Definition
A virtual machine is an isolated compute instance with its own vCPU, memory, storage, and network identity, provisioned on shared physical hardware and billed for the time it runs.
A virtual machine (VM) is an isolated compute environment that behaves like a standalone server: it has its own operating system, vCPU and memory allocation, storage, and network identity. A hypervisor multiplexes many VMs onto shared physical hardware while keeping them strongly isolated from each other.
How cloud VMs are provisioned
Creating a VM on a modern cloud platform is a three-choice operation:
- Region — where the machine physically runs (for LayerRail: East US, East US 2, Central US, UK South, or South Africa North).
- Boot image — the operating system snapshot the machine starts from, such as Ubuntu, Debian, AlmaLinux, or Rocky Linux.
- Plan — the vCPU, memory, and storage bundle, on shared or dedicated cores, with optional GPU capacity.
The control plane schedules the instance, attaches networking, and returns SSH access — typically within a minute or two.
Billing model
Cloud VMs are duration-billed: the meter starts when provisioning begins and stops when deletion completes. Per-minute billing with a monthly cap keeps both short experiments and always-on workloads predictable.
VMs versus containers
Containers share a host kernel and start faster; VMs carry a full OS and offer stronger isolation plus the freedom to run anything — including Windows, game servers, and kernel-dependent software. Many platforms (LayerRail included) run managed services like PostgreSQL and Kubernetes on top of VM-backed infrastructure.
Cloud providers commonly bill VMs per minute or even per second, so a machine deleted after ten minutes of testing costs a fraction of a cent rather than a full hour.
Usage in APIs
Cloud platforms expose VMs through create, resize, and delete operations in a console, CLI, or API. A create request selects a region, boot image, and plan; the control plane then schedules the instance and returns connection details such as a public IP and SSH access.
Definition and structure
| Isolation | Hypervisor-enforced |
|---|---|
| Resources | vCPU, memory, storage, network |
| Plan Types | Shared, dedicated, GPU |
| Billing | Per minute while running |
Best practices
- Pick the smallest plan that fits your steady-state load and resize when real usage data says otherwise — resizing is cheap, overprovisioning is not.
- Place VMs in the region closest to your users, and keep dependent services in the same region to avoid cross-region latency.
- Delete unused machines promptly; duration-billed resources keep charging until deletion completes.
Historical context
| Introduced | Est. ~1972 |
|---|---|
| Origin | Mainframe virtualization (Virtual Machine) |
| Evolution | Cloud VMs and GPU instances |
Recommended reading
Frequently asked questions
What is the difference between shared and dedicated VM plans?
Shared plans place multiple tenants on the same physical CPU cores, which keeps prices low but allows some performance variance. Dedicated plans reserve cores for your instance alone, giving consistent performance for production workloads.
When should I choose a GPU VM?
Choose a GPU plan when your workload benefits from massively parallel computation — model inference, training, rendering, or transcoding. GPU capacity is metered separately, so delete or downsize the instance when the job finishes.
Do stopped VMs still cost money?
On most platforms a stopped VM still reserves storage and often the compute slot itself, so charges continue until the resource is deleted. Check your provider's billing rules — on LayerRail, billing stops when deletion completes.
Build your next layer.
Run compute, networking, and managed services from one project-centered console.