In this lesson
- find out why most transactions today do not run directly on Ethereum
- understand the difference between an optimistic and a ZK rollup
- learn to judge how much a layer 2 network relies on Ethereum
Ethereum's main network is like a motorway with a toll gate: safe, but when everyone drives alone it is expensive and jammed. A rollup is a bus. It picks up hundreds of passengers, goes through the gate once and they split the toll.
Why layer 2#
Ethereum processes only a limited number of transactions per second – after all, every node computes everything. Instead of making the main network bigger (and nodes would need expensive hardware), computation moved to layer 2 networks (L2). Ethereum serves as judge and archive.
How a rollup works#
A rollup does three things:
- It takes transactions from users and orders them.
- It computes the result – who holds how much now.
- It sends to Ethereum the packed transaction data and the result, so it can be checked at any time.
Thanks to the data on Ethereum the rollup's state can be rebuilt even if its operator disappears – as long as someone saved the data before it was deleted (blobs, see lesson 3).
Optimistic vs. ZK rollup#
| Optimistic rollup | ZK rollup | |
|---|---|---|
| how it is checked | the result is assumed valid; whoever finds an error challenges it | a mathematical validity proof for every batch |
| withdrawal to Ethereum | after a challenge window, usually 7 days | once the proof is verified, usually hours |
| examples (Oct 2026) | Base, Arbitrum One, OP Mainnet | Starknet, Linea, ZKsync Era |
Fast bridges get around the seven-day wait of optimistic rollups – for a fee someone pays you out immediately and waits themselves. Bridges are the next lesson.
The sequencer: one place where things can jam#
Transactions on a rollup are ordered by a sequencer. On all big rollups today it is a single server run by the operator. When it goes down, the rollup stops:
- Base: 5 Aug 2025 for 33 minutes, 25 Jun 2026 for over an hour.
- Arbitrum: in December 2023 for about an hour and a quarter during a rush of transactions.
The money is not lost during an outage, but you cannot move it.
Security levels#
The L2BEAT website rates rollups by how much they depend on people:
- Stage 0 – the operator can do practically anything; a "rollup with training wheels".
- Stage 1 – a proof system works, but a security council can override it.
- Stage 2 – only the code decides.
As of 11 October 2026: Base, Arbitrum One and OP Mainnet are Stage 1. None of the big rollups is Stage 2.
Layer 2 = extra risk
Money on an L2 is as safe as Ethereum plus the rollup's code, its sequencer and the people who can change it. For larger amounts check on L2BEAT which stage the rollup is at.
Common myths#
- "An L2 is a different blockchain." A rollup stores data on Ethereum and relies on its security.
- "I can withdraw from an L2 to Ethereum immediately." From an optimistic rollup the standard is 7 days.
- "Rollups are fully decentralised." All the big ones have a single sequencer and a security council.
Summary
- Layer 2 networks process transactions outside Ethereum and send it the data and the result.
- An optimistic rollup waits for challenges (7 days); a ZK rollup proves validity with a proof.
- The sequencer is usually a single server – when it fails, the rollup stops.
- Stages 0–2 show how much a rollup depends on people; the big ones are Stage 1.