Node operators

Run a node on hardware you already operate.

Sell capacity.

Your hardware is paid for.

UniNet runs as one program on your machine. Tenants get containers. The host stays yours.

At the bottom, your machine. Above it runs an ownership boundary: below that line nothing changes, and the host OS, root and remote access stay yours. Above it, the UniNet node, whose top is the capacity you publish, cut into cells. Idle cells are drawn as outlines. Reservations arrive, claim an idle cell, and a tenant's container settles into it with its workload inside; from time to time a tenant leaves and its room is freed. Busy containers pay your identity, by the chain and in UNE. No rate or yield is shown, because none is published.

Operator questions

Six questions, answered.

Current state, not target state.

What runs on my machine?

The UniNet node, a single program, and the hosted chains and containerized jobs scheduled through it.

Can it be isolated?

Container-level at most, and only on a Linux host with crun or runc. No confidential computing - you, holding root, remain inside the boundary.

How much of the box does it take?

You publish CPU, memory and storage, and choose shared or dedicated allocation.

How is reliability recorded?

As a score against your identity, from the blocks your node signed. Placement skips a node that scores too low and prefers higher scores.

How do I get paid?

By the chain, in UNE, to your identity: a share of the fees of each block your node signs, and a daily charge from every chain that reserves room on it. The rates are a test network’s.

What if a tenant disappears?

A chain that stops paying is frozen at the first unpaid day, and after fourteen your node frees the room. What those days cost stays owed.

Placement

How a chain picks its nodes.

Full nodes and low scores are skipped. Of the rest, the most reliable are chosen, and the reservation lands on all of them in one step.

Eight nodes, each showing how much of its published capacity is already reserved and its reliability score out of 10,000. A new chain asks for room on three. Full nodes are skipped, and so are nodes scoring below 9,000; of the rest, the three most reliable are chosen. The reservation lands on all three in one step, each answers with a signed grant, and the chain then pays its hosts by the day. The scores shown are illustrative; the threshold is the protocol's.

Evidence

What is built, and how far.

Read from the same source as the build status page.

Capability	On your hardware	Status
Reliability scoring	A score recorded against your identity.	Available
Workload isolation	Container-level at most; needs crun or runc. Not hardware-backed.	Available
Allocation	Shared or dedicated, per machine.	Available
Atomic reservation	Capacity inspectable across nodes; reservations are atomic.	Available
Lifecycle control	Start, stop and remove containers on your node.	Available
Payment	Fee shares and daily hosting charges, paid by the chain to your identity.	Available

Every subsystem: build status. Setup: running a node.

Metering and payment

UNE is the unit of account for resources.

Reserved capacity is charged in it by the day and paid to the nodes that hold it. The rates are a test network's settings, so no price target or yield figure is published. This is not investment material.

A node's life in five stages. It is installed; 1,000 UNE is staked behind it, as behind every validator; the validators vote it in and it takes a seat; it hosts chains and signs blocks, and its reliability rises with each block it signs; and it earns a share of the fees of each block it signs and a daily charge from each chain it hosts, paid to your identity in UNE. A validator that votes twice in one round loses half its stake and its seat. That is the only thing slashed: downtime costs a node its reliability score, not its stake. No rate or yield is shown, because none is published.

Talk to an engineer What works today