Ethereum mainnet launched on 30 July 2015. Ten years on, running the nodes behind it is a different discipline from the early days. Here is what we think operators should take from the decade.
Between launch and today the network moved from proof of work to proof of stake, client software multiplied, and chain state kept growing. Tessellate Labs is new, so this is an engineering reading of public history, not a record of our own operations. Each section ends with how the lesson shapes the way we design infrastructure.
A short, practical look at Ethereum’s first decade from the point of view of the people who run its nodes: what changed, and what stayed hard.
This article covers four lessons that have shaped how node infrastructure is planned and operated:
- Launch to the Merge: Ethereum mainnet launched on 30 July 2015. In September 2022 the Merge moved it to proof of stake, so a node became two programs working together, an execution client and a consensus client. Operators had to learn to run, upgrade and monitor both.
- Client diversity: A bug in client software run by most of the network can affect most of the network. Client diversity has been a recurring operational concern, which is why we run Reth with Lighthouse and Geth with Prysm on Ethereum.
- Archive state growth: Archive state growth shapes hardware planning. Mainnet archive state is of the order of 3 TB, so we serve history and the debug_ and trace_ namespaces from a Reth archive tier on local NVMe.
- Reliability as the difference: See why reliability is the difference as node infrastructure consolidates around a handful of providers. When many applications depend on a few endpoints, one outage is felt widely, and measured, published availability is what lets teams choose with evidence.
- What to plan for: Plan for continued protocol upgrades, client releases that need staged rollout, and rollup chains running alongside Ethereum itself. Practise restores, keep a second client available, and budget for state growth.