Prepare and test your skills

Prepare and test your skills

Google Cloud compute solutions differ in how much infrastructure Google manages and how much control you keep. The three core options are Compute Engine (opens in a new tab) for self-managed virtual machines (VMs) and bare metal, Google Kubernetes Engine (GKE) (opens in a new tab) for managed Kubernetes container orchestration, and Cloud Run (opens in a new tab) for serverless containers. Choosing among them depends on application architecture, scaling needs, and how much operational work you want to handle.
Compute Engine is Google Cloud's Infrastructure as a Service (IaaS) product for creating and running VM and bare metal instances. It gives you full control over the operating system, kernel settings, boot files, CPU processors, memory configurations, storage types, and guest OS. It supports NVIDIA GPUs and Google TPUs attached to instances, and its minimum uptime service level objective (SLO) is 99.5%, varying by region, Network Service Tier, and deployment configuration. You manage compute instances, network rules, traffic routes, and backup systems yourself, though load balancing and autoscaling can improve reliability.
Compute Engine offers flexible machine and storage choices for different workloads:
Compute Engine can run individual VMs or managed instance groups (MIGs) that are managed as a single entity. Regional MIGs automatically provision VMs across multiple zones in a region, and autoscaling adds or removes VMs based on demand. Choose Compute Engine for custom architectures, specialized kernel settings, lift-and-shift migrations, commercial off-the-shelf applications, or any workload needing administrative control over virtual servers.
Google Kubernetes Engine (GKE) is a managed Kubernetes platform for deploying, scaling, and orchestrating containerized applications. Google manages the Kubernetes control plane for security and system availability, and Google manages the operating system in both modes. GKE has two modes: Autopilot, where Google manages cluster nodes and adjusts capacity without manual intervention, and Standard, where you manage cluster node pools. GKE supports NVIDIA GPUs and Google TPUs for containerized workloads.
GKE scales containers and nodes in layers, and supporting services manage traffic, configuration, and data protection. Horizontal Pod Autoscalers add or remove container replicas, Vertical Pod Autoscalers adjust pod requests and limits, the cluster autoscaler adds or removes nodes, and Kueue orchestrates batch jobs. Cloud Service Mesh controls network traffic into clusters and between clusters, Config Sync applies GitOps-style configuration from a single source, and Backup for GKE provides periodic backup and restore of cluster persistent data. Choose GKE for distributed applications, general-purpose microservices, APIs, and standard containerized workloads that need robust cluster orchestration.
Cloud Run is a serverless platform for running containerized applications without managing servers or clusters. Google handles all infrastructure management, including the operating system, kernel, network setup, and capacity management. You package workloads as containers and run them as web services, asynchronous tasks, or batch jobs. Cloud Run services are regional resources that are replicated across multiple zones in their region for availability.
Cloud Run starts, runs, and scales container instances without manual intervention. For web services, Cloud Run adds instances when incoming requests or CPU consumption rises and removes instances when traffic decreases. It supports NVIDIA RTX PRO 6000 Blackwell and L4 GPUs, but does not support other GPU models or Google TPUs. Choose Cloud Run for containerized web applications or jobs when you want to focus on code and data rather than infrastructure operations.
The choice among Compute Engine, GKE, and Cloud Run is mainly a trade-off between configuration flexibility and management effort. Compute Engine and GKE give more configuration flexibility, but you are responsible for more resource management, while Cloud Run deploys workloads to a preconfigured platform with minimal management effort. A simple decision path is to use Cloud Run if you want Google to manage all infrastructure, use GKE if you need cluster orchestration for standard containerized applications such as microservices or APIs, and use Compute Engine if you want to manage VMs or bare metal instances yourself.
| Compute option | Service model | Primary unit | Scaling | Best for |
|---|---|---|---|---|
| Compute Engine | Infrastructure as a Service (IaaS) | VM or bare metal instance | Manual or automatic with MIGs | Custom architectures, specialized kernel settings, lift-and-shift, full VM control |
| Google Kubernetes Engine (GKE) | Managed Kubernetes | Container in a pod, managed by cluster | Horizontal and Vertical Pod Autoscalers for pods, cluster autoscaler for nodes, Kueue for batch | Distributed apps, microservices, APIs, standard containerized workloads needing orchestration |
| Cloud Run | Serverless containers | Container instance | Automatic based on requests or CPU consumption | Containerized web apps or jobs without server administration |
Exam tip: Choose Cloud Run when Google should manage all infrastructure, GKE when you need Kubernetes cluster orchestration for containers, and Compute Engine when you need VM or bare metal control.
Compute service models describe how management responsibility is divided between you and Google, from raw virtual machines to fully managed platforms. Google Cloud compute services provide virtualized processing power and memory for running applications along this spectrum. The three main models are Infrastructure as a Service (IaaS), Platform as a Service (PaaS), and serverless computing, and the right choice depends on how much control you need versus how much operational management you want Google to take on.
Infrastructure as a Service (IaaS) is the compute model that gives you the most control over the environment. You manage the operating system (OS), middleware, and applications, while Google manages the underlying virtualization, servers, storage, and networking. Google Cloud IaaS options include Compute Engine (opens in a new tab) for VM and bare metal instances where you choose machine series, storage, accelerators, and guest OSs, and Google Cloud VMware Engine (opens in a new tab) for migrating and running VMware workloads natively on Google Cloud. With VMware Engine, Google manages the hardware and VMware platform, but you keep full control over your VMs, guest OSs, and applications. IaaS fits infrastructure architects, machine learning (ML) infrastructure engineers, and platform engineers who need to design and manage custom environments from scratch.
Platform as a Service (PaaS) is the compute model that balances control with ease of use for deploying workloads quickly. Google manages the underlying infrastructure, operating system, and runtime environments, so you focus on running your workloads instead of maintaining the platform. In Google Cloud, PaaS appears as managed platforms that still leave workload and platform choices with you. Data scientists, artificial intelligence (AI) researchers, and ML engineers often prefer PaaS because they can provision resources and deploy workloads without deep infrastructure management.
Google Cloud PaaS examples cover batch, containers, and high-performance clusters:
Serverless computing is the compute model that abstracts away the most infrastructure management. Cloud Run is a fully managed platform that runs highly scalable containerized applications on serverless infrastructure, so containers can respond to web requests or events without you managing servers or clusters. Other serverless deployment options include App Engine flexible environment and Cloud Run functions. In serverless computing, Google handles the infrastructure behind the scenes, and you provide only the application code or container that runs. Choose serverless when you want request-driven or event-driven workloads to run without managing servers or clusters.
The three models differ mainly in where the management boundary sits between you and Google. IaaS keeps the operating system and applications with you, PaaS moves the platform and runtime to Google while you keep your workloads, and serverless moves all infrastructure management to Google while you keep only code or containers. This boundary determines how much flexibility you keep and how much operational work you hand off.
| Model | What you manage | What Google manages | Example services | Choose when |
|---|---|---|---|---|
| IaaS | OS, middleware, applications | Virtualization, servers, storage, networking | Compute Engine, Google Cloud VMware Engine | You need the most control and custom environments |
| PaaS | Workloads and some platform choices | Underlying infrastructure, OS, runtime environments | Batch, GKE, Cluster Director | You want a balance of control and ease of use |
| Serverless | Application code or container | All infrastructure management | Cloud Run, App Engine flexible environment, Cloud Run functions | You want to run request- or event-driven workloads without managing servers or clusters |
Exam tip: IaaS gives the most control, serverless gives the most abstraction, and PaaS sits between them by managing the platform while leaving you responsible for your workloads.
Cloud-based compute means running applications on virtualized infrastructure in a public cloud instead of on servers you own and maintain. The cloud provider manages the physical data center and virtualization layer, while you focus on your applications and how many resources they use. This shift brings three main advantages: less operational work, the ability to scale resources up and down with demand, and lower cost through paying only for what you use.
Reduced operational overhead means you hand off layers of the technology stack to the cloud provider instead of managing them yourself. In an on-premises environment you own everything, including cooling, physical security, hardware maintenance, the virtualization hypervisor, and your applications. In a private hosting environment you outsource physical security, power, and cabling, but you still manage the hypervisor and workloads. In a public cloud environment the provider also manages hardware maintenance and virtualization, and you can use fully managed services to hand off even more runtime work.
| Responsibility | On-premises environment | Private hosting environment | Public cloud environment |
|---|---|---|---|
| Physical security, power, and cabling | You | Service provider | Service provider |
| Hardware including maintenance | You | Depends on provider | Service provider |
| Virtualization platform | You | You | Service provider |
| Applications and app resources | You | You | You, with option to use fully managed services |
Because Google Cloud manages the underlying infrastructure, teams spend less time on patching, power, cooling, and hardware upkeep. That time can be redirected to building and improving workloads rather than keeping the data center running.
Elastic scaling is the ability to automatically add resources when traffic rises and remove them when traffic falls. An on-premises data center is rigid because capacity is fixed by the servers you purchased, which often leads to over-provisioning for peaks and wasted capacity the rest of the time. Google Cloud resources scale elastically with workload fluctuations, so you have the resources you need when you need them and avoid paying for idle capacity.
Google Cloud provides autoscaling for different compute styles:
Together, these options improve performance during peaks and reduce costs during quiet periods without manual resizing of servers.
Consumption-based pricing means total cost follows actual usage plus discounts, rather than the fixed cost of owning hardware. A complete total cost of ownership model includes hardware and software plus data center operating costs such as power, cooling, maintenance, and support services. Cloud workloads can lower that total because elastic scaling removes unused capacity and pricing rewards steady or flexible usage patterns.
Google Cloud offers several tools to lower the billing rate and track spending:
Cost Management (opens in a new tab) tools such as budgets and alerts help track spending and send alerts when budgets are exceeded. Tagging and labeling resources by project or department makes it possible to see which workloads drive costs, forecast future spending, and find underutilized resources to rightsize or shut down.
Exam tip: Choose committed use discounts for steady, predictable usage and Spot VMs for flexible, fault-tolerant workloads that can use unused capacity at a lower price.
Use Cloud Run when you want Google to manage all infrastructure, use Google Kubernetes Engine (GKE) when you need Kubernetes cluster orchestration for standard containerized applications such as microservices or APIs, and use Compute Engine when you need to manage virtual machines or bare metal instances yourself. Compute Engine and Google Kubernetes Engine (GKE) give more configuration flexibility but require more resource management, while Cloud Run deploys workloads to a preconfigured platform with minimal management effort.
Infrastructure as a Service (IaaS) gives the most control, with you managing the operating system, middleware, and applications while Google manages virtualization, servers, storage, and networking. Platform as a Service (PaaS) balances control with ease of use, with Google managing the underlying infrastructure, operating system, and runtime environments while you focus on workloads. Serverless computing abstracts away the most management, with Google handling all infrastructure management while you provide only the application code or container.
Compute Engine scales through managed instance groups that are managed as a single entity and add or remove virtual machines based on demand. Google Kubernetes Engine (GKE) scales in layers using Horizontal Pod Autoscalers for replicas, Vertical Pod Autoscalers for requests and limits, the cluster autoscaler for nodes, and Kueue for batch jobs. Cloud Run provides built-in automatic scaling that adds container instances when incoming requests or CPU consumption rises and removes instances when traffic decreases.
Committed use discounts provide discounts in exchange for committing to a level of usage over a period of time for steady, predictable usage. Spot VMs provide access to unused virtual machine capacity at a lower cost compared to regular virtual machines for flexible, fault-tolerant workloads. Cost Management tools such as budgets and alerts help track spending, while tagging and labeling resources helps identify which workloads drive costs.
Last updated