Prepare and test your skills

Prepare and test your skills

Contrast Financial Models and Cost Structures is the comparison of how organizations pay for IT under traditional data centers versus cloud computing. Traditional technology uses capital expenditure (CapEx) with large fixed upfront investments in owned hardware, while cloud computing uses operating expenditure (OpEx) with variable payments based on measured consumption.
The traditional capital expenditure (CapEx) model requires an organization to buy and own physical infrastructure before applications can run. The organization purchases servers, storage, software licenses, and data center capacity, and it also pays for power, cooling, rack space, maintenance, and IT operations staff. Because capacity must be bought for peak demand and future growth, much of that hardware sits idle during normal periods.
The full burden of ownership stays with the organization in the CapEx model. Hardware costs are spread over a useful life of several years, while each host carries ongoing monthly facility costs and staffing for server, network, and security administration. This creates a rigid cost structure where teams over-provision as a safety net for spikes and failover, which increases waste and makes budgets hard to adjust when needs change.
The cloud operating expenditure (OpEx) model replaces upfront purchases with pay-as-you-go use of shared cloud resources. An organization provisions virtual machines, containers, storage, and managed services when needed and releases them when demand drops, so spending rises and falls with actual workload use. This shift gives more agility to scale, budget, and choose where and when to deploy applications.
Variable cost works because cloud capacity can expand and contract automatically instead of sitting idle. Compute Engine managed instance groups (MIGs) (opens in a new tab) add virtual machines (VMs) when utilization such as CPU rises and remove VMs when utilization falls. Google Kubernetes Engine (GKE) (opens in a new tab) adjusts nodes and Pods to match application needs, and Cloud Run automatically scales instances up with incoming traffic and scales down to reduce costs when traffic decreases.
| Model | When cost is paid | How capacity is handled | Budget impact |
|---|---|---|---|
| Traditional CapEx | Upfront, before use, plus ongoing facility and staffing costs | Buy for peak demand and keep extra servers idle | Fixed and rigid, hard to adjust to change |
| Cloud OpEx | Ongoing, based on measured consumption | Scale resources up and down with demand | Variable and flexible, tied to actual use |
Total cost of ownership (TCO) (opens in a new tab) is the complete cost of running a workload, not just the purchase price of hardware and software. A fair on-premises to cloud comparison includes power, cooling, maintenance, rack space, network use, operating system licenses, operations, migration, and support. Google Cloud compares these options over one-year and five-year periods to show how elastic scaling and managed services change long-term value.
A value-driven and TCO-aware mindset connects cloud spending to business results. Unit costing and unit economics correlate spending with measures such as revenue per customer or transaction volume. When higher spending comes from profitable growth, further cloud investment is justified, while spending without matching value signals inefficiency that should be optimized.
Keeping variable cloud spending under control depends on visibility, accountability, and discounted pricing. Cloud Billing budgets and alerts (opens in a new tab) track actual spending against defined limits and notify stakeholders before major overruns occur. Tagging and labeling attach business metadata such as project, team, or department to resources, so costs can be allocated to the team that created them and analyzed by resource type or time period.
Google Cloud offers pricing options that lower the billing rate for predictable or flexible use:
Exam tip: CapEx means fixed upfront investment in owned data centers, while OpEx means variable consumption-based payment for cloud resources actually used.
Distinguish Operational Responsibilities and Maintenance is the comparison of full-stack ownership on premises with the shared responsibility model in the cloud. Traditional environments require one organization to manage everything from physical servers to applications, while cloud technology has the provider manage physical infrastructure, cooling, and hardware maintenance and the customer manage software, operations, and data protection.
The shared responsibility model divides operational ownership between the cloud provider and the customer. Google Cloud manages physical infrastructure, cooling, and hardware maintenance. The customer remains responsible for operating system security and compliance, drivers, compatibility, security patches, software updates, and for designing incident and operations management, change management, Identity and Access Management (IAM), backup and recovery, high availability, and disaster recovery.
| Responsibility area | Traditional on-premises | Cloud shared responsibility |
|---|---|---|
| Physical infrastructure, cooling, hardware maintenance | Organization | Cloud provider |
| Operating system, drivers, compatibility | Organization | Customer |
| Security patches and software updates | Organization | Customer |
| IAM, backup and recovery, high availability, disaster recovery | Organization | Customer |
| Incident and operations management, change management | Organization | Customer |
For bare metal (opens in a new tab), Google Cloud technical support focuses on infrastructure-related issues and does not support the operating system, drivers, or compatibility. Customers contact the operating system vendor for those software issues.
Operational monitoring and observability in Google Cloud use metrics, logs, and traces to detect potential failures before errors occur. Metrics are numerical measurements tracked at time intervals, such as request rate, error rate, orders placed, and payments received, and they can serve as service-level indicators (SLIs). Logs are time-stamped records of discrete events such as failures, errors, or state changes, and they can also be used for SLIs. Traces follow the journey of a single user or transaction through separate applications or components, helping identify bottlenecks and how long each journey takes.
Cloud Monitoring (opens in a new tab) and Cloud Logging track response times and error rates and analyze default service metrics for component dependencies. Teams can also create custom metrics with the Google Cloud SDK to monitor business-specific activity. Together, metrics, logs, and traces show how components depend on each other and where performance is degrading.
Proactive troubleshooting and resource optimization mean collecting the right signals and acting on them without waiting for an outage. Proactive troubleshooting means enabling logging across all workload components and activating logs such as Cloud Storage access logs and VPC Flow Logs. Logging has associated costs, so exclusion filters on log sinks can exclude selected logs from being stored.
Resource optimization depends on utilization signals from services such as Google Kubernetes Engine (GKE), Compute Engine, and Managed Service for Apache Spark. Teams monitor CPU consumption, network I/O metrics, and disk I/O metrics, while Cloud Monitoring analyzes performance trends, experiment effects, alerts, and retrospective analyses. Active Assist (opens in a new tab) provides insights and recommendations to adjust resource allocation. For alerts, teams focus on critical metrics, set thresholds that minimize alert fatigue, and ensure timely responses.
Maintenance during VM migration (opens in a new tab) is a customer task that moves a workload from an older virtual machine (VM) series to a newer series by creating a new VM. Moving from first- or second-generation VMs to third-generation or later follows an ordered process where the source VM is prepared and stopped before its data is carried forward to the new instance:
A left-to-right process flow showing the five ordered VM migration steps: prepare source, stop and snapshot, create new VM, start and reattach, then reconfigure, with SUSE and H3 special rules.
Special rules apply to specific operating systems and machine types. On SUSE Enterprise Linux Server, rebuild the Initial RAM filesystem (initramfs) with all drivers to avoid hardware-specific dependencies and prevent boot failures. Moving from an H3 VM to any other machine series is not supported.
Exam tip: The cloud provider manages physical infrastructure, cooling, and hardware maintenance, but the customer still owns OS patches, drivers, IAM, backups, high availability, disaster recovery, and application updates.
Evaluate Scalability and Resource Provisioning is the comparison of rigid manual scaling on premises with automatic elasticity in the cloud. On-premises technology requires hardware to be bought and managed in advance, while cloud technology lets resources grow and shrink on demand without buying physical servers.
On-premises provisioning means buying, installing, and managing physical servers in an owned data center before they are needed. Capacity is fixed by the hardware owned, so planning, budgeting, and scaling have little agility because adding power means ordering equipment, waiting for delivery, and setting it up by hand. The team also carries the full burden of server management, including Windows, SQL Server, upgrades, and maintenance.
Legacy on-premises applications (opens in a new tab) share common limits that make scaling harder:
Because capacity cannot be added quickly, organizations must over-provision for the highest possible peak plus a safety margin for failover. That extra hardware sits idle and wastes money during normal periods, yet it can still fall short during an unexpected surge.
Cloud elasticity is the ability to increase and decrease compute, storage, and other resources on demand without buying physical hardware. A cloud architecture gives full agility to scale and budget, with more flexibility to choose where and when to deploy applications and with serverless options that grow and shrink automatically. Google manages the underlying infrastructure, so management overhead drops and resilience improves.
| Approach | How capacity is added | Agility and cost effect |
|---|---|---|
| On-premises | Manual purchase and setup of servers in advance | Slow to change; pay for peak capacity even when idle |
| Cloud | Automatic allocation from the provider's pool, including serverless architecture | Fast to change; scale up for peaks and scale down to reduce cost when demand is low |
Choose on-premises provisioning only when a workload cannot move because of legacy hardware or regulatory rules. Choose cloud elasticity when demand varies by season, growth, or events and paying only for what is used matters.
Planning capacity and using autoscaling in Google Cloud means forecasting future needs, managing quotas, and letting policies adjust resources automatically. Usage data flows from workloads into Cloud Monitoring (opens in a new tab) and BigQuery, where it is collected and analyzed to identify trends, spot bottlenecks, and forecast demand. Forecasts combine historical usage patterns, growth projections, and business requirements such as service-level agreements (SLAs) and performance targets.
Capacity planning (opens in a new tab) uses historical metrics such as CPU utilization, memory usage, and network traffic to reveal periods of high demand. Growth projections account for more users, new products and features, and marketing campaigns. Sizing must also cover temporary spikes from seasonal variations like holiday shopping or end-of-quarter sales, planned events like product launches, and enough extra capacity for the primary and disaster recovery (DR) system to support graceful failover during natural disasters and cyberattacks.
Autoscaling depends on those signals to adjust resources automatically in response to workload fluctuations. Autoscaling policies watch metrics like CPU utilization, memory usage, and queue depth, then add compute instances and storage when demand rises and remove them when demand falls. This keeps performance strong during peaks while minimizing costs when utilization is low.
Exam tip: On-premises scales by over-provisioning hardware in advance for the peak, while cloud with autoscaling scales resources up and down automatically based on CPU, memory, and queue depth.
The traditional capital expenditure (CapEx) model requires upfront purchase of owned servers, storage, software licenses, and data center capacity for peak demand plus ongoing facility and staffing costs. The cloud operating expenditure (OpEx) model replaces upfront purchases with pay-as-you-go use of shared resources that are provisioned when needed and released when demand drops, so spending rises and falls with actual workload use.
Total cost of ownership (TCO) is the complete cost of running a workload, not just the purchase price of hardware and software. A fair comparison includes power, cooling, maintenance, rack space, network use, operating system licenses, operations, migration, and support over one-year and five-year periods.
Google Cloud manages physical infrastructure, cooling, and hardware maintenance. The customer remains responsible for operating system security and compliance, drivers, compatibility, security patches, software updates, Identity and Access Management (IAM), backup and recovery, high availability, disaster recovery, incident and operations management, and change management.
On-premises provisioning requires manual purchase, installation, and management of physical servers in advance, so organizations over-provision for peak demand and keep extra servers idle. Cloud elasticity automatically increases and decreases resources on demand without buying hardware, with autoscaling policies watching metrics like CPU utilization, memory usage, and queue depth to add capacity when demand rises and remove it when demand falls.
A retail organization runs its e-commerce application in an on-premises data center. To handle massive traffic surges during seasonal sales, the company purchased enough physical servers to meet peak demand. This traditional setup has resulted in several challenges:
How does migrating to a cloud environment resolve these resource provisioning challenges?
It enforces resource type-specific quotas to cap infrastructure usage at baseline levels during peak traffic periods
It relies on static zonal reservations to permanently maintain maximum virtual machine capacity year-round
It replaces horizontal scaling with manual vertical scaling so administrators can upgrade server memory ahead of traffic spikes
It provides elasticity and autoscaling to dynamically add or remove compute resources based on real-time demand
Last updated