Skip to content

Overview to running nodes on Celestia

There are many ways you can participate in the Celestia networks.

Celestia node operators can run several options on the network.

Consensus:

  • Validator node: This type of node participates in consensus by producing and voting on blocks.
  • Consensus node: A celestia-app full node to sync blockchain history.

Data Availability:

  • Bridge node: This node bridges blocks between the Data-Availability network and the Consensus network.
  • Light node: Light clients conduct data availability sampling on the Data Availability network.

You can learn more about how to set up each different node by going through each tutorial guide.

Data availability nodes

Non-archival data availability nodes

Node typeMemoryCPUDiskBandwidth
Light node500 MB RAMSingle core20 GB SSD56 Kbps
Bridge node64 GB RAM8 cores8 TiB NVME1 Gbps

Archival data availability nodes

Node typeMemoryCPUDiskBandwidth
Light node (unpruned headers)500 MB RAMSingle core111.7 KB/block56 Kbps
Bridge node64 GB RAM8 cores160 TiB NVME*1 Gbps

*Archival disk requirement is based on the current v6 32MB/6s throughput at maximum capacity. In reality the requirement can be much lower and we advise regularly checking disk usage and having at least 1 month worth of max throughput of extra disk space (8TiB).

Consensus nodes

Non-archival consensus nodes

Node typeMemoryCPUDiskBandwidth
Validator32 GB RAM16 cores2 TiB NVME1 Gbps
Consensus node32 GB RAM16 cores2 TiB NVME1 Gbps

When upgrading to 128mb/6s, validators must use hardware that passes the CPU benchmark. If your server does not pass, upgrade to a more powerful machine.

For a list of CPUs tested for 128mb/6s, see the release notes. Other CPUs are acceptable if they pass the benchmark. Recommended CPU specs:

  • 32 or more cores
  • GFNI (Galois Field New Instructions) support
  • SHA-NI (SHA New Instructions) support

Archival consensus nodes

Node typeMemoryCPUDiskBandwidth
Consensus node64 GB RAM16 cores160 TiB NVME*1 Gbps

*Archival disk requirement is based on the current v6 32MB/6s throughput at maximum capacity. In reality the requirement can be much lower and we advise regularly checking disk usage and having at least 1 month worth of max throughput of extra disk space (8TiB).

Please provide any feedback on the tutorials and guides. If you notice a bug or issue, feel free to make a pull request or write up a Github issue!