Virtual machines sized by you, not chosen from a list.
vCPU, memory and SSD move independently. There is no plan that forces you to buy four cores to get eight gigabytes, and no upgrade that rebuilds the machine — you change the shape while it runs and the bill follows within the same month.
27 Aug 2026 2 vCPU · 4 GB RAM · 80 GB SSD
Ubuntu 24.04 LTS 10.1.1.14 • Running
8 Sept 2026 4 vCPU · 8 GB RAM · 160 GB SSD
Ubuntu 24.04 LTS 10.1.1.22 • Running
16 Sept 2026 1 vCPU · 2 GB RAM · 40 GB SSD
Ubuntu 24.04 LTS 10.1.1.31 • Stopped
Figures here are demo data. The shapes and the ratio rules are the real ones.
Every column, including the two you usually have to work out.
Most rate cards give you a monthly number and leave the tax and the hourly equivalent as an exercise. All four are printed here, because all four are on the invoice.
| Shape | vCPU | Memory | NVMe SSD |
|---|---|---|---|
| Nano | 1 | 1 GB | 25 GB |
| Starter | 1 | 2 GB | 50 GB |
| StandardMost deployed | 2 | 4 GB | 80 GB |
| Boosted | 4 | 8 GB | 160 GB |
| Pro | 8 | 16 GB | 320 GB |
Every shape above is the same three multiplications, and any combination inside the limits is valid — these five are only the popular points. The month cap is what an on-demand instance costs if you never switch it off: an hour is (monthly × 1.2) ÷ 730, and the month stops at 120 % of the monthly figure. How the billing works.
Two rules, and then the machine is whatever you say it is.
The bounds are the outside edges of one instance. The ratios exist so a shape is schedulable on real hardware — they are the only thing standing between you and an arbitrary machine.
1 to 16 GB of memory per core
A single core can carry anything from one gigabyte to sixteen. Below that the memory is wasted on a machine that cannot use it; above it, the shape stops fitting a host cleanly. Two cores with four gigabytes — the most-deployed shape — sits at 2 GB per core, comfortably inside.
1 to 50 GB of SSD per GB of memory
The disk is bounded by the memory rather than the core count, because that is what actually tracks a working set. Four gigabytes of memory admits anything from 4 GB to 200 GB of SSD; the 80 GB in the Standard shape is 20 GB per GB of RAM.
Change the shape of a machine you are already running.
No rebuild, no new address, no reinstall. Price the change first, apply it, and pay for the part of the month you actually had the new shape rather than a rounded cycle.
A quote before the commitment
Every resize has a quote call behind it, so the console can show you the new monthly figure and the prorated difference before anything changes. You are never surprised by the arithmetic afterwards, because you saw it first.
Down as well as up
Resizing down credits the unused prepaid portion back to the wallet, immediately usable on anything else. It is a credit, not a cash refund — that is stated plainly rather than discovered at the end of a quarter.
How a machine starts, and how you get into it.
One kept-current template, your own ISO if you need something else, your keys injected at create time, and a browser console for the day SSH is exactly what is broken.
Ubuntu 24.04 LTS, kept current
One customer-selectable OS template, maintained rather than left to rot in a catalogue. An unrestricted OS catalogue is on our own list of things we do not do yet, and we would rather write that here than imply a menu we do not have.
Your own ISO, your own image
Browse the shared ISO library or upload your own, then boot any instance from it to install something we do not ship. The marketplace is a shortcut, not a gate.
SSH keys at create time
Keys stored on the account are injected when the instance is built, and the API can generate a pair as well as import one. No password is emailed to anybody.
Arbitrary cloud-init
Paste any user-data script and it runs once on first boot. Every marketplace stack is exactly this and nothing more, which is why one leaves an ordinary machine behind.
A browser console session
Mint a one-time console URL for an instance and get in without SSH — the thing you need precisely when the network configuration you just applied was wrong.
Rebuild, recycle, restore
Reinstall in place from a template without destroying and redeploying. A deleted instance goes to a recycle bin first, where it can be restored, or expunged with its disks when you mean it.
Numbers with a known age.
A console that shows you a stale figure is worse than one that shows you nothing, because you act on it. Power state is polled every 45 seconds and metrics are sampled every 60, and both intervals are published so you know what you are looking at.
Scheduled power actions
Reboot, stop and start windows on a cron expression in your own account timezone — so a machine only needed in office hours is only billed for them on an on-demand instance. The API returns the next few firings so you can check the schedule means what you think it means.
Alerts that reach something
Threshold rules on instance metrics notify by email or by a signed webhook, with a delivery log and a retry on the webhook. A ticket can be raised against a specific instance, so an alert in your monitoring can open one in ours.
The rest of it, without the paragraphs.
Shape
- 1–64 vCPU, metered per core per month
- 1–512 GB memory, metered per GB
- 10–16,384 GB NVMe SSD, metered per GB
- 1–16 GB of RAM per core
- 1–50 GB of SSD per GB of RAM
- Live resize with a quote first
Lifecycle
- Deploy in about 58 seconds
- Rebuild in place from a template
- Recycle bin with restore
- Expunge with the disks when you mean it
- Scheduled reboot, stop and start
- Start, stop and reboot from the console
Access
- SSH keys injected at create time
- Key pairs imported or generated
- Arbitrary cloud-init user data
- One-time browser console session
- Shared ISO library and your own uploads
- Root on the machine, always
Observability
- CPU, memory, disk and network every 60s
- 14 days of retained history
- Power state polled every 45s
- Threshold alerts by email or webhook
- Per-instance activity log
- Tickets pinned to one instance
Billing
- Committed monthly or on-demand hourly
- Hourly is (monthly × 1.2) ÷ 730
- A month never bills past 120 %
- Renewal anchored to the 5th
- First period prorated per day
- Downgrades credited to the wallet
Automation
- A REST endpoint for every console action
- Idempotency-Key on state changes
- Scoped keys pinnable to one instance
- 202 plus a task handle for long operations
- Signed webhooks on state change
- An account-wide API kill switch
About compute specifically.
Can I really pick any combination of the three?
Any combination inside the bounds that also satisfies the two ratio rules: 1–16 GB of RAM per core and 1–50 GB of SSD per GB of RAM. The console shows the live ratio readout as you move the steppers and refuses the shape before you pay for it, not after.
Does resizing restart the machine?
Resize is applied to the running instance rather than rebuilding it, and the bill is adjusted for the part of the month you had the new shape. Quote the change first and you will see the new monthly figure and the prorated difference before anything happens.
Which operating systems can I run?
Ubuntu 24.04 LTS is the only customer-selectable template today — an unrestricted OS catalogue is something we openly list as not shipped. For anything else, boot the instance from the shared ISO library or from an ISO you upload yourself.
How do I get in if I break SSH?
Mint a one-time browser console session for the instance from the console or over the API. It reaches the machine without going through your own network configuration, which is usually the thing that just broke.
How current are the metrics?
Metrics are sampled every 60 seconds and kept for fourteen days; power state is polled separately every 45. Both intervals are published rather than implied, so you always know how old the number in front of you is.
What does hourly billing actually cost?
An hour is (monthly × 1.2) ÷ 730, and a month never bills past 120 % of the monthly figure. An on-demand instance is capped at 120 % of the same shape’s committed price — so on-demand is a premium for flexibility, capped.
Is there a second zone for failover?
No. ZONE01 is the only zone Comsky runs, and instances, volumes and snapshots stay in it. A second region is on our own list of things we do not do yet; if your design needs one today, we would rather you knew before you deployed.
Start with Comsky Cloud.
Grow into the ecosystem.
One account opens every Comsky product — and one invoice covers them, however many you run.