Google Compute Engine

GCP's IaaS offering allowing one to spin up virtual machines and networks of virtual machines. Google manages the hardware layer, we are fully responsible for software layer.

• innovative pricing
  • sub-hour billing
    • paying by the minute with 10 minute minimum
    • discounted pricing for sustained usage
  • preemptible instances
    • up to 70% cheaper
    • short-lived - up to 24 hours before being terminated
• robust networking features
• instance metadata and startup scripts
• wide range of machine types
  • can specify custom machine attributes
• HTTP and network load balancing
• Google Cloud Monitoring
  • create uptime checks
  • get notified when there are issues
  • creat dashboards from generated metrics
• persistent disk snapshots
  • makes it easy to backup or migrate data
• APIs for auto-scaling and group management
  • group management allows for pools of virtual machines (instance groups) based on a template
  • auto-scaling allows for machines to be automatically scaled within an instance group based on specified metrics

Compute Engine Resources

• instances provide CPU and memory
• standard, SSD, and local SSD persistent disks
  • standard and SSD are network attached
  • local SSD is directly attached storage used for intensive workloads
• images for Linux, UNIX, and Windows Server
• support for disk snapshots
• networks with firewall rules
  • networks allow one to create many virtual machines that communicate with each other
  • firewall rules dictate traffic coming in or out of the network
• forwarding rules for load balancing
  • balance traffic across the network

When we spin up a Compute Engine virtual machine instance we:

• define a region and zone for the instance
• define the physical specifications of the machine
• attach persistent disks to the machine
• attach a boot disk with an OS, bootloader, and file system

Compute Engine in developer console

Compute Engine (opens in a new tab)

In the navigation:

• VM Instances
  • list of instances for the current project
• Instance groups
  • managing pools of VMs
• Instance templates
  • templates for spinning up instance groups
• Disks
  • persistent data attached to instances
  • disks can be created independently and then attached to instances
• Snapshots
  • backed up data
• Images
  • available OSs to configure VMs with
• Metadata
  • project-wide metadata
  • each instance has its own metadata, but project-wide metadata can be configured here that all instances have access to
• Health checks
  • used with load balancing

Networking in developer console

VPC Networks (opens in a new tab)

Compute Engine Billing

• instances are charged for while running (and can be switched off)
• disks are charged for bt allocated amount (not amount in use)
• networks are charged for on certain operations
  • ingress (inbound traffic) generally has no charge
  • egress (outbound traffic) it depends
    • same zone, generally no charge
    • between zones / regions charges vary depending on geographical location
• pricing calculator (opens in a new tab)
• Compute Engine pricing (opens in a new tab) is region-specific

CLI - gcloud compute

Cloud SDK includes gcloud compute for managing GCE resources, e.g.

# create a compute instance
$ gcloud compute instances create <instance_name> --zone <zone> --project <id>
#   [1]    [2]      [2]    [          3         ]
# 1 - gcloud SDK
# 2 - compute command
# 3 - command group / resource
# 4 - commands
 
# get all compute command groups / resources
$ gcloud compute -h
 
# get all operations that can be performed on a specific command group / resource
$ gcloud compute instances -h

Accessing Instances

• access Linux and UNIX instances of SSH

  • access via Cloud SDK

    $ gcloud compute ssh <instance>

  • over standard SSH

  • from the developer console

• access Windows Server instances over Remote Desktop Protocol (RDP)