Skip to main content

config-hardware

Hardware requirements for Arbitrum chains

Arbitrum chains, including Arbitrum One, Nova, and customizable Arbitrum L2/L3 chains, have varying hardware needs depending on the node type. Requirements are generally modest compared to Ethereum L1 nodes, with a focus on accessibility for decentralized participation. The specs are for nodes handling limited RPC requests; scale up for high traffic. Full nodes form the base, with validators and sequencers building on them. Archive nodes require more storage for historical data.

Key considerations:

  • OS: Linux or macOS recommended; Docker is commonly used for setups.
  • Bandwidth: At least 100 Mbps for syncing and RPC; higher for sequencers to minimize latency.
  • Storage: NVMe SSD preferred for speed; data grows over time (e.g., Arbitrum One chain data ~1TB+).
  • Scaling: Increase CPU/RAM for multiple RPC requests or during disputes (e.g., BoLD challenges). Use cloud providers like AWS t3.xlarge for baseline.
  • Arbitrum chain-specific: Storage requirements are similar. If running AnyTrust, you will need to run your own DAS or use a third-party DA service provider.
  • General Tip: Start with minimums and monitor; overprovision for reliability. No internet access needed post-sync, but initial sync requires a stable connection.

Full node (base for most setups)

Runs the chain state, serves RPCs, and syncs data. Non-archive by default.

ComponentMinimumRecommendedNotes
CPU4 cores (multi-core processor)8+ cores (e.g., Intel i7 or AMD Ryzen 7)Single-core performance key; upgrade if node lags. For limited RPC.
RAM16 GB32 GB+Scale with RPC load; more for simultaneous requests.
Storage1 TB NVMe SSD2 TB+Depends on chain traffic; Arbitrum One ~1TB, grows ~100GB/month.
Bandwidth100 Mbps1 GbpsFor initial sync (days to weeks).

Validator node

Monitors assertions and challenges fraud. Runs on top of a full node with staker software enabled. BoLD upgrade makes it more efficient, assuming one honest party.

ComponentMinimumRecommendedNotes
CPU4 cores8 coresNeeds fast computation for disputes; BoLD reduces multi-challenge load.
RAM8 GB16-32 GBSimilar to full node; more if running defensive mode.
Storage1 TB NVMe SSD2 TB+Includes chain data
Bandwidth100 Mbps500 Mbps+For monitoring and bonding (e.g., 3,600 ETH equivalent stake).
OtherBonding funds; WASM supportVPS/cloud for uptimePermissioned for low-TVL chains; permissionless needs higher bonds/resources.

Sequencer node

Orders transactions, posts batches. Centralized in main chains (run by Offchain Labs), but customizable for Arbitrum chains (e.g., via RaaS like Gelato/Caldera). Focus on low latency; hardware similar to full node but optimized for throughput.

ComponentMinimumRecommendedNotes
CPU4-8 cores16+ coresHigh clock speed for tx processing; latency-sensitive.
RAM16 GB32-64 GBHandles mempool and batching; more for high TPS Arbitrum chains.
Storage1 TB NVMe SSD2 TB+For logs and state; fast I/O critical.
Bandwidth1 Gbps10 Gbps+Low-latency network; private mempool integration.
OtherDocker/Nitro stackDedicated server/cloudIn Arbitrum chains using AnyTrust, lower due to DAC; self-host or RaaS for control.

Archive node

Optional for full history querying; builds on full node.

ComponentMinimumRecommendedNotes
CPU4 cores8+ coresSimilar to full, but more for pruning disabled.
RAM16 GB32 GB+-
Storage2 TB SSD4 TB+Historical data balloons size (e.g., 2-5x full node).

Decisions to Make:

  • Node Type: Full for basic, validator for security contribution, sequencer for Arbitrum chain sovereignty.
  • Chain Mode: Storage requirements are similar for Rollup or AnyTrust.
  • Hosting: Self-host (control, cost) vs. RaaS/cloud (ease, fees). Start local, scale to VPS.
  • BoLD Enablement: For validators, reduces hardware during disputes but requires an upgrade.
  • Monitoring: Use tools like Prometheus; budget for growth (e.g., +20% resources yearly).
  • When to Upgrade: If CPU >80% or sync lags, add resources. Test on Sepolia.

Consult official docs for updates; requirements evolve with traffic.