Migrate to Virtual Machines for the PCA Exam: Runbook, BYOL, Benchmarking

GCP Study Hub
Ben Makansi
February 14, 2026

Lift-and-shift migrations show up on the Professional Cloud Architect exam in a way that catches people off guard. The questions are not about deep cutover engineering. They are about whether you know the named GCP service for the job, the operational practices that wrap around it, and how licensing carries over. I want to walk through the four ideas you should have ready: the Migrate to Virtual Machines service itself, the migration runbook, BYOL, and benchmarking.

Migrate to Virtual Machines

Migrate to Virtual Machines is the GCP service for moving VMs from on-premises data centers, VMware environments, Amazon EC2, or Azure VMs into Compute Engine. The service runs an automated workflow that supports live migration with minimal downtime, so you do not have to shut the source VM down to make the move.

The important property for the exam is that it preserves what is already on the VM. Operating system, installed applications, configuration, and data all carry over. You are not re-architecting the workload. You are picking it up from one environment and dropping it onto Compute Engine. That is the textbook definition of a lift-and-shift, and Migrate to Virtual Machines is the named service that performs it.

If you see a question describing a customer with a fleet of on-prem VMs or AWS EC2 instances who wants to move to Google Cloud without rewriting their applications, Migrate to Virtual Machines is the answer. The distractors will often be unrelated services, or they will suggest re-platforming work that is not warranted.

Migration Runbook

A migration runbook is a step-by-step guide that documents the migration procedure end to end. It is not a GCP product. It is an operational practice that the Cloud Architect role is expected to be familiar with.

The runbook does four things:

  • Standardizes the procedure so multiple migrations follow the same path.
  • Identifies dependencies between systems so you do not knock out a downstream service when you cut over an upstream one.
  • Defines fallback strategies if a migration fails partway through.
  • Defines troubleshooting steps so the team is not improvising under pressure.

If a question describes an organization preparing to migrate a large fleet and asks what they should produce before starting, a migration runbook is a defensible answer. The exam treats it as a baseline best practice for anyone doing migrations at scale.

BYOL (Bring Your Own License)

BYOL is the practice of carrying existing software licenses with the VM during migration. The classic example is a Windows Server license attached to an on-prem VM. When you migrate that VM to Compute Engine, you bring the license along instead of paying for a new one bundled with the GCE instance.

Mechanically, you specify the license during import or during instance creation. The licensing terms remain intact and compliant in the new environment. This applies whether you are coming from on-premises, Azure VMs, or AWS EC2.

The exam framing here is usually cost optimization or compliance. If the customer already paid for licenses and wants to avoid double-paying after migrating, BYOL is the correct posture. You are not buying new licenses bundled into Compute Engine. You are migrating the workload and the license together.

Benchmarking

Benchmarking is the pre-migration practice of measuring your current system's performance and running controlled tests in Compute Engine to determine the right configuration on the target side. It exists because the obvious migration plan, "match the old machine specs," is often wrong. Workloads behave differently on different hardware, and you do not want to over-provision or underutilize.

The process has three steps:

  1. Evaluate the current CPU and memory usage of the source VMs or servers. This is your baseline.
  2. Deploy Compute Engine instances with specifications similar to the source. Run realistic load tests against them. Install the Ops Agent so Cloud Monitoring and Cloud Logging can capture the performance data you need to analyze the results.
  3. Adjust based on what you measure. If the standard predefined machine types do not match your workload, use custom machine types to fine-tune CPU and memory allocation independently.

Custom machine types are the lever that makes benchmarking actionable. If the workload sits awkwardly between two predefined shapes, you do not have to settle. You build a custom shape that fits.

An exam question on benchmarking will usually describe a customer who migrated and is now seeing performance issues, or one who is planning a migration and wants to right-size the target environment. The answer involves measuring the source workload, testing on Compute Engine with the Ops Agent, and tuning the machine type, possibly to a custom configuration.

How these four ideas fit together

The Professional Cloud Architect exam treats VM migration as a small but recurring topic. The pattern is consistent. Migrate to Virtual Machines is the service that does the move. The migration runbook is the operational artifact you produce before you do the move. BYOL handles licensing during the move. Benchmarking determines the target configuration around the move. If you can name each of those and explain what it is for, the migration questions on the exam are straightforward.

My Professional Cloud Architect course covers VM migration alongside the rest of the compute material.

arrow