GCP for DevOps — Machine Types & Images

3 minute read

Machine Types & Images

Choosing the right machine type and image is critical for cost, performance, and reliability. GCP offers a wide range of predefined and custom machine types, plus public and custom images.

Machine Type Families

┌─────────────────────────────────────────────────────┐
│              Machine Type Families                   │
├───────────┬───────────┬───────────┬─────────────────┤
│ General   │ Compute   │ Memory    │ Accelerator     │
│ Purpose   │ Optimized │ Optimized │ Optimized       │
├───────────┼───────────┼───────────┼─────────────────┤
│ E2        │ C2        │ M1        │ A2 (GPU)        │
│ N2        │ C2D       │ M2        │ G2 (GPU)        │
│ N2D       │ C3        │ M3        │                 │
│ N1        │ H3        │           │                 │
│ T2D       │           │           │                 │
│ T2A (Arm) │           │           │                 │
├───────────┼───────────┼───────────┼─────────────────┤
│ Web apps  │ CI/CD     │ Databases │ ML training     │
│ Dev/test  │ Gaming    │ In-memory │ Rendering       │
│ APIs      │ HPC       │ caching   │                 │
└───────────┴───────────┴───────────┴─────────────────┘

Common Machine Types for DevOps

Machine Type	vCPUs	Memory (GB)	Use Case
`e2-micro`	0.25	1	Free tier, tiny workloads
`e2-small`	0.5	2	Small dev servers
`e2-medium`	1	4	Dev/test, light services
`e2-standard-2`	2	8	General purpose
`e2-standard-4`	4	16	Build agents, CI/CD
`e2-standard-8`	8	32	Medium workloads
`n2-standard-4`	4	16	Production workloads
`c2-standard-8`	8	32	Compute-intensive builds

# List available machine types in a zone
$ gcloud compute machine-types list --zones=us-central1-a \
    --filter="name~'^e2'" \
    --format="table(name,guestCpus,memoryMb)"

NAME             GUEST_CPUS  MEMORY_MB
e2-highcpu-16    16          16384
e2-highcpu-2     2           2048
e2-highcpu-32    32          32768
e2-highcpu-4     4           4096
e2-highcpu-8     8           8192
e2-medium        1           4096
e2-micro         2           1024
e2-small         2           2048
e2-standard-16   16          65536
e2-standard-2    2           8192
e2-standard-32   32          131072
e2-standard-4    4           16384
e2-standard-8    8           32768

Custom Machine Types

When predefined types don’t fit, create a custom machine type with exact CPU and memory.

# Create a VM with a custom machine type (6 vCPUs, 24 GB RAM)
$ gcloud compute instances create custom-vm \
    --zone=us-central1-a \
    --custom-cpu=6 \
    --custom-memory=24GB \
    --image-family=debian-12 \
    --image-project=debian-cloud

# Custom with extended memory (more memory per CPU than standard)
$ gcloud compute instances create high-mem-vm \
    --zone=us-central1-a \
    --custom-cpu=4 \
    --custom-memory=40GB \
    --custom-extensions \
    --image-family=debian-12 \
    --image-project=debian-cloud

Public Images

GCP provides maintained images for popular operating systems.

# List available image families
$ gcloud compute images list --format="table(family,name,project)" \
    --filter="family~'debian|ubuntu|centos|rocky'" | head -20

FAMILY              NAME                                  PROJECT
debian-11           debian-11-bullseye-v20260301          debian-cloud
debian-12           debian-12-bookworm-v20260301          debian-cloud
ubuntu-2004-lts     ubuntu-2004-focal-v20260301           ubuntu-os-cloud
ubuntu-2204-lts     ubuntu-2204-jammy-v20260301           ubuntu-os-cloud
ubuntu-2404-lts     ubuntu-2404-noble-v20260301           ubuntu-os-cloud
rocky-linux-8       rocky-linux-8-v20260301               rocky-linux-cloud
rocky-linux-9       rocky-linux-9-v20260301               rocky-linux-cloud

# List Container-Optimized OS images (for running containers)
$ gcloud compute images list --filter="family:cos" \
    --format="table(family,name,project)"

FAMILY              NAME                                  PROJECT
cos-stable          cos-stable-109-17800-xxx              cos-cloud
cos-beta            cos-beta-113-18244-xxx                cos-cloud

Custom Images

Create reusable images with your tools pre-installed.

# Step 1: Create a VM and install your software
$ gcloud compute instances create image-builder \
    --zone=us-central1-a \
    --machine-type=e2-standard-2 \
    --image-family=ubuntu-2204-lts \
    --image-project=ubuntu-os-cloud

# Step 2: SSH in and install software
$ gcloud compute ssh image-builder --zone=us-central1-a
# (Inside VM)
$ sudo apt-get update && sudo apt-get install -y docker.io kubectl helm
$ exit

# Step 3: Stop the VM
$ gcloud compute instances stop image-builder --zone=us-central1-a

# Step 4: Create an image from the stopped VM's disk
$ gcloud compute images create devops-base-v1 \
    --source-disk=image-builder \
    --source-disk-zone=us-central1-a \
    --family=devops-base \
    --description="Ubuntu 22.04 with Docker, kubectl, Helm"

Created [devops-base-v1].

# Step 5: Use the custom image family
$ gcloud compute instances create build-agent \
    --zone=us-central1-a \
    --machine-type=e2-standard-4 \
    --image-family=devops-base \
    --image-project=my-project

Boot Disk Types

Disk Type	ID	IOPS (Read)	Throughput	Use Case
Standard PD	`pd-standard`	0.75/GB	120 MB/s	Dev/test, low cost
Balanced PD	`pd-balanced`	6/GB	240 MB/s	General purpose
SSD PD	`pd-ssd`	30/GB	480 MB/s	High performance
Extreme PD	`pd-extreme`	Configurable	2,400 MB/s	Databases

# Create a VM with an SSD boot disk
$ gcloud compute instances create fast-vm \
    --zone=us-central1-a \
    --machine-type=e2-standard-4 \
    --boot-disk-size=100GB \
    --boot-disk-type=pd-ssd \
    --image-family=debian-12 \
    --image-project=debian-cloud

Machine Type Recommender

GCP automatically analyses VM usage and recommends right-sizing.

# View recommendations
$ gcloud recommender recommendations list \
    --project=my-project \
    --recommender=google.compute.instance.MachineTypeRecommender \
    --location=us-central1-a \
    --format="table(content.overview.resourceName,content.overview.recommendedMachineType.machineType,primaryImpact.costProjection.cost.units)"

RESOURCE_NAME  RECOMMENDED_MACHINE_TYPE  COST_SAVINGS
build-agent    e2-standard-2             -$24/month
web-server     e2-medium                 -$12/month

Common Mistakes

Mistake	Why It’s Wrong	Fix
Using N1 for new workloads	E2/N2 are newer and cheaper	Use E2 for general, N2 for production
Over-provisioning machine types	Wasting money	Use Recommender to right-size
Not using image families	You pin to a specific image version	Use `--image-family` for latest
Using `pd-standard` for CI/CD	Slow disk = slow builds	Use `pd-ssd` or `pd-balanced`
Not creating custom images	Reinstalling software every time	Bake images with required tools

Quick Reference

Command	Description
`gcloud compute machine-types list --zones=ZONE`	List machine types
`gcloud compute images list`	List public images
`gcloud compute images create NAME --source-disk=DISK`	Create custom image
`gcloud compute images list --project=my-project`	List custom images
`--custom-cpu=N --custom-memory=MGB`	Custom machine type
`--boot-disk-type=pd-ssd`	SSD boot disk

Next → Firewall Rules

Share on

Twitter Facebook LinkedIn