Run a Metanode

Under Construction

This page is under construction and its content can change. It covers what a metanode runs in its first phase and the hardware it needs. Setup instructions will be added in a later update.

Introduction

In the first phase, a metanode operator is a validator operator who also provides the minimum services that Polkadot Products need:

• Producing blocks for Asset Hub and People Chain as collators.
• Serving smoldot light clients over WebRTC for the relay chain, Asset Hub, and People Chain, which requires additional node configuration covered in a later update.
• Serving remote procedure call (RPC) requests for the relay chain, Asset Hub, and People Chain.

You can run the validator and the collators on separate machines or together on a single machine. Hardware Requirements covers both options.

Hardware Requirements

The following table lists the recommended minimum requirements for each service on its own machine and for a single machine that runs all three services:

Resource	Validator	Asset Hub Collator	People Chain Collator	Single Machine
CPU	8 physical cores at 3.4 GHz	4+ cores (8+ for optimal performance)	4+ cores (8+ for optimal performance)	16 physical cores at 3.4 GHz (24 for optimal performance)
Memory	32 GB DDR4 ECC	32 GB (64 GB for optimal performance)	32 GB (64 GB for optimal performance)	64 GB DDR4 ECC (128 GB for optimal performance)
Storage	2 TB+ NVMe SSD	200+ GB NVMe SSD, pruned	200+ GB NVMe SSD, pruned	2.4 TB+ NVMe SSD
Network	500 Mbit/s symmetric	500+ Mbit/s, stable connection	500+ Mbit/s, stable connection	1 Gbit/s symmetric

Every machine needs a public IP address and must run Linux with kernel 5.16 or later. When you choose hardware, also consider the following details:

• CPU: The validator must meet the processor requirements in Validator Requirements, including disabled simultaneous multithreading. Prioritize single-threaded performance over core count.
• Storage: Prioritize latency over throughput. Fast disk I/O is critical for collator block production. Validator and single-machine storage requirements grow with the chains.
• Network: The collator figures are higher than the 100+ Mbit/s in Run a Block-Producing Collator because metanode collators also serve RPC requests and light clients.
• Relay chain data: To use storage efficiently, point each collator at your validator's RPC endpoint with --relay-chain-rpc-urls, as described in Run a Block-Producing Collator. The collator then reads relay chain data from the validator and runs a lightweight relay chain node instead of keeping a full relay chain database, which removes most of the collator's relay chain database storage.
• Single machine: CPU and storage are the sum of the per-service figures, including 24 cores for optimal performance. Memory and network are lower than the sum because the services do not all reach peak load at the same time: 64 GB of memory instead of 96 GB (128 GB instead of 160 GB for optimal performance), and 1 Gbit/s of network instead of 1.5 Gbit/s.

Validator Hardware Alone Is Not Enough

The hardware listed in Validator Requirements is not enough to also run the collators. Running the validator and the Asset Hub collator on a machine that meets only the validator requirements leaves too few resources for validation.