Intrigued by the art of cloud architecture? Discover how to design, develop, and manage robust, secure, scalable, and dynamic solutions on Google Cloud as you prepare for the Professional Cloud Architect exam!
When creating virtual machines in Compute Engine, users must choose the right mix of CPU, memory, and storage. An Instance Template defines a virtual machine configuration that can be reused to create identical virtual machines or managed instance groups. For deeper backups or configuration cloning, a Machine Image captures the disk data, metadata, and configuration of an active virtual machine. The lifecycle of a virtual machine moves from the provisioning state, where Google Cloud allocates resources, to staging for setup, and finally to running once the virtual machine is fully operational.
Architects must choose between Standard VMs and Spot VMs based on workload reliability requirements. Standard VMs offer highly available resources that run continuously until a user stops or deletes them, making them ideal for core business applications. In contrast, Spot VMs offer spare compute capacity at a significant discount, but they can be reclaimed by Google Cloud at any time if the capacity is needed elsewhere.
When Google Cloud needs to reclaim resources, a Spot VM receives a 30-second termination notice before entering the stopping state. Applications running on these instances must be designed to save their state, stop work gracefully, and drain active connections within this short window. Because of this volatility, choose Spot VMs only for fault-tolerant tasks like batch processing or data analytics where a sudden interruption will not cause system failure.
Compute resources require tailored networking configurations depending on their environment. For serverless platforms like Cloud Run or Cloud Functions to access private databases within a VPC, a Serverless VPC Access connector is required to bridge the serverless environment with the private network. In Google Kubernetes Engine (GKE), container pods use Alias IP ranges to receive their own unique IP addresses directly from the VPC subnet, enabling efficient pod-to-pod and external communication without extra translation steps.
Connecting dedicated enterprise workloads requires robust integration solutions. Google Cloud VMware Engine deploys private VMware SDDCs directly onto Google Cloud bare-metal infrastructure, which relies on high-speed private connections to interface with standard Google Cloud services. Traffic flows securely between these environments through designated network boundaries, ensuring that on-premises VMware configurations can scale into the cloud without modifying existing IP addressing plans or management tools.
Managing compute infrastructure at scale requires automated orchestration tools rather than manual setups. Terraform serves as the primary infrastructure-as-code tool, allowing developers to define resources in configuration files and maintain an accurate state of the environment over its lifecycle. When a new virtual machine boots, custom startup scripts execute during the provisioning phase to install dependencies, download application code, and configure initial services before the virtual machine enters the running state.
Once virtual machines are active, administrators use VM Manager to maintain system health, enforce configurations, and apply software updates. This service automates patch management by scanning the fleet for vulnerabilities, scheduling patch windows, and executing updates across multiple operating systems. This structured process reduces security risks and prevents manual configuration drift across large-scale deployments.
For containerized workloads, Google Kubernetes Engine (GKE) provides two distinct operational modes. Choose GKE Autopilot when you want Google to manage the node infrastructure, scaling, and security configurations automatically, leaving your team to focus only on deploying pods. Choose GKE Standard if your team requires deep administrative control over individual virtual machines, custom networking setups, and specific node configurations.
GKE manages compute capacity dynamically through two layers of autoscaling. The Horizontal Pod Autoscaler monitors application metrics like CPU usage and adds more container pods when demand rises. If the existing virtual machines in the Node Pool run out of space to host these new pods, the Cluster Autoscaler automatically provisions new worker nodes to expand the cluster capacity.
Google Cloud offers serverless solutions that eliminate the need to manage virtual machines or container clusters. Choose Cloud Run to deploy stateless web applications and APIs packaged as container images, giving developers flexibility in programming languages and system libraries. Choose Cloud Functions for event-driven architectures where small, single-purpose blocks of code need to execute quickly in response to cloud events, such as file uploads or database updates.
Both serverless platforms automatically adjust resources to match incoming request volume. When traffic spikes, the platform instantly spins up new container instances; when traffic drops, the system scales down, even scaling down to zero instances to eliminate costs when idle. This rapid lifecycle transition ensures that organizations only pay for the exact millisecond compute resources are active.
Prepare and test your skills
Prepare and test your skills
Choose GKE Autopilot when you want Google to manage the node infrastructure, scaling, and security configurations automatically, leaving your team to focus only on deploying pods. Choose GKE Standard if your team requires deep administrative control over individual virtual machines, custom networking setups, and specific node configurations.
Spot VMs offer spare compute capacity at a significant discount but can be reclaimed by Google Cloud at any time, making them suitable only for fault-tolerant tasks like batch processing or data analytics where a sudden interruption will not cause system failure. Standard VMs offer highly available resources that run continuously, making them ideal for core business applications.
A Serverless VPC Access connector is required for serverless platforms like Cloud Run or Cloud Functions to access private databases within a VPC. It bridges the serverless environment with the private network.
GKE uses the Horizontal Pod Autoscaler to monitor application metrics like CPU usage and add more container pods when demand rises. If the existing virtual machines in the Node Pool run out of space, the Cluster Autoscaler automatically provisions new worker nodes to expand the cluster capacity.
StudyCerts
StudyCerts