Layer 2 Scaling: Rollups and Channels
2026-06-16
Base-layer blockchains are deliberately slow. Every full node re-executes and stores every transaction, which is what makes them trustless, and also what caps throughput at a few to a few dozen transactions per second. You cannot run global payments on that directly. Layer 2 (L2) systems fix this by moving execution off the main chain while still anchoring security to it.
The Blockchain Trilemma
The framing most people use: a chain wants decentralization, security, and scalability, and naive designs can only have two. Making the base layer faster usually means fewer nodes can keep up, eroding decentralization. L2s sidestep the trilemma by keeping the slow, decentralized, secure base layer as a settlement court, and doing the high-volume work on top.
Rollups: Execute Off-Chain, Settle On-Chain
A rollup processes hundreds or thousands of transactions off-chain, then posts a compressed batch plus enough data back to the base chain so anyone can reconstruct the state. Users get the base chain's security guarantees because the data and the dispute mechanism live there. The two flavors differ in how they convince the base chain the batch was honest.
Optimistic rollups assume the batch is valid by default and open a challenge window (typically about a week). If anyone spots fraud, they submit a fraud proof, the bad batch is reverted, and the cheater is penalized. Cheap and EVM-compatible, but withdrawals back to L1 wait out the challenge period unless you use a liquidity provider.
zk-rollups post a validity proof (a zero-knowledge proof) with every batch, mathematically demonstrating the new state is correct. No challenge window is needed because correctness is proven, not assumed, so finality and withdrawals are faster. The cost is heavier proof generation and historically harder EVM compatibility, both improving quickly.
The Real Bottleneck: Data Availability
Rollups are cheap at execution but still must publish their data so the world can verify and reconstruct state. Posting that data to L1 became the dominant cost. Ethereum's "blob" upgrade (EIP-4844) added a dedicated, cheaper space for rollup data, slashing L2 fees. Data availability, making sure the data was actually published, is now the frontier of scaling research.
Payment Channels (Lightning)
A different approach for payments: two parties lock funds in a shared on-chain contract, then exchange signed balance updates off-chain, instantly and for free, settling to the chain only when they close the channel. Connected channels form a network (Bitcoin's Lightning Network) where you can route a payment through intermediaries without trusting them. It is superb for high-frequency micropayments, less suited to general computation.
How to Think About It
- Base layer (L1): the slow, maximally secure settlement court.
- Rollups: general-purpose scaling that inherits L1 security via fraud or validity proofs.
- Channels: instant, near-free payments for parties that transact repeatedly.
The endgame most builders now assume is a modular stack: a secure base layer handling settlement and data, with the vast majority of activity happening on rollups above it. That is how a chain settling a handful of transactions per second becomes infrastructure for millions of users.