The State of Ethereum Scalability

Published in

Carbonocom

15 min readJan 20, 2023

💡 With 15 to 45 transactions per second while serving a growing ecosystem of decentralized applications, Ethereum is very far from becoming the global supercomputer it aims to be. Ethereum’s transactions are slow, expensive, or both- a big obstacle in the way of massive adoption. A problem of this size brings an equally significant opportunity for improvement. The race towards scalability is currently taking different technical routes that fall into three categories: on-chain improvements through sharding and off-chain optimistic and zero-knowledge rollups. All with innumerable challenges and nuances that make it hard to predict who will be the winner(s).

This report was originally published in August 2022. Download the PDF version here. Since the publication, The Merge took place successfully, and Consensys entered the zero-knowledge race with their own testnet.

Intro

Ethereum’s major upgrade, The Merge, is finally about to happen. Initially planned to start in 2019, Ethereum’s transition has undergone significant delays. Today there’s an almost certain deployment month: The Merge has successfully passed all rehearsals to date, and it is almost ready for the big leap in the second half of September.

Some people mistakenly believe that The Merge is going to improve Ethereum’s scalability and decrease fees. But the only way The Merge will help push the scalability agenda is by giving way to the next step of the roadmap, The Surge, where developers will finally tackle throughput.

Ethereum developers have been aware for years that they would not be able to provide the desired improvements in transaction speed and costs in the short term through on-chain scalability implementations. It has been clear that the fastest path to scalability would have to happen through off-chain solutions: the development of rollups.

Rollups are secondary layers that process transactions and bundle them together, then settle those bundles on Ethereum’s main chain as single transactions. Different rollups have different approaches to data compression (to decrease the size of information transmission) and transaction processing (to reduce the computing power required to validate operations).

“Rollups move computation (and state storage) off-chain but keep some data per transaction on-chain. To improve efficiency, they use a whole host of fancy compression tricks to replace data with computation wherever possible.” An Incomplete Guide to Rollups

Currently, there are three main routes in the scalability race. Sharding is Ethereum’s on-chain solution and will be the main course of The Surge. But the deployment date (2023 at the earliest) makes it a non-viable solution for the present needs. Optimistic rollups (ORs) and zero-knowledge rollups (ZK-rollups or ZKRs) are the two most important works in progress: off-chain Layer 2s with different approaches to data transmission and validation and with varying flavors of Ethereum compatibility. Optimistic rollups are enjoying a headstart at the moment, while ZK-rollups are trying to get up to speed.

The pace of improvements has increased lately. Optimistic rollups are working hard to grow their userbase and ecosystem of developers and dApps, in many cases with superb product and marketing moves. ZK-rollups, more challenging technically, were expected to take longer to become viable products. Still, in recent weeks, three projects from Scroll, Matter Labs, and Polygon have announced upcoming launches of their respective zero-knowledge general-purpose rollups.

With The Merge around the corner, ready to free Ethereum developer time for scalability improvements, with the growth of optimistic rollups led by Optimism and Arbitrum, and the up-and-coming launch of the first all-purpose zk-rollups by Scroll, Matter Labs, and Polygon, the scalability race is about to enter warp speed.

Scalability context and concepts

Ethereum Roadmap

Ethereum’s current roadmap is a five-year plan expected to make Ethereum scalable, secure, and sustainable to “bring Ethereum into the mainstream and serve humanity.”

The Merge. Currently approaching its final stretch, The Merge will change Ethereum’s Proof of Work validation method to Proof of Stake. The transition will reduce energy consumption and set the stage for future scaling.
The Surge. Introduces sharding: the splitting of Ethereum’s database to increase scalability.
The Verge. The introduction of verkle trees will reduce node size, which in turn promotes decentralization and scalability.
The Purge. Reduces the hard drive space needed for validators. Making validation less costly increases the likelihood of decentralization.
The Splurge. Adds a miscellaneous set of other improvements.

EVM compatibility and EVM equivalence

The Ethereum Virtual Machine, or EVM, is a virtual software stack that runs in Ethereum’s nodes. The EVM is similar to server software, installed in the decentralized network of Ethereum nodes instead of in physical servers. The EVM is what gives the Ethereum blockchain its programmability because it allows the execution of transactions and smart contract code.

The ability to interact as frictionlessly as possible with Ethereum’s Virtual Machine is an outstanding feature of any Layer 2 solution.

Vitalik Buterin identifies up to five approaches to rollup equivalence with the EVM. Still, they can be summed up into two groups:

EVM compatibility guarantees that applications deployed on an L2 can communicate successfully with the L1. Yet, it involves some low-level code customization, which creates differences between the EVM-compatible applications and the Ethereum code. Customizations are tradeoffs projects need to decide on, where lower compatibility can sometimes improve performance but reduce security and decentralization as rollups move away from canonical Ethereum rules.
EVM equivalence, on the other hand, means that rollups are developed in full compliance with Ethereum’s specifications. EVM equivalence guarantees that everything that can run on an EVM equivalent L2 can also run seamlessly on the L1. This reduces friction and guarantees full interoperability, but it might raise performance issues, especially in the case of ZK-rollups.

Validation 101

A blockchain operation (any transaction or interaction with a smart contract) can be explained as a validated transition of states.

The state is the snapshot of the current data on a blockchain — for example, the current balance of funds for all blockchain users. When Alice sends 2 ETH to Bob, a new block is added to the blockchain, including the latest information.

For that change to be guaranteed and secured by the protocol, the blocks need to be validated before the operations it contains are considered final and are immutably added to the blockchain. Validation includes checking block data and metadata to find inconsistencies or falsehoods. Validation answers questions like if state changes are possible (do senders have enough funds?) if PoW or PoS has worked properly, if timestamps are consistent, as well as other operations with the data and metadata of blocks that help perform security checks.

One of the main differences between Layer 2 approaches is validation methods. It is one of the areas where developers make decisions and tradeoffs between speed, performance, and reliability.

The three leading scalability solutions

On-chain: Sharding

Sharding is the on-chain scaling solution chosen by Ethereum. Sharding consists of splitting a database into fragments called shards and distributing them to spread the load.

Sharding has always been Ethereum’s weapon of choice in the path to scalability, but the Ethereum Foundation excluded its development from The Merge to avoid delays. It remains, nevertheless, the core contribution of the next phase, The Surge.

In the meantime, developers brought up some less complex approaches to sharding. Danksharding and proto-danksharding are simplified approaches to sharding that are currently in discussion.

Proto-danksharding, a streamlined version of danksharding, is already being debated as an Ethereum Improvement Proposal, EIP-4844. It would implement a new transaction type that would significantly reduce transaction speeds and costs of current Layer 2s. This, along with a proposed new data type, blobs, would help optimize the data availability problem of rollups.

Off-chain: Optimistic rollups

Optimistic rollups are an off-chain scalability solution. They are a Layer 2 whose main feature is that they consider all transactions valid by default, which reduces computation and therefore increases the ability to process more transactions.

On the other hand, it requires more extended withdrawal periods.

Validation is performed as a side process by external agents. These agents monitor both the L1 And L2 blockchains in search of fraudulent transactions and earn rewards if they find and report one. The protocol gives external agents seven days to review the data, which slows down the refund process: users who want to take their funds from the L2 to the L1 have to wait for that 7-day grace period to give time to the independent agents to check the validity.

Off-chain: ZK-rollups

Zero-knowledge proof is a somewhat recent mathematical concept that allows a prover to demonstrate to a verifier that certain information is true without revealing how the prover has obtained this information. Zero-knowledge proofs add tremendous value in terms of blockchain privacy, as, for example, they allow transactions without disclosing the source address or balance. This method reduces the need for data, which enables ZK-rollups to process many more transactions, even than their optimistic counterparts. On the other hand, they require highly costly computation that relies on complex, new technology that is even, in some cases, still in the research phase.

The zero-knowledge approach allows ZK-rollups to perform validation before settling transactions on the mainnet. Validity proof drastically improves the recovery delay times over-optimistic fraud proofs. With transactions properly validated, there is no need for a 7-day grace period, which makes withdrawals immediate. But fraud proofs are a bigger technical challenge, with major software and even hardware requirements responsible for the delay in deploying total zero-knowledge solutions.

The scalability ecosystem today

It is generally considered that ZK-rollups will be the long-term winner in the scalability race. Validity-proof costs are expected to decrease, which, combined with other features like the aforementioned privacy and better composability, make ZK-rollups a stronger candidate for the future.

But we should not underestimate the advantages of a headstart. Optimistic rollups, led by Arbitrum and Optimism, launched in 2021, have been attracting projects, developers, and funding at a faster pace, and they have been battle-tested for months.

ZK-rollups to date have only been developed as app-specific solutions, like dYdX (derivatives trading), Loopring (trading & payments), Immutable X (gaming). General-purpose zero-knowledge blockchains are only starting to be announced by projects like Polygon, Scroll, and Matter Labs.

Optimistic rollups also have advantages in terms of Ethereum equivalence. Some features of the Ethereum protocol make zk-prove processes heavy in terms of computation, while Arbitrum and Optimism, the most prominent optimistic rollups, claim full EVM equivalence. In the meantime, EVM equivalence is only starting to show up in the upcoming ZK-rollups announced by Polygon and Scroll (Matter Labs’ zkSync will only achieve EVM compatibility, not equivalence).

Finally, some collateral improvements, like EIP-4844, are expected to level the playing field between the two off-chain scalability approaches. EIP-4844 simplifies data availability for rollups, making the fraud-proof process much lighter for optimistic rollups.

The scalability race is incredibly complex and intricate, with projects making progress concurrently, which affects the evolution of other projects. It is impossible to predict, at this stage, what the future will look like. In the meantime, these are today’s most prominent players in the OR vs. ZKR ecosystem.

Polygon

Polygon is one of the oldest scalability projects in the ecosystem. It was founded as Matic Network in 2017 as a plasma-driven scaling approach with a Proof-of-Stake sidechain. It is mainly known for Polygon PoS (technically a sidechain, not a Layer 2), well known as a fast, reliable, and cheap alternative to Ethereum. Over time, Polygon has become a “Swiss army knife” of scaling solutions.

Polygon offers a set of products in different stages of development that expand Ethereum’s features in ways that range from Proof of Stake sidechains (Polygon PoS) to zero-knowledge implementations in various stages of development (Hermez, Nightfall, Zero), to development frameworks (Polygon Edge).

MATIC is Polygon’s native token and is used to pay transaction fees and staker rewards in Polygon PoS.

Polygon has been the chosen layer for many mainstream brands’ immersion into Web 3. They are partners with Meta, Disney, Reddit, or HTC.

Optimism

Optimism was born from Plasma Group, a non-for-profit research organization investigating ways of scaling Ethereum. In 2020 Plasma turned into OptimismPBC, where PBC stands for Public Benefit Corporation, “a corporation created to generate social and public good, and to operate in a responsible and sustainable manner.”

Optimism has an innovative approach to decentralization. In May 2022, together with the announcement of a token launch, the Optimism PBC unveiled its roadmap toward fully decentralized governance fueled by the OP token. The OP is used to pay fees and reward sequencers (the L2 equivalent of miners), and as the governance token of the Optimism Collective.

Optimism’s L2 was initially deployed on the mainnet in January 2021. It is an optimistic rollup that could potentially process up to 40,000 transactions per second. Right before the launch on mainnet, the Optimism protocol upgraded to OVM 2.0, an improvement focused on EVM equivalence that reduces all friction with Ethereum.

Arbitrum

Arbitrum was developed by Offchain Labs, a research and development team founded to build a scaling suite for Ethereum. Arbitrum’s L2, launched in 2021, can allegedly process up to 40,000 transactions per second.

Arbitrum also uses an optimistic approach to validation. Fraud disputes are settled off-chain and reduced to a single step within a transaction. This makes Arbitrum’s conflict resolution smaller and cheaper than Optimism’s but less secure and decentralized (because it runs off-chain and is based on custom code instead of on the Ethereum EVM).

Arbitrum uses the Arbitrum Virtual Machine, an EVM-compatible development that allows interoperability with Ethereum, but that is not EVM equivalent, unlike Optimism.

Arbitrum does not have a token: transaction fees are paid in ETH.

In April 2022, Arbitrum launched Arbitrum Odyssey. In this ambitious marketing initiative, teams were invited to participate in a team competition that involved bridging and using Arbitrum to access monetary and NFT rewards. The subsequent usage spike led to an increase in fees mirrored Ethereum’s, to the point that Odyssey had to be halted.

StarkNet

StarkNet is a zero-knowledge Layer 2 built by Israeli software company StarkWare.

StarkNet enables general computation smart contracts that support composability. StarkNet operates as a rollup that adds state diff data (the differences between initial and final states) on-chain. This allows them to improve Ethereum’s TPS (StarkNet reported a record 3k TPS in January 2021 and claims to be able to process up to 300k).

StarkNet is built with Cairo programming language, with means lower compatibility with Ethereum. Because of its technical complexity, StarkNet has been mostly used as an app-specific L2, instead of a general purpose one that allows all types of smart contract deployments.

dYdX was the most popular application built with StarkWare’s software, but it will soon move to a custom-made Cosmos-based L1. The reasons given publicly had to do with StartWare’s lack of decentralization, but some speculations say the cost of using StarkWare was the main driver (projects need to pay StarkWare to use their proprietary software). StarkWare has announced a roadmap to decentralized governance and a token launch since.

zkSync

ZKSync is an app-specific, zero-knowledge rollup that settles transactions on Ethereum, introduced by Dutch developing company Matter Labs in 2019, and launched on the mainnet in 2020.

zkSync supports payments, token swaps, and NFT minting. In its current form, it does not support smart contracts, so it is generally used as an app-specific rollup solution.

During its almost two years of life, it has helped reduce the cost of transfers by ~50x compared with Ethereum and secured close to 4M transactions. Its most prominent user has been the open-source projects grant community, Gitcoin.

zkSync is EVM-compatible, and it uses a proprietary development framework, Zinc. These features separate zkSync’s current version from full EVM equivalence.

zkSync has announced plans for decentralization and the launch of a token, but they both seem halted while the team works on zkSync 2.0, their next iteration of the current product.

Next up: Polygon, Matter Labs, and Scroll unveil their zkEVM projects

During the same week in late July 2022, three unaffiliated projects announced new evolutions of their ZK-rollups. Polygon and Matter Labs, already current actors in the scalability scenario, and Scroll, a newcomer to the race.

Polygon released the code for Polygon zkEVM, their EVM-equivalent zero-knowledge rollup, intended to take their ZK-rollup’s throughput improvements and add to it the excellent performance derived from full compatibility with the Ethereum Virtual Machine. Polygon’s zkEVM will be able to run any transaction, any smart contract faster and cheaper than Ethereum without sacrificing security or decentralization.

Matter Labs initiated a 100-day countdown to the launch of zkSync 2.0 on mainnet. ZkSync 2.0 is an iteration of zkSync launched on testnet in early 2022. zkSync is EVM-compatible and will support Solidity and Vyper, programming languages used by Ethereum developers. The code will be fully released by the end of 2022.

Scroll announced the “pre-alpha version of Scroll for external testers” intending to start the conversation with developers and users. Scroll has worked directly with the PSE (Privacy and Scaling Explorations) group at the Ethereum Foundation, and it is holding its zkEVM rollup to the highest standards of compatibility -EVM equivalence- and decentralization.

Insights

There is no one-size fits all solution neither for users or developers who need to decide what scalability solution to invest money or effort in. Every use case, whether it is getting a loan from a DeFi protocol or launching an NFT collection, will need an individual evaluation of the current situation and possible future before making a decision. It is possible that a smaller set of projects will take the leading position in the future, making decisions easier. But we are nowhere near that situation, and very few elements can help us predict a likely outcome.

It has been frequently stated that ZK-rollups are the favorite candidate. They provide higher and safer throughput and desired side effects like privacy protection. But at the moment, they are underdeveloped and still depend on R&D before becoming an available solution for general use.

In the meantime, Optimism and Arbitrum are building momentum, attracting users, developers, and dapps, and creating real-world use cases. And the compatibility with Ethereum, especially in Optimism’s case, drastically reduces the friction of deploying a project on their OR.

Momentum is a crucial success factor in this race, as network effects can reduce the price of transactions in Optimistic and ZK rollups. When an L2 processes a bundle and gets sent to the mainnet, all the transactions inside that bundle share the cost of a unique Ethereum transaction. The higher the number of transactions in a bundle, the lower the fees per transaction.

In the short term, it doesn’t seem like transaction costs will be the primary driver of adoption. Coming from the high costs of Ethereum, any L2 able to provide transactions for less than $1 is already orders of magnitude better. But higher throughput and lower prices can unlock new use cases, like a gaming platform where every action (a trade, a trip, a fight) is reflected on-chain, which can only happen with negligible transaction costs.

Where are the opportunities in the scalability race

Native tokens

Polygon and Optimism have native tokens which operate similarly to Ethereum. In the same way, Ethereum accrues value through the increased demand for blockspace and its use as a fee payment system, the same can happen with MATIC and OP.

Optimism, furthermore, has included governance utility to its token, linking its value to funding public good projects (like open source code, educational initiatives, etc.).

New rollups

Each project has had to make many design decisions and make numerous tradeoffs. We have seen some of them -optimism vs. zero-knowledge, EVM compatible VS EVM equivalent- but hundreds of possible combinations exist in the definition of an L2.

We will probably see some new contenders coming up with new combinations, maybe specialized for specific use cases. We shouldn’t consider the current landscape of L2s final.

New projects

We will see “old” projects deploying in new L2s (Aave’s recent launch of their V2 on Optimism drove Optimism’s TVL up) and new projects leveraging niche advantages in different L2s. As we have seen in the L1 ecosystem, specific use cases (DeFi, NFTs, gaming…) will likely find a better fit in particular layers.

Furthermore, lower transaction costs can set the stage for creative approaches to blockchain technology. They will help some existing trends thrive (like on-chain gaming) and probably make way for unseen applications.

What are the risks

UX is an even bigger challenge in L2s

Web 3 user experience is not for the faint of heart. User experience design is immature in an already green industry, and the process of transferring funds to L2s adds new layers of complexity that create friction and uncertainty for the user. Layer 2s, on a fundamental level, is supposed to be the key to the massive adoption of crypto technology because they build the foundations for acceptable transaction costs and time. But demand will probably not show up until the experience of using an L2 is not dramatically simplified.

New layers, new dangers

Layer 2 requires a new set of tools to interact with, and they constitute new touchpoints subject to hacks. Bridges (such as Ronin, Wormhole, or Nomad) and wallets (such as Solana’s Slope) are frequent victims of hacks or exploits. And every time a wallet or a bridge gets hacked, the damage is felt in every corner of the whole layer, leaving the protocol impotent.