Blockchain Scalability: The Trilemma, Layer 2 Solutions, Parallel Execution & the Path to Millions of TPS (2025)

Blockchain scalability — the ability to process thousands or millions of transactions per second — is the single biggest technical challenge in crypto. Bitcoin can only handle 7 transactions per second (TPS). Ethereum manages about 15 TPS. Meanwhile, Visa processes 24,000 TPS. To reach mainstream adoption serving billions of users, blockchains need to scale by orders of magnitude — without sacrificing the decentralization and security that make them valuable in the first place. This guide explains the blockchain trilemma, compares every major scaling approach (L2 rollups, parallel execution, sharding, DA layers), and maps the path to millions of TPS.

The Blockchain Trilemma

The blockchain trilemma, coined by Ethereum co-founder Vitalik Buterin, states that a blockchain can optimize for only two of three properties simultaneously:

• Decentralization: Anyone can run a node and validate transactions. Low hardware requirements enable widespread participation.
• Security: The network is resistant to attacks, censorship, and data manipulation. High cost to attack (51% attack cost).
• Scalability: The network can process a high volume of transactions quickly and cheaply.

Every blockchain makes a tradeoff. Bitcoin maximizes decentralization and security at the expense of speed. Solana maximizes speed and security but requires powerful hardware (reducing decentralization). Ethereum L1 prioritizes decentralization and security, then scales through Layer 2 rollups that inherit its security properties.

Blockchain Speed Comparison

Scaling Approach 1: Layer 2 Rollups

Layer 2 rollups are Ethereum's primary scaling strategy. They process transactions off the main chain (Layer 1) and post compressed transaction data back to Ethereum for security. This gives you the speed of a fast chain with the security of Ethereum.

Optimistic Rollups

Optimistic rollups (Arbitrum, Optimism, Base) assume all transactions are valid by default and only check them if someone submits a fraud proof during a dispute window (typically 7 days). This makes them simple and fast to build. Arbitrum is the largest L2 with $10B+ TVL. Base (by Coinbase) is the fastest-growing. The drawback is the 7-day withdrawal period for moving assets back to L1.

ZK-Rollups

ZK-rollups (zkSync, StarkNet, Polygon zkEVM, Scroll) generate mathematical validity proofs that cryptographically guarantee every transaction is correct. No dispute window needed — once the proof is verified on L1, the transactions are finalized. ZK-rollups are technically superior but more complex to build. They offer faster finality and stronger security guarantees than optimistic rollups.

Scaling Approach 2: Parallel Execution

Traditional blockchains process transactions sequentially — one after another. Parallel execution processes independent transactions simultaneously, dramatically increasing throughput:

• Solana: Processes non-conflicting transactions in parallel using the Sealevel runtime. Achieves ~5,000 TPS with 400ms finality.
• Monad: Upcoming EVM-compatible chain with parallel execution — targeting 10,000+ TPS while maintaining full Ethereum compatibility.
• Sei: Parallel DeFi-optimized chain with built-in order book and parallel transaction processing.
• Sui: Object-based parallel execution achieving 120,000+ TPS theoretical. Transactions that touch different objects execute simultaneously.
• Aptos: Block-STM parallel execution engine processing conflicting transactions optimistically.

Scaling Approach 3: Data Availability Layers

The biggest cost for rollups is posting transaction data to Ethereum L1. Data availability (DA) layers provide cheap, dedicated data storage:

• Celestia: The first modular DA layer, providing cheap data storage for rollups at a fraction of Ethereum L1 costs. Celestia separates data availability from execution and consensus.
• EigenDA: DA layer built on EigenLayer, using restaked ETH for security. Cheaper than Ethereum blobs while inheriting Ethereum economic security.
• Ethereum blobs (EIP-4844): Ethereum's native DA solution via proto-danksharding. The Dencun upgrade (March 2024) reduced L2 fees by 90-99%.
• Avail: DA layer from Polygon team, optimized for rollup data publishing.

Scaling Approach 4: Sharding

Sharding splits a blockchain into multiple parallel chains ("shards") that process different transactions simultaneously. Each shard handles a subset of the network's activity, multiplying total throughput. Near Protocol uses 100 shards. Ethereum originally planned execution sharding but pivoted to a rollup-centric roadmap where rollups effectively act as execution shards while Ethereum L1 becomes the shared settlement and DA layer.

Scaling Solutions Compared

Key Takeaways

• The blockchain trilemma forces tradeoffs between decentralization, security, and scalability
• L2 rollups (Arbitrum, zkSync, Base) achieve 2,000-10,000 TPS while inheriting Ethereum security
• Parallel execution (Solana, Monad, Sui) processes independent transactions simultaneously
• Data availability layers (Celestia, EigenDA) reduce rollup costs by 90%+
• Ethereum's Dencun upgrade (2024) cut L2 fees by 90-99% with proto-danksharding
• ZK-rollups offer mathematically guaranteed transaction validity — the gold standard for security
• The future is modular: rollups for execution, DA layers for data, L1 for settlement
• Combined, these approaches put blockchains on a path to millions of TPS

Frequently Asked Questions

What is the blockchain trilemma?

The blockchain trilemma states that a blockchain can only optimize two of three properties: decentralization, security, and scalability. Bitcoin maximizes security and decentralization (7 TPS). Solana maximizes speed and security (5,000 TPS, higher hardware needs). Ethereum prioritizes security and decentralization on L1, then scales through Layer 2 rollups that inherit its security while processing 2,000-10,000+ TPS.

How fast are different blockchains?

Bitcoin: ~7 TPS. Ethereum L1: ~15 TPS. Arbitrum/Optimism/Base L2s: ~2,000-4,000 TPS. ZK-rollups (zkSync, StarkNet): ~2,000-10,000 TPS. Solana: ~5,000 TPS. Sui: ~120,000 TPS theoretical. Monad: ~10,000 TPS (upcoming). Visa comparison: ~24,000 TPS. Note: these are theoretical maximums — real-world throughput varies with network conditions and congestion.

What are the best blockchain scaling solutions?

Layer 2 rollups (Arbitrum, zkSync, Base) for Ethereum ecosystem scaling. Parallel execution chains (Solana, Monad, Sui) for high-throughput monolithic scaling. Data availability layers (Celestia, EigenDA) to reduce rollup costs. No single solution wins — the future combines modular rollups with DA layers and parallel execution. Ethereum's modular approach and Solana's monolithic approach will likely coexist and serve different use cases.

What is a Layer 2 rollup?

A Layer 2 rollup processes transactions off the main blockchain and posts compressed data back to L1 for security. Optimistic rollups (Arbitrum, Base) assume validity unless challenged. ZK-rollups (zkSync, StarkNet) generate mathematical proofs guaranteeing correctness. Both inherit L1 security while achieving 100-1000x higher throughput. L2 rollups are Ethereum's primary scaling strategy.

What is data availability and why does it matter?

Data availability (DA) ensures transaction data is published and accessible for anyone to verify. DA is the biggest cost for rollups — posting data to Ethereum L1 is expensive. Dedicated DA layers like Celestia and EigenDA provide cheap data storage, reducing rollup costs by 90%+. Ethereum's Dencun upgrade introduced blob data (EIP-4844), cutting L2 fees by 90-99%. Full danksharding will expand DA further.