Node OperationsCompute Node

Compute Node

A compute node participates in OON (Omne Offchain Network) job execution without running consensus. It syncs the chain, serves RPC, and dedicates resources to offchain compute workloads.

Role vs capability: --role compute is a node role — it determines how the node participates in the network. Storage and Archive are orthogonal capabilities that can be layered on top.

Start a compute node

omne-node start \
  --data-dir ~/.omne/compute \
  --network devnet \
  --role compute \
  --bind 0.0.0.0:9947 \
  --p2p-port 30306 \
  --storage rocksdb \
  --bootstrap-peers "/ip4/<GENESIS_IP>/tcp/30303/p2p/<GENESIS_PEER_ID>"

Register with the OON

After the node syncs, register it as a compute provider:

curl -X POST http://127.0.0.1:9947 \
  -H "Content-Type: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "method": "omne_oonRegisterNode",
    "params": [{
      "operator": "<OPERATOR_ADDRESS>",
      "capabilities": ["wasm", "evm"],
      "bond_ogt": 1000,
      "max_concurrent_jobs": 4
    }],
    "id": 1
  }'

Registration parameters

ParameterDescription
operatorOperator account address
capabilitiesSupported runtimes: wasm, evm
bond_ogtOGT bond amount (minimum: 1,000)
max_concurrent_jobsMaximum parallel job slots

Verify registration

curl -s http://127.0.0.1:9947 \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","method":"omne_oonNodeStatus","params":[],"id":1}' | jq

Compute vs Validator

AspectCompute nodeValidator
--rolecomputevalidator
Consensus❌ Does not produce blocks✅ Produces blocks
OON compute✅ Dedicated✅ Idle capacity only
Staking requiredBond only (1,000 OGT)Validator stake (15–28 OGT) + bond
Best forDedicated offchain workloadsBlock production + opportunistic compute

Combining with other capabilities

A compute node can also serve as an archive or storage node:

# Compute + Archive
omne-node start \
  --role compute \
  --receipt-archive true \
  ...
 
# After startup, also register for OMP storage via RPC

Use cases

  • Dedicated OON workers — run offchain compute jobs (WASM/EVM) without consensus overhead
  • Horizontal scaling — add compute capacity to the network without increasing validator count
  • Specialised workloads — nodes tuned for compute-heavy tasks (AI inference, data processing)