Google Cloud Platform (GCP) provides a comprehensive ecosystem of cloud services spanning compute, storage, databases, networking, security, and data analytics. As a fully-managed cloud provider, GCP abstracts infrastructure complexity while giving developers fine-grained control over resources when needed. Two critical mental models underpin GCP services: global resources (like IAM and Cloud Storage) that span regions, and regional/zonal resources (like Compute Engine and Cloud SQL) that remain within specific geographic boundaries. Understanding this distinction helps you architect for both performance and compliance—global services optimize for availability and durability, while regional services optimize for latency and data residency requirements. With Google Cloud Next '26 introducing the Agentic Data Cloud, AI-native capabilities are now embedded across databases, analytics, and compute, making GCP's AI/ML integration deeper than ever.
What This Cheat Sheet Covers
This topic spans 14 focused tables and 170 indexed concepts, 166 flashcards. Below is a complete table-by-table outline of this topic, spanning foundational concepts through advanced details.
A jump-to index of every table row in this cheat sheet.
An interactive map of every table and concept in this topic.
Table 1: Compute Services
VMs, managed instance groups, and specialized compute options form the foundation of most GCP workloads. Machine family selection—general-purpose, compute-optimized, memory-optimized, or accelerator-optimized—drives both performance and cost more than almost any other decision.
| Service | Example | Description | |
|---|---|---|---|
gcloud compute instances create my-vm --machine-type=e2-medium --zone=us-central1-a | • Customizable virtual machines running Linux or Windows • supports predefined, custom, and specialty machine types across N1, N2, E2, C2, C3, C4, M-series, and GPU/TPU-attached families | ||
n2-standard-4 (4 vCPUs, 16 GB)e2-micro (2 vCPUs, 1 GB) | Predefined vCPU/memory configurations: general-purpose (E2, N1, N2, N2D, N4), compute-optimized (C2, C3, C4), memory-optimized (M1, M2, M3, M4), and accelerator-optimized (A2, A3, A4, G2) families. | ||
gcloud compute instance-groups managed create my-mig --template=web-template --size=3 | Collection of identical VMs managed as a unit with autoscaling, auto-healing, load balancing, and rolling updates. | ||
gcloud compute instance-templates create web-template --machine-type=n2-standard-2 | • Reusable VM configuration blueprints defining machine type, boot disk, and network settings • required by MIGs for autoscaling | ||
--provisioning-model=SPOT | • Interruptible instances at up to 91% discount with no maximum runtime limit • can run indefinitely until Google reclaims capacity; ideal for fault-tolerant batch jobs. | ||
--preemptible flag | • Short-lived instances at up to 80% discount that Google terminates with 30-second notice • max 24-hour runtime; Spot VMs are the preferred successor with no runtime cap. | ||
gcloud batch jobs submit --config=job.json | • Fully managed batch job scheduler provisioning VMs automatically per job • no custom job-scheduler setup required; ideal for HPC, ML training, and large-scale data processing. | ||
--shielded-secure-boot --shielded-vtpm | Hardware-rooted VM security with Secure Boot, vTPM, and Integrity Monitoring to defend against rootkits, bootkits, and malicious firmware tampering. | ||
--confidential-compute flag | • Encrypts VM memory in-use using AMD SEV or Intel TDX hardware-level encryption • prevents cloud provider, hypervisor, or co-tenant access to data in RAM. | ||
gcloud compute disks create my-disk --size=100GB --type=pd-ssd | • Network-attached block storage persisting independently of VM lifecycle • available as Standard HDD (pd-standard), Balanced SSD (pd-balanced), SSD (pd-ssd), or Extreme SSD (pd-extreme). | ||
Attached directly to VM host | • Physically attached ephemeral NVMe storage with sub-millisecond latency and highest IOPS • data lost when instance stops; never use for data requiring persistence. | ||
gcloud compute disks snapshot my-disk --snapshot-names=backup-1 | • Incremental point-in-time backups of Persistent Disks stored in Cloud Storage • first snapshot is full, subsequent ones capture only changed blocks | ||
gcloud compute sole-tenancy node-templates create | • Physically isolated servers where only your VMs run • meets compliance and bring-your-own-license requirements that prohibit multi-tenancy | ||
--custom-cpu=6 --custom-memory=20GB | • User-defined vCPU and memory configurations when predefined types don't fit • charged per vCPU-hour and GB of RAM |