
Optimizing compute and storage spend means matching size, price, and placement to actual demand. Right-sizing adjusts Compute Engine machine types and counts from Cloud Monitoring signals and Recommender suggestions, avoiding waste from over-provisioning and bottlenecks from under-provisioning. Discount choice follows predictability: resource-based CUDs commit to regional capacity for one or three years, spend-based CUDs commit to product spend more flexibly, sustained use discounts apply automatically, and enterprise discounts can stack with commitments. Spot VMs offer more than 80% lower prices for fault-tolerant batch or parallel work but move from running to preempted when capacity is reclaimed, so they must not host high-availability needs. Cloud Storage Object Lifecycle Management and Autoclass move aging data to lower-cost tiers or delete it, and single-region buckets cost less than dual- or multi-region buckets, so replication is used only when reliability requires it.
Continue with this study guide: Cost optimization/resource optimization (CapEx/OpEx)
Business continuity is the design of data protection and recovery so work resumes within agreed time and loss limits after zone, region, or data-level failure. The Recovery Time Objective (RTO) defines acceptable downtime, while the Recovery Point Objective (RPO) defines acceptable data loss, with tighter targets requiring more redundancy and faster restore paths. Google Cloud aligns replication, backups, location controls, and key management to those targets.
Protecting and restoring data means combining replicated storage with backups and deletion safeguards. Objects in Cloud Storage dual-regions and multi-regions are stored in at least two separate areas, with Turbo replication giving 15-minute RPO in a dual-region and cross-bucket replication copying asynchronously to another location. Firestore regional databases synchronously replicate across at least three zones for 99.99% availability, while multi-region databases replicate across five zones in three regions for 99.999% availability. Firestore scheduled backups run daily or weekly with up to 14 weeks retention and restore to a new database at lower cost, while point-in-time recovery (PITR) reads up to seven days back at one-minute granularity with zero RTO for reads and one-minute RPO for surgical document repair. Data protection and recovery (opens in a new tab) is strengthened by object holds, soft delete, Object Versioning, Bucket Lock, Object Retention Lock, Backup and DR Service snapshots, and transfer tools for copies outside Google Cloud.
| Option | Retention window | Restore behavior |
|---|---|---|
| Scheduled backups | Up to 14 weeks | Full restore to new database; lowest restore cost |
| PITR reads and writes | Up to 7 days | Surgical document fix; lowest RTO and RPO |
| PITR exports | Export from last 7 days, stored indefinitely | Long archive; highest recovery cost |
Exam tip: Scheduled backups retain for 14 weeks and cost less for full restores, while PITR covers only 7 days at 1-minute granularity but gives lower recovery time and less data loss for surgical fixes.
Aligning architecture to RTO, RPO, and location means choosing replication and controls that meet targets within compliance limits. Cloud Bigtable production instances use at least two clusters in unique zones with automatic replication and failover, BigQuery provides 99.99% SLA with regional or multi-region dataset placement, and Cloud Storage bucket choice sets geographic resilience. Locality-restricted workloads add four decisions: whether multi-regional services are allowed, whether access is location-restricted, whether data at rest must stay in one region, and which regions can run the app. Access is enforced with Google Cloud Armor and VPC Service Controls, creation with Organization Policy Service, and residency with domain-restricted sharing and location constraints. When data cannot leave one region, teams add a same-area copy on another cloud or on-premises, or use Cloud External Key Manager (Cloud EKM) and customer-supplied encryption keys so keys stay outside Google infrastructure, making key-manager availability an RTO dependency. For data-level disasters, BigQuery uses scheduled exports and transfer copies plus append-only partitions to shorten RTO, alongside the Firestore backup and PITR choice.
Implementing standby and failover patterns means deciding how much capacity runs before disaster to meet RTO and RPO. Cold standby keeps only backups, images, and templates with lowest cost but largest RTO, rebuilding with Deployment Manager, startup-script restores, testing, and Cloud DNS switching. Warm standby keeps a minimal always-on database replica with asynchronous replication on the smallest viable machine, launching app servers from snapshots only during failover over Cloud Interconnect plus virtual private network (VPN). Hot standby runs full active-active environments across regions behind a global load balancer with health checks, using Cloud Spanner, Cloud Storage multi-regional services, Cloud SQL high availability, managed instance groups (MIGs), regional GKE clusters, and regional persistent disks for near-zero RTO and RPO at highest cost. Synchronous replication supports RPO zero but adds latency, while asynchronous replication lowers latency but risks recent writes, so transfer frequency must be shorter than the RPO target. Automation with autohealing, startup scripts, reservations, and monitoring keeps failover fast, and critical operations must depend only on data-plane components, not on management-plane APIs.
| Pattern | Standby resources | Choose when |
|---|---|---|
| Cold standby | Backups, snapshots, images, templates; no running servers | Cost matters most; hours of RTO acceptable |
| Warm standby | Minimal always-on database replica plus snapshots | Balance of budget and availability needed |
| Hot standby | Full active-active environments with continuous replication | Downtime and loss must be near zero |
Continue with this study guide: Business continuity
Cloud Billing budgets and alerts only send notifications about actual versus planned spend and do not block use, while quotas block provisioning beyond the limit and Organization Policy Service restricts how resources may be used. Budgets are subject to billing-data delay, so prevention relies on quotas and Organization Policy constraints, strengthened by least-privilege Identity and Access Management (IAM), reviewed infrastructure as code, and drift detection.
Cold standby with only backups, snapshots, images, and templates is lowest cost but has the largest Recovery Time Objective (RTO), warm standby with a minimal always-on database replica balances budget and availability, and hot standby with full active-active environments provides near-zero RTO and Recovery Point Objective (RPO) at highest cost. Choose cold standby when cost matters most and hours of RTO are acceptable, warm standby when a balance of budget and availability is needed, and hot standby when downtime and loss must be near zero with continuous replication across regions behind a global load balancer.
Firestore scheduled backups run daily or weekly with up to 14 weeks retention and restore to a new database at lower cost, while point-in-time recovery (PITR) covers up to seven days at one-minute granularity for surgical document repair. Scheduled backups provide full restores with the lowest restore cost, while PITR reads provide zero Recovery Time Objective (RTO) for reads and one-minute Recovery Point Objective (RPO) for surgical fixes, and PITR exports from the last seven days can be stored indefinitely at the highest recovery cost.
Source code flows into Cloud Build for build and test, into Artifact Registry for versioned storage, and into Cloud Deploy for ordered promotion across development, test, and production. Automated tests act as gates between stages, and Binary Authorization blocks deployment to Google Kubernetes Engine (GKE) or Cloud Run unless images carry required attestations, while Artifact Analysis scans images and monitors for new vulnerabilities.
Last updated