Blockchain Scalability: The Trilemma of Speed, Security, and Decentralization
Why blockchains are slow and what Layer 2 solutions do about it
Throughput vs Decentralization
Bitcoin’s base layer processes on the order of single-digit TPS; Visa quotes peaks near five figures for its network. Public blockchains replicate work across many nodes—that redundancy buys auditability at the cost of raw throughput.
Ethereum’s 2021 congestion pushed swap fees into hundreds of dollars at peaks; that pain accelerated rollups and better fee markets. The question is not “TPS bragging rights” but user cost and latency at sustained load.
Trilemma (Engineering Tradeoffs)
Vitalik Buterin popularized the framing: decentralization, security, scalability—pick stresses, not absolutes. High-TPS chains often raise hardware requirements; maximally decentralized bases may push work to L2.
Layer-1 Approaches
Ethereum’s roadmap included the Merge (September 2022, PoS) and later scaling via blobs (EIP-4844, March 2024) to cut L2 data costs. Solana pursues high single-chain throughput with beefier validators.
Neither path removes engineering tradeoffs—measure outages, client diversity, and operator costs, not slogans.
Rollups
Optimistic rollups post data and allow a challenge window (often ~7 days on Ethereum L1). ZK rollups post proofs—verification cost on L1, prover cost off L1.
Both inherit security assumptions from the settlement layer; understand the bridge and upgrade keys.
App-Specific Chains
Hyperliquid targets on-chain order books and matching with its own consensus (HyperBFT). GaiaEx routes trading to that stack—specialization trades generality for latency-sensitive workloads.
Modular Stacks
Separate execution, settlement, data availability, and proving. Celestia/EigenLayer-class projects focus on DA; Ethereum aims at durable settlement.
Users should judge chains on running costs, uptime, and bridge risk—not marketing TPS. Read incident postmortems; they are more informative than roadmaps.