Instances
Skytells virtual machines are provisioned instantly in your chosen region and size. Each instance is a dedicated compute node with full networking identity, health monitoring, and lifecycle management from the Console.
An instance is a virtual machine — a dedicated compute node provisioned and managed from the Skytells Console. Instances are not containers inside a shared runtime; they are standalone machines with their own CPU, memory, storage, operating system, and network identity.
Instances are the foundation of Skytells Infrastructure. You choose where they run, how large they are, and what network and firewall rules they operate under. Skytells handles the hypervisor layer, availability, and the control plane.
What instances are for
Virtual machines are the right choice when you need:
- Full OS access — run any operating system image, kernel configuration, or system-level software.
- Custom runtimes — install languages, tools, and services that are not containerized or that require persistent system state.
- Persistent workloads — long-running compute that should not be tied to a deployment lifecycle: ML training jobs, data pipelines, game servers, build infrastructure.
- Isolated compute — workloads that must not share kernel or system resources with other tenants.
- Agent and automation runtimes — isolated environments for agentic systems that operate browsers, run scripts, or execute human-level tasks. See Eve for how Skytells uses isolated VMs for agent execution.
If your workload is a web service, API, worker, or managed database, Projects is the better starting point — it handles networking, domains, and deployments automatically.
Instant provisioning
Skytells instances are provisioned in seconds. From the moment you confirm a new instance in the Console, the machine is live: IP addresses assigned, networking active, and the instance visible in your list with a real-time status.
There is no queuing and no manual networking configuration — if a VPC and firewall group are already configured, they are attached at creation time.
Provisioning time is measured in seconds, not minutes. Skytells' infrastructure layer is designed for instant availability — you should see your instance reach Running status before you have time to switch tabs.
Deploying an instance
From the Console
- Open the Skytells Console and navigate to Infrastructure in the left sidebar.
- Select All Instances and click New Instance (or the equivalent action in the top-right).
- Choose a region — select the geographic location closest to your users or your other infrastructure.
- Choose a size — CPU cores and memory for your workload. Larger sizes are available for memory-intensive workloads like ML inference or in-memory caches.
- Select or confirm the OS image if prompted.
- Select a VPC to place the instance in. The instance will receive a private IP address within that VPC's address space. See VPCs.
- Optionally attach a firewall group to control inbound and outbound traffic from creation. See Firewalls.
- Review and confirm. The instance will appear in your list with Provisioning status, transitioning to Running in seconds.
Choose the region carefully at creation time. Instances cannot be migrated across regions after provisioning. If you need the same workload in a different region, provision a new instance there.
Instance detail
Each instance has a dedicated detail page in the Console. Navigate to it by clicking the instance name or ID in the All Instances list.
The detail page gives you operational visibility and control:
Health and status
Live status indicator: Running, Stopped, Provisioning, or Error. Review uptime and the last-known health check result.
Metrics
CPU utilization, memory usage, and network I/O over configurable time windows. Useful for right-sizing decisions and identifying workload spikes.
Networking identity
The instance's private IP address (within its VPC), public IP address (if accessible), and the VPC and firewall group it belongs to.
Size and region
The provisioned CPU, memory, and storage configuration, and the region the instance is running in.
Lifecycle controls
Start, stop, restart, or decommission the instance. Lifecycle actions are available from the detail page when the platform exposes them for your account tier.
Networking
Every instance is placed inside a VPC. Instances in the same VPC communicate with each other over private IP addresses — no public routing, no firewall rules required for internal traffic between them.
Public access to an instance is governed by its attached firewall group. By default, no inbound traffic is allowed unless a firewall rule explicitly permits it.
| Traffic type | How it works |
|---|---|
| Instance → Instance (same VPC) | Private IP, no firewall required |
| Internet → Instance | Requires a firewall rule permitting inbound on the relevant port |
| Instance → Internet | Permitted by default unless an outbound rule restricts it |
Availability
Skytells maintains high availability at the infrastructure layer:
- Instances run on hardware maintained and monitored by Skytells.
- The control plane (Console management, status reporting, metrics) is independent from the instance's OS — it continues to function even if the instance OS is unresponsive.
- Region selection distributes exposure across independent availability zones where supported.
For workloads requiring redundancy, provision multiple instances across regions and manage traffic distribution at the network or DNS level.
Related
- VPCs — private network boundaries for instance communication.
- Firewalls — rule groups controlling instance access.
- Infrastructure Overview — when to use instances vs. project apps.
- Eve — Eve runs human-level tasks in isolated VMs using this infrastructure layer.
- Security — platform security practices that govern infrastructure.
How is this guide?
Overview
Skytells Infrastructure gives you instant access to virtual machines, isolated enterprise networks, and VPCs — provisioned in seconds from the Console. It is raw compute and networking, separate from the project-level app and database runtime.
GPU Infrastructure
Skytells provides on-demand and reserved GPU compute across a global network — from individual H100 and A100 nodes to distributed multi-GPU configurations spanning multiple regions. Purpose-built for AI training, inference, and large-scale parallel workloads.