What you need
| Minimum | Notes | |
|---|---|---|
| Disk | whatever you can spare | cap it with --max-storage; beyond it new shards go to the next node |
| Memory | about 100 MiB | measured: 23–30 MiB idle; the rest is page cache for stored data |
| CPU | any 64-bit core | measured: 1–8 millicores idle |
| Network | one reachable UDP port (default 7700) | today: a public IP or a forwarded port |
| OS | Linux with Docker, or anything Rust builds on |
Behind a home router? Today your node needs its UDP port forwarded. Nodes that peers can’t reach are left out of their routing tables and get no shards. Hole punching and relays (no port forwarding) are the next milestone.
Join the testnet
- Ask for an invite. You get access to the source and image, the testnet issuer key (so your node only stores paid data) and the control-plane URL.
- Start the node with Docker. It runs under
slab-runner, which applies signed releases and rolls back a release that keeps crashing:
docker run -d --name slabd --restart unless-stopped --network host \
-v slab-data:/data -v "$PWD/issuer:/keys:ro" \
slab:<tag> slab-runner slabd --data-dir /data --listen 0.0.0.0:7700 \
--issuer-key /keys --ctl-url https://<ctl-url> --max-storage 100G
Or from source, without Docker:
cargo build --release --bin slabd
target/release/slabd --data-dir ~/slab-data --listen 0.0.0.0:7700 \
--issuer-key issuer.pub --ctl-url https://<ctl-url> --max-storage 100G
The node finds the network through bootstrap.slab.me.uk, saves its peers and rejoins from them after a restart.
- Forward UDP 7700 on your router if the node is behind NAT. If the public port differs, add
--announce-port <port>. - Register for (simulated) payouts. Your node signs a statement naming the account that gets paid. It never sees that account’s secret:
docker exec slabd slabd --data-dir /data register-proof --account <your account id> > node.proof
slab wallet --file op.json register-node node.proof --stake-cents 100
What happens next
- Your node joins the routing table and starts receiving shards of slabs whose addresses are near its ID.
- Once per epoch (a day on the testnet) the control plane challenges it to prove it holds what it claims. Passing nodes are paid for storage, and for shards they serve.
- Missing three audits in a row, or answering with a bad proof, costs half the stake. Restarts are fine: a node never restarts for an update while an audit answer is due.
FAQ
Can I see what my node stores? No. Data is encrypted on the uploader’s machine before it is cut into shards. Your node holds pieces of ciphertext under content addresses, and no node holds enough pieces to rebuild a file on its own.
Is there a token? No. Prices are in euros (simulated during the testnet). There is no chain, no gas and no wallet on the node.
Can I run several nodes? Yes, but Slab caps how many of a slab’s shards one host holds (6 of 16 today), and is moving to the same limit per operator. Several nodes on one machine add capacity, not more pay per slab.
What if I stop? Repair notices missing shards and rebuilds them on other nodes. Stop cleanly
(docker stop) so the node finishes what it is doing.