What is a virtual machine?

Virtual machines use virtualization technology to create virtual hardware—or a virtual version of a computer on a physical machine. The physical machine on which the VMs run is called the host, and the VMs running on the host are called guests. 

Each guest VM runs on an isolated partition on the host, completely separated from other guests. You can host multiple VMs on a single host machine, often a server, running on a software layer known as the hypervisor.

The hypervisor abstracts the host machine’s physical resources, such as CPUs, GPUs, TPUs, memory, storage, or networks, into a pool that can be provisioned and dynamically allocated to guest VMs as needed, providing more flexibility and increasing overall efficiency.

• Type 1: bare-metal hypervisors: These run directly on the host's physical hardware. They're commonly used in enterprise data centers for their high performance and efficiency. For example, Google Cloud uses a KVM-based hypervisor to power its Compute Engine virtual machines.
• Type 2: hosted hypervisors: These run as an application on top of an existing operating system, making them suitable for desktop use. The most common use case for these is for development and testing.

While both VMs and containers are used to isolate applications, they do it in fundamentally different ways. A virtual machine virtualizes the entire physical hardware stack, including the operating system. This makes each VM a self-contained, isolated environment, but it also means that VMs tend to be larger and take up more resources.

By contrast, containers are more lightweight because they virtualize only the OS layer. Instead of bundling a full operating system with each application, a container shares the host's OS kernel. This enables containers to use fewer resources than VMs and start up faster while remaining isolated. This makes containers attractive for new application development. Since many of the applications developed over the last 10 years were written for containers, many workloads including e-commerce, back-office, and AI are “container native.

There are, however, some considerations to keep in mind when running VMs. One of the biggest potential challenges of virtual machines is that running multiple operating systems and a hypervisor layer can come with a performance cost if the host machine isn’t robust enough. In addition, virtual hardware may not be as efficient as the physical hardware of a physical machine. Finally, most cloud providers offer virtual machines that are fixed in the CPU and memory they provide, leading to inefficient usage of resources.

Many of these concerns though can be overcome by choosing to use VMs offered by a cloud service provider. Cloud VMs provide many advantages over traditional VMs since they offer organizations access to the computing power of an entire data center’s worth of computers, rather than a single machine. In addition, Google Compute Engine virtual machines sizes offer custom machine types. Rather than selecting from predefined machine types that might include excess capacity, you can tailor the CPU-to-memory ratio specifically for your workloads, so you only pay for resources you actually use. This targeted approach minimizes waste and can significantly reduce your cloud spend, especially when migrating from on-premises to Google Cloud or from other cloud providers. Compute Engine also delivers virtual machine types optimized for specific customer needs for enterprise workloads, high memory configurations, or demanding workloads like machine learning or high performance computing.

Google Cloud also offers shielded virtual machines for extra security and verifiable integrity of your VM instances. Google Cloud shielded virtual machines leverage advanced platform security capabilities and controls that protect enterprise workloads from threats like remote attacks, privilege escalation, and malicious insiders.