Google Cloud Compute
Google Cloud compute services run applications as virtual machines, Kubernetes workloads, managed containers, or queued jobs. You choose how much infrastructure control to keep and how much operation to hand to Google.
itCloud computing | OpenSkills.info
Course pathWalk it in order
Look it upDip in anytime
Go furtherLeaves this page
Don't Panic
Don't Panic — Google Cloud Compute
Google Cloud compute is the collection of places where your software gets to run while everyone pretends the word “compute” settled the matter. It did not. The useful question is what kind of work has arrived: a virtual machine that needs an operating system, a Kubernetes Pod, a container serving requests, or a finite task that must finish and depart politely.
Compute Engine is the control-heavy end of the arrangement. It gives you a VM and hands you its image, disks, network, service account, and the small but persistent matter of operating it. For one machine that can be fine. For a fleet, use an instance template and a managed instance group, then remember that autoscaling adds capacity and autohealing replaces failed capacity. Neither mechanism quietly rescues data that lived on the machine it replaced.
GKE is for work that genuinely needs Kubernetes: Pods, Kubernetes APIs, and its workload model. A container image alone is not an admission ticket. Cloud Run is the managed-container choice when an application fits its contract: services receive requests or events, jobs finish tasks, and worker pools run continuous non-HTTP work. Its instances are disposable, which is a wonderfully efficient way to discover whether important state was hiding on local disk.
Batch belongs to finite, queueable computation. It provisions compute for tasks and removes it after the job ends; it is not an interactive application wearing a queue-shaped hat. Across all four models, identity and reachability remain separate: a network path decides whether traffic can arrive, while a service account decides what the workload may do.
Start with the Intro for the responsibility ladder and each service's contract. Use the Slides to compare the four execution models and their failure boundaries. Keep the Cheatsheet nearby when you need the fleet pattern, state-placement rules, scaling signals, or Spot VM fit. Then try the Exercise: the answer matters less than being able to explain why the alternatives do not fit.
Where this skill leads
Relevant careers
See how this topic contributes to broader role-level skill maps.
Sources
- https://docs.cloud.google.com/docs/compute-area/choose-compute-options
Supports
- Google Cloud compute selection begins with workload requirements and the required level of operating system control, orchestration, and scaling
- Compute Engine provides self-managed operating system control, GKE provides Kubernetes, Cloud Run provides serverless managed execution, and Batch provides job-queue orchestration
- https://docs.cloud.google.com/compute/docs/overview
Supports
- Compute Engine is an infrastructure service with self-managed virtual machine and bare metal instances
- Compute Engine supports Linux and Windows operating systems, durable storage, and management through console, command line, API, and client libraries
- https://docs.cloud.google.com/compute/docs/regions-zones
Supports
- Zones are deployment areas within regions
- Resource location and scope affect availability and failure design
- https://docs.cloud.google.com/compute/docs/disks
Supports
- Compute Engine offers durable network block storage and local disk choices with different persistence, replication, and performance behavior
- Disk placement and backup behavior must match the workload's durability requirement
- https://docs.cloud.google.com/compute/docs/instance-groups
Supports
- Managed instance groups create and maintain matching VMs from an instance template
- Managed instance groups support autohealing, autoscaling, regional multi-zone deployment, and automatic updates
- Unmanaged instance groups do not provide the same automated fleet services
- https://docs.cloud.google.com/compute/docs/autoscaler
Supports
- A managed instance group autoscaler adds VMs when load grows and removes VMs when demand falls
- Autoscaling behavior is defined by the group's autoscaling policy
- https://docs.cloud.google.com/load-balancing/docs/load-balancing-overview
Supports
- Cloud Load Balancing distributes traffic across multiple application instances
- Google Cloud offers application and network load balancers with internal, external, regional, and global scopes
- Load balancers route traffic to healthy backends
- https://docs.cloud.google.com/compute/docs/access/service-accounts
Supports
- A service account is an identity used by an application or compute workload
- Applications on VMs can use attached service account credentials without embedded secret keys
- Google recommends a user-managed service account with only the required IAM roles
- https://docs.cloud.google.com/compute/docs/instances/spot
Supports
- Spot VMs use excess Compute Engine capacity at lower prices
- Compute Engine can reclaim Spot VMs at any time
- Spot VMs fit fault-tolerant workloads that can withstand preemption
- https://docs.cloud.google.com/kubernetes-engine/docs/concepts/kubernetes-engine-overview
Supports
- GKE is a managed implementation of Kubernetes for containerized applications
- Google manages the Kubernetes control plane
- Autopilot manages worker nodes while Standard provides more infrastructure control
- Kubernetes workloads run as Pods on cluster nodes
- https://docs.cloud.google.com/kubernetes-engine/docs/concepts/autopilot-overview
Supports
- Autopilot is a managed GKE operating mode focused on workloads
- Google manages nodes, scaling, security, and other infrastructure settings in Autopilot
- https://docs.cloud.google.com/run/docs/overview/what-is-cloud-run
Supports
- Cloud Run is a managed application platform for code, functions, and containers
- Cloud Run provides service, job, and worker-pool resource types
- Cloud Run services handle requests and events, jobs run tasks to completion, and worker pools handle continuous non-HTTP work
- Cloud Run container instances are disposable and their writable file systems are not durable
- Cloud Run supports revisions, traffic rollout, scaling, service identity, and Google Cloud integrations
- https://docs.cloud.google.com/batch/docs/get-started
Supports
- Batch schedules, queues, and executes finite processing workloads
- A Batch job contains tasks and runnables plus the resources needed to execute them
- Batch provisions and removes Compute Engine resources for jobs
- https://docs.cloud.google.com/monitoring/docs/monitoring-overview
Supports
- Cloud Monitoring provides metrics, dashboards, alerting, and integrations for Google Cloud resources and applications
- Compute Engine application metrics and logs can be collected through the Ops Agent
- https://docs.cloud.google.com/monitoring/agent/ops-agent
Supports
- The Ops Agent is the primary telemetry agent for Compute Engine instances
- The Ops Agent collects logs, metrics, and traces
- https://github.com/GoogleCloudPlatform/awesome-google-cloud
Supports
- The fetched list includes Functions Framework, Online Boutique, Bank of Anthos, and Awesome Cloud Run as compute ecosystem resources
- Functions Framework is described as an open source set of libraries for portable Google Cloud functions
- Online Boutique is described as a multilingual microservices sample running on GKE
- Bank of Anthos is described as a GKE sample integrating several Google Cloud products
- https://github.com/steren/awesome-cloud-run
Supports
- The fetched list curates Cloud Run documentation, quickstarts, interactive tutorials, samples, languages, tools, and deployment resources
- https://cloud.google.com/blog/products/compute/5-best-practices-compute-engine-cost-optimization
Supports
- Compute Engine launched in 2012
- https://docs.cloud.google.com/compute/docs/release-notes-archive
Supports
- Compute Engine Autoscaler became available in beta in November 2014
- https://cloud.google.com/blog/products/containers-kubernetes/google-clouds-container-platform-for-the-next-decade-of-ai
Supports
- Google Container Engine launched in 2015
- https://cloud.google.com/blog/products/serverless/announcing-cloud-run-the-newest-member-of-our-serverless-compute-stack
Supports
- Cloud Run entered beta in April 2019 as a managed environment for stateless HTTP-driven containers
- https://cloud.google.com/blog/products/serverless/knative-based-cloud-run-services-are-ga
Supports
- Cloud Run became generally available in November 2019
- https://cloud.google.com/blog/products/compute/google-cloud-spot-vm
Supports
- Spot VMs entered preview in October 2021 and use excess capacity that can be reclaimed
- https://cloud.google.com/blog/products/compute/new-batch-service-processes-batch-jobs-on-google-cloud
Supports
- Batch entered preview in July 2022 as a managed service for scheduling batch jobs
- https://cloud.google.com/blog/products/serverless/cloud-run-jobs-and-second-generation-execution-environment-ga
Supports
- Cloud Run jobs became generally available in April 2023 for run-to-completion work
- https://cloud.google.com/blog/products/application-development/run-your-ai-inference-applications-on-cloud-run-with-nvidia-gpus/
Supports
- Cloud Run added NVIDIA L4 GPU support in preview in August 2024
- https://cloud.google.com/blog/products/compute/compute-engine-managed-instance-groups-get-scale-in-controls
Supports
- Abrupt scale-in can conflict with workloads whose instances take many minutes to initialize
- https://cloud.google.com/blog/products/compute/google-cloud-spot-vm-use-cases-and-best-practices
Supports
- Spot VM workloads need fault tolerance, preemption handling, and durable progress outside interruptible compute
- https://cloud.google.com/compute
Supports
- Google Compute Engine is Google Cloud's virtual machine product
- https://cloud.google.com/kubernetes-engine
Supports
- Google Kubernetes Engine is Google Cloud's managed Kubernetes product
- https://cloud.google.com/run
Supports
- Cloud Run is Google Cloud's managed container product
- https://cloud.google.com/batch
Supports
- Google Cloud Batch is Google Cloud's batch scheduling product
- https://aws.amazon.com/ec2/
Supports
- Amazon EC2 provides virtual machine compute
- https://aws.amazon.com/eks/
Supports
- Amazon EKS provides managed Kubernetes
- https://aws.amazon.com/fargate/
Supports
- AWS Fargate provides managed container compute
- https://azure.microsoft.com/products/virtual-machines/
Supports
- Azure Virtual Machines provides virtual machine compute
- https://azure.microsoft.com/products/kubernetes-service/
Supports
- Azure Kubernetes Service provides managed Kubernetes
- https://azure.microsoft.com/products/container-apps/
Supports
- Azure Container Apps provides managed container application hosting