We publish the ratio.

VPS and VDS are metered by the second and priced by the hour, with a one-hour minimum. A dedicated chassis is billed monthly, in whole months, because it is a machine we rack for you. Twenty minutes costs one hour. Nine hundred hours in a month cost the monthly figure and not a cent more, because the monthly price is a cap. Billing stops at the moment you destroy the instance, not at the end of the day, and there is no charge for a stopped instance's compute — only for the volume it is still holding, at $0.04/GB-month.

You are billed $0.005 a gigabyte for the excess and nothing else changes. Going over an allowance does not throttle your instance, does not suspend it, and is not priced differently at different harbours. Suspension is for non-payment and for the acceptable use policy, and nothing else. Allowances pool across every instance you run in the same harbour, so twenty VPS-1 at Tokyo share 40 TB rather than 2 TB each. Inbound is never counted, and traffic between your own instances across the private ring is never counted. You get an email at 80% of the pool and again at 100%.

You can bring your own: a /24 or larger for IPv4, a /48 or larger for IPv6, announced from AS207214 on a free BGP session on any dedicated chassis, with a letter of authority and an RPKI ROA naming your ASN as origin. You cannot take ours. A Yunagi address belongs to the harbour's /22 and stays in it when you leave — that is true of every provider that owns its own space, and anyone who tells you otherwise is describing a lease they do not control.

Full root or Administrator on all three families from the first boot, with a serial console and a VNC console in the browser for when you lock yourself out. There is no Yunagi agent, no monitoring daemon and no management network interface inside your instance; the only thing we run on your behalf is the anti-spoof filter at the top of the rack, which drops packets whose source address is not yours. On a dedicated chassis you also get the BMC on an isolated jump network, and its credentials are rotated the moment the chassis is released.

Threads, yes, on the VPS families, at four to one, and it is written into the contract rather than into a footnote. Memory, never. Disk capacity, never — we do not thin-provision and we do not sell space we have not bought. On VDS and dedicated the ratio is one to one, and a VDS core that is not exclusively yours is a fault we owe you a credit for.

Anything lawful in the harbour's jurisdiction, including adult content and including a VPN for your own use. Not permitted: open proxies and Tor exit relays, because the abuse volume falls on the /22 everyone else is sharing; anything that scans or attacks a third party; and proof-of-work mining on VPS or VDS, where the threads are shared — mine on a dedicated chassis if you want to, it is your silicon. Outbound port 25 is closed on new accounts and opened on request after thirty days of history. On an abuse report you have 48 hours to respond before we act, except for child sexual abuse material and an attack in progress, which are immediate and final.

There is no minimum term and no notice to give: destroy the instance and billing stops. Monthly prepayment is refunded pro rata to the day for the unused remainder within the first thirty days; annual prepayment is refunded pro rata less ten per cent at any time. From us, you get seven days' written notice before any host maintenance that requires your instance to reboot, and ninety days before a plan is retired — and if a plan is retired, your price holds for the life of the instances already running on it.

vCPU and memory move up and down inside a family with one reboot, about forty seconds. Disk grows in place and never shrinks, because shrinking a filesystem under a running machine is how people lose data; to go smaller, snapshot and redeploy. Moving between families — VPS to VDS, VDS to dedicated — is a snapshot and a redeploy, and support will sit with you and do it at no charge in your first thirty days. Prices change from the second the resize completes, prorated.

A snapshot is a block copy taken on the host your instance is already running on: free within your slots, about thirty seconds for 40 GB, roughly 400 ms of I/O pause, and stored on the same rack. It protects you from your own mistakes. An off-node backup is written nightly to a different rack — or, for five per cent more, to a second harbour — encrypted with AES-256-GCM under a key you may hold yourself, with seven dailies and four weeklies kept. It protects you from the rack. They cost differently because they are different objects, and a page that calls both 'backups' is hiding the distinction.

An engineer, on every account including a $4.20 VPS. There is no paid support tier and no queue that money moves you up. We do not publish a response time, because a number we could always hit would be too long to be worth printing and a number worth printing would be one we would eventually miss. Availability is what the SLA covers; support is not part of it. Status and maintenance windows are posted at status.yunagi.cloud before they are posted anywhere else, and the notice board carries pricing and capacity changes in plain language on the day they take effect.