Prepare and test your skills

Prepare and test your skills

Private and hybrid cloud architectures are two ways to place workloads to balance control with flexibility. A private cloud (opens in a new tab) is dedicated infrastructure used by one organization, while a hybrid cloud (opens in a new tab) connects that private environment to a public cloud so data and workloads can move between them in a controlled way.
A private cloud gives one organization exclusive use of compute, storage, and networking resources. The private environment sits inside the organization's trust boundary, while the public cloud sits outside it, so sensitive systems remain under direct organizational ownership. This separation matters when policy or law requires highly secure data to stay on premises and when data sovereignty rules require personally identifiable information (PII) to stay in a specific country or region. Organizations keep production systems with strict privacy, compliance, and performance needs in the private cloud for this reason. The tradeoff is less elasticity, because capacity must be bought and installed in advance, so a private cloud fits best when control outweighs on-demand scale.
A hybrid cloud is an architecture where application components are distributed across a private environment and a public cloud such as Google Cloud. Network connectivity carries requests and data flows from the private environment to public cloud services and back, so the two sides operate as one system. The distributed application depends on that connection and on a common communication model that defines how its parts talk to each other. Hybrid architectures combine private control with public capabilities in practical ways:
Exam tip: Keep production and performance load testing in the private environment and use the public cloud for development and functional testing, because the infrastructures are not identical so cloud performance results do not validate on-premises performance.
Hybrid cloud balances dedicated security with public flexibility by placing each workload where its constraints are met. Regulated or sovereign data stays in the private environment under organizational control, while elastic or innovative work runs in the public cloud and reaches that data through controlled interfaces. A common example unlocks legacy systems that hold critical data by placing an API management layer such as Apigee (opens in a new tab) in front of them, so new cloud applications can call the on-premises system with added security, analytics, and scalability and with minimal changes. Workload portability makes this balance practical across environments. Containers with Google Kubernetes Engine (GKE) (opens in a new tab) provide a common runtime layer, so the same workload can run on both sides without being rewritten, while a zero trust service mesh controls separation between environments and open source foundations such as Kubernetes reduce vendor lock-in.
The environment hybrid pattern (opens in a new tab) splits environments by purpose rather than by application. Development and functional testing run in the public cloud, while production, staging, and performance and reliability testing run in the private environment. Data flows from developers through cloud test instances for functional checks, and only validated code moves to the private production system. This split requires functional equivalence without assuming identical performance, so teams use the same database engine and version or a tested equivalent version. The network boundary enforces separation through the mirrored network pattern, which blocks direct communication between environments and allows cross-environment traffic only in a controlled manner, plus governance that controls what data can be used for cloud testing.
Moving to hybrid follows an ordered journey where each workload can be in a different phase at the same time:
Start with a workload that is not business-critical and has minimal dependencies, yet is typical enough to serve as a blueprint for later waves.
A multicloud strategy (opens in a new tab) uses more than one cloud service provider (CSP) to meet business and technical goals. The strategy can reduce vendor lock-in, allow use of best-in-breed services from multiple providers, and improve business continuity by distributing workloads across clouds. It also adds management, security, integration, performance, skills, and cost challenges, so the choice should balance business priorities rather than a single driver.
Multicloud is commonly chosen for data sovereignty (opens in a new tab), cost reduction, unique capabilities, vendor lock-in avoidance, and resilience. Each driver has its own considerations:
A single cloud provider can offer reduced complexity, built-in integrations among services, and cost optimization such as committed use discounts, so multicloud is not automatically better. The choice depends on sovereignty needs, desired services, lock-in risk, and tolerance for complexity:
| Approach | Main strengths | Main tradeoffs and considerations |
|---|---|---|
| Single cloud | Reduced complexity; built-in integrations; committed use discounts | May lack a local region for sovereignty; vendor lock-in risk; limited access to other providers' capabilities |
| Multicloud | Data sovereignty; best-in-breed services; lock-in avoidance; resilience to outages | Greater management complexity; harder consistent security; software integration work; consistent performance challenges; expensive skills; harder pricing and tool management without unified cost visibility |
Building a multicloud technical team can be expensive unless managed by a third party. Managing pricing and tools from each CSP is harder without unified dashboards, where the Looker cloud cost management solution (opens in a new tab) can be used where applicable.
Multicloud resilience is the ability to keep business-critical applications available by distributing them across more than one CSP. If one CSP region fails, traffic flows from users to the second CSP in the same region, but only when both providers support the required capabilities in that region. When data residency rules require sensitive data such as PII to stay in a specific country or region, using two CSPs in one region can provide outage resilience while meeting those restrictions. Deploying on multiple clouds for disaster recovery can raise short-term costs but can prevent outages that cause long-term financial and reputational damage. Long-term value varies with organization size, technology scale, criticality, and industry.
Before adopting multicloud for resilience, assess data movement and transfer charges, security and manageability complexity, capability parity in the selected region, and whether the team has the required skills.
Exam tip: Multicloud resilience depends on capability parity: both CSPs in the selected region must offer the required capabilities, and traffic can be routed to the second CSP only if that condition is met.
Open source solutions can ease hybrid and multicloud adoption and minimize vendor lock-in, but planning must still address interoperability, manageability, cost, and security. Building on a platform that contributes to open source can simplify adoption. Open cloud provides maximum choice and abstracts complexity, while Google Cloud offers flexibility to migrate, build, and optimize across hybrid and multicloud environments while meeting regulatory needs. Google is a large contributor to open source and helped develop well-known technologies like Kubernetes (opens in a new tab), which as a managed service can reduce manageability and security complexity.
A multicloud strategy should start from business requirements, translate them into technical objectives, and weigh drivers together rather than in isolation. For shared ownership, consider building a Cloud Center of Excellence (COE) (opens in a new tab). A COE team helps internal teams serve the business during the transition, speeds adoption, drives standardization, and aligns business strategy with cloud investments. Migration remains a journey with Assess, Plan, Deploy, and Optimize phases that can apply to each workload individually.
On-premises infrastructure (opens in a new tab) is computing capacity that an organization owns and runs in its own data center, while public cloud infrastructure (opens in a new tab) is computing capacity provided by a cloud service provider and consumed on demand. The main contrast is financial and operational: on-premises requires upfront capital spending and hardware management, while public cloud shifts spending to usage-based operating costs with automated, elastic scaling.
On-premises infrastructure is computing capacity owned and operated in a private data center or private computing environment. It is where existing workloads stay during a cloud-first approach and often hosts legacy applications that hold critical data. On-premises deployment may be required when public cloud use is impossible for technical or organizational reasons, such as legal restrictions or the need to keep highly secure data on premises. The organization is responsible for hardware and self-managed software, including self-hosted databases like MySQL and on-premises hardware load balancing and web application firewall (WAF) capabilities. Because capacity is tied to owned physical resources, the organization must plan, maintain, and operate everything itself.
Public cloud infrastructure is computing capacity from a CSP that offers on-demand pricing, quick setup, and high compute capacity. It fits temporary projects, research and development, and workloads that must scale quickly. Public cloud supports autoscaling, which automatically adjusts compute instances, storage, and other resources based on fluctuations using metrics such as CPU utilization, memory usage, and queue depth. Managed services reduce operational work: Cloud SQL (opens in a new tab) provides a managed database, GKE Autopilot manages cluster nodes, scaling, and security settings, and Cloud Load Balancing and Google Cloud Armor replace hardware load balancing and WAF. A single provider can also offer reduced complexity and built-in integrations among services.
The financial contrast is a shift from capital expenditure (CapEx) (opens in a new tab) to operational expenditure (OpEx). On-premises requires upfront investment in hardware and data center resources, which is a capital expense. Public cloud uses on-demand pricing where organizations pay for consumed resources, which is an operating expense. This shift can reduce large upfront CAPEX and lower total cost of ownership for some applications. Public cloud costs can still include data transfer charges, and hybrid or multicloud setups can add management complexity and cost. Committed use discounts are a public cloud cost optimization option.
Exam tip: On-premises spending is usually a capital expense for owned hardware, while public cloud spending is usually an operating expense based on consumption.
Operationally, on-premises requires manual hardware maintenance and self-managed software, while public cloud provides automated scaling and managed services. On premises, capacity is tied to physical hardware, and the team handles load balancing, WAF, and self-hosted databases. In public cloud, autoscaling adjusts resources to meet demand, which maintains performance during peaks and minimizes cost when utilization is low. Managed services move routine configuration, scaling, security, and load balancing to the provider. Cloud Monitoring (opens in a new tab) provides visibility into performance, availability, and health so teams can optimize resources.
On-premises and public cloud differ across cost, operations, and control in ways that affect which workloads fit where:
| Dimension | On-premises | Public cloud |
|---|---|---|
| Cost model | Upfront CapEx for hardware and data center resources | On-demand OpEx based on consumption; committed use discounts available |
| Provisioning and scaling | Capacity tied to owned hardware; scaling depends on physical resources | Quick setup and high capacity; autoscaling adjusts compute and storage to demand |
| Maintenance | Organization maintains hardware and self-managed software such as MySQL and hardware load balancing/WAF | Managed services such as Cloud SQL, GKE Autopilot, Cloud Load Balancing, and Google Cloud Armor reduce work |
| Control and compliance | Can be required for highly secure data, sovereignty, or legal restrictions | May be restricted for sensitive data; residency can require local regions |
| Best fit | Existing workloads, legacy systems with critical data, or workloads that cannot move | New workloads, modernization, analytics and AI, elastic services, and global architectures |
Transition follows a migration journey with four phases that can apply to each workload individually, so different workloads can be in different phases at the same time:
A hybrid approach can combine both models, with existing workloads on premises and new workloads in public cloud communicating over a hybrid connection. Migration can be phased using one or more approaches:
This lets an organization choose the model that best fits each workload instead of applying one model to everything.
On-premises infrastructure is computing capacity owned and operated in an organization's own data center with upfront capital expenditure (CapEx) and manual hardware and software maintenance, while public cloud infrastructure is computing capacity from a cloud service provider (CSP) with on-demand operational expenditure (OpEx) pricing, quick setup, autoscaling, and managed services. On-premises fits existing workloads, legacy systems with critical data, or highly secure and regulated workloads that must stay on premises, while public cloud fits new, elastic, analytics, artificial intelligence, and temporary workloads that must scale quickly.
A private cloud is dedicated compute, storage, and networking infrastructure used exclusively by one organization inside its trust boundary, while a hybrid cloud is an architecture that connects that private environment to a public cloud such as Google Cloud so data and workloads can move between them in a controlled way. Private cloud provides exclusive control for strict privacy, compliance, data sovereignty, and performance needs but requires capacity to be bought in advance with less elasticity, while hybrid cloud combines private control with public capabilities such as artificial intelligence and analytics, disaster recovery, development and testing, and temporary high-compute capacity.
Use a multicloud strategy that uses more than one cloud service provider (CSP) when goals include data sovereignty in a new region, access to unique best-in-breed capabilities, avoidance of vendor lock-in, or improved resilience by routing traffic to another provider if one region fails. A single cloud provider can offer reduced complexity, built-in integrations, and cost optimization such as committed use discounts, while multicloud adds management, security, integration, performance, skills, and cost challenges and requires capability parity in the selected region for resilience.
A financial services organization is modernizing its application delivery lifecycle and must satisfy the following business and technical requirements:
Which architectural approach should the organization adopt to meet these requirements?
Expand the private on-premises data center to host dedicated development, testing, and production environments on identical physical hardware
Implement a hybrid cloud architecture that keeps production workloads in the private data center, runs development and testing in the public cloud, and uses Google Kubernetes Engine (GKE) for container-based workload portability
Migrate all development, testing, and production environments to a single public cloud provider and apply Sensitive Data Protection to de-identify customer data
Adopt a multicloud architecture that distributes production workloads and customer data across two different public cloud service providers within the same geographic region
Last updated