Optimism (OP) — Ethereum L2 Optimistic Rollup on the OP Stack

Consensus mechanism

Optimism does not have its own consensus algorithm in the traditional sense. It is an optimistic rollup, which means transaction ordering and execution happen off-chain on the L2, but security — the guarantee that a malicious operator cannot rewrite history or steal funds — is delegated to Ethereum L1. The flow works like this. A single sequencer (currently OP Labs) receives user transactions, orders them, executes them against the OP Stack state, and produces 2-second blocks. The sequencer batches these transactions and posts the raw data to Ethereum as EIP-4844 blobs on a roughly 30-90 second cadence post-Ecotone; a separate proposer posts state-root commitments. From the moment a state root is posted, a 7-day (604,800-second) challenge window opens: anyone running a verifier node can re-execute the transactions, find a discrepancy, and invoke the Cannon fault-proof game — an interactive bisection protocol that drills down to a single disputed MIPS instruction, executes it on-chain via the MIPS VM contract, and slashes the dishonest proposer. As of June 10, 2024 this challenge is permissionless. Honest finality is therefore inherited from Ethereum PoS (Gasper, 768-second economic finality), with Cannon providing the bridge that enforces the inheritance.

Performance context

Raw throughput numbers — typical 30 TPS, theoretical ceiling 2,000 TPS — sound modest next to Solana's reported 3,000 typical TPS or Hyperliquid's 200,000-TPS ceiling. But those comparisons mislead. Optimism's throughput is bounded by Ethereum blob bandwidth, not by execution capacity, and is shared across the entire Superchain (Base alone routinely processes around 1,500 TPS on the same OP Stack engine). Block time of 2 seconds matches Polygon, Avalanche and Berachain. The number that genuinely matters for users is the L1 data fee per transaction post-Dencun, which for a standard ERC-20 transfer typically runs under one cent — comparable to BSC, an order of magnitude cheaper than Ethereum L1, and the actual driver of activity migration.

Comparison insights

• Versus Arbitrum (the largest peer optimistic rollup): both inherit Ethereum PoS and both share the 604,800-second (7-day) challenge-period finality. The technical gap is in block time and throughput — Arbitrum Nitro runs 250ms sub-second blocks against Optimism's 2-second cadence, and Arbitrum's typical TPS of 40 edges Optimism's 30. Arbitrum's BoLD challenge protocol and Optimism's Cannon fault-proof system reached permissionless states within roughly a year of each other; security models are now broadly comparable.
• Versus Base (built on the same OP Stack codebase): Base inherits OP Stack technology but currently dwarfs OP Mainnet on activity, with roughly $3.8 billion in TVL and a typical throughput closer to 1,500 TPS thanks to aggressive sequencer tuning and Coinbase-driven user acquisition. Critically, Base's reported finality of 780 seconds reflects a different policy choice on withdrawal handling; the underlying fault-proof model remains the same 7-day game.
• Versus Polygon (now POL, with ZK rollup roadmap): Polygon PoS is technically not a rollup at all but a PoS sidechain with Heimdall v2 / CometBFT checkpointing — its 5-second finality is dramatically faster than Optimism's 7-day exit window, but the security model is its own validator set rather than Ethereum L1. TPS-wise Polygon's 60 typical edges Optimism, but the tradeoff is sovereign-chain trust assumptions, not rollup-inherited L1 security.
• Versus Starknet (validity rollup, Cairo VM): Starknet uses STARK proofs for cryptographic finality in approximately 5,400 seconds (90 minutes), an order of magnitude faster than Optimism's challenge window. The cost is non-EVM tooling — Cairo is a different language with different account-abstraction primitives — and currently lower typical throughput (3 TPS vs 30 TPS). Optimism's bet is EVM equivalence and bytecode-level compatibility; Starknet's is cryptographic over economic security.
• Versus Scroll (zkEVM rollup) and X Layer (zkEVM validium on Polygon CDK): Scroll achieves faster finality than Optimism (~3,600 seconds) via Halo2-based validity proofs with bytecode-equivalent EVM, while X Layer uses off-chain data availability — a strictly weaker security model than a true rollup. Optimism's edge is maturity — a five-year operational track record, the OP Stack as the most-forked rollup framework in production, and the Superchain interop framework binding sister chains together. Scroll's ~$10.8M TVL and X Layer's ~$73.9M TVL remain a fraction of activity flowing through OP Stack chains collectively.