The Rise of Layer 2 Scaling on Ethereum

Ethereum has continued to mature and establish itself as the second-largest digital asset by market cap following The Merge and its transition to Proof-of-stake (PoS). As the first smart contract network, Ethereum introduced groundbreaking blockchain technology that let developers build a growing array of new user applications. Today, Ethereum must battle against an increasing number of competing smart contract layer 1 ecosystems. Scalability is arguably the primary axis in which Ethereum must fight to maintain its stronghold on crypto users. Alternative blockchains, which can offer lower settlement times and cheaper transaction costs (better scalability), could be seen as a threat to Ethereum’s dominance.

For these reasons, Ethereum’s current roadmap is fixated primarily around increasing the overall Ethereum network and ecosystem’s scalability. This scaling roadmap consists of two main upgrades: direct on-chain scaling upgrades, which have shifted in focus over time, and indirect off-chain scaling upgrades. We’ll explore both and discuss why layer 2 networks are likely just beginning to enter the spotlight.

Ethereum’s Scaling Challenges

The total number of active users and applications using the Ethereum network has grown drastically over time. This pushes Ethereum’s current scalability towards its upper bound and drives up the costs associated with routine network transactions. Certain upgrades have helped reduce major spikes in these transaction fees. However, Ethereum is still far from offering consistently-cheap transactions and suffers from occasional instances in which the average user can’t make financially prudent on-chain transactions due to prohibitive costs.

Today, Ethereum’s base layer can only handle about 27 transactions per second (TPS), compared to alternative networks, such as Avalanche, at 4,500 TPS or Solana at 50,000 TPS.¹ As of today, these networks haven’t had the same level of widespread adoption or usage for a variety of reasons. One way in which adoption can be measured is by the relative total value locked (TVL) across smart contract chains shown below.

As digital asset platforms aim to onboard users at an exponentially higher rate, it’s critical for Ethereum to defend itself against competitors that make tradeoffs for better scalability.

While The Merge itself didn’t offer major improvements to Ethereum scaling, it did lay the foundation for the next stage of growth focused on lower costs and higher transaction throughput. 

On-Chain Scaling

One way a network can improve its scalability is through direct, on-chain upgrades designed to improve transaction throughput capacity. The blockchain trilemma, a tradeoff between two of three desirable attributes — decentralization, security, and scalability — creates a limiting factor for those optimizing a given digital ledger. Ethereum developers previously appeared to be focused on improving the network’s on-chain scaling without making large tradeoffs or compromises through what’s dubbed “sharding.”

This original design was created with the idea that most transaction executions would occur on Ethereum’s base layer. The act of separating the Ethereum database into a planned 64 individual execution shards, which are simply smaller individual chains, would increase the number of transactions that can be validated in a given period by distributing the data load necessary to archive data. However, the complexity associated with this endeavor has driven developers to look for other solutions that keep the base layer simple and less prone to security issues. Ethereum’s most notable founder and developer, Vitalik Buterin, made no reference to this type of execution sharding in his extensive updated roadmap released in November of 2022, yet its future implementation is still possible.²

Instead, the idea of sharding has morphed from sharding execution into sharding data. In the new sharding design, known as “danksharding,” the base layer is being kept simple with a focus on decentralization and security, while outsourcing most of the execution to layer 2 platforms. Since execution is occurring outside the main chain, layer 2 platforms post data (a list of past transactions) to Ethereum’s base layer to keep everyone in consensus on the state of all chains. Therefore, future scaling efforts are centered around optimizing the cost and efficiency at which data is posted to and consumed by the base layer.

The next on-chain scaling upgrade, which is currently planned for the second half of 2023, is known as EIP-4844 and introduces a concept known as “proto-danksharding.³” This upgrade will create a new, separate type of transaction that doesn’t compete with the gas usage of typical transactions, known as a blob-carrying transaction. These transactions function similarly to normal transactions, but introduce an added piece of data known as a blob, which will allow batches of transactions from layer 2 networks to settle far cheaper.

Proto-danksharding is thought of as a stepping-stone to get to full danksharding, as it uses a useful element intended for the future sharded Ethereum chain, a new transaction format, but doesn’t actually introduce data sharding itself. The future of on-chain Ethereum scaling has shifted over time, but the community’s view that layer 2s and rollups will be critical to the network’s future appears to be consensus.

Off-Chain Scaling

Creating additional base layer throughput can also be done without design changes to a network’s base layer itself. Instead, off-chain solutions allow for transactions to occur in a separate environment from the main chain. The most popular off-chain scaling solution, and the main subject of this paper, are Layer 2 networks.

Layer 2 solutions inherit their security from the underlying Ethereum consensus. These networks have a primary goal of increasing transaction speed, reducing base layer network congestion (thereby lowering individual transaction costs), while inheriting decentralization and security from the underlying layer 1 blockchain.⁴ Layer 2 networks allow for additional scalability, while not forcing design upgrades or tradeoffs onto the layer 1 blockchain and largely isolating any potential risks to solely the users of that layer 2 scaling protocol. 

The most popular types of layer 2 networks are known as rollups. Rollups allow transactions to take place in a separate environment from the layer 1 chain, but still rely on Ethereum’s security by batching, compressing, and relaying transaction data back to mainnet. This process of combining and reducing data requirements via transaction rollups allows for a drastic reduction in fees and an increase in transaction throughput. 

There are two primary types of rollups: Optimistic and Zero-Knowledge (ZK) Rollups. Each of these has a different design and makes certain tradeoffs.

Optimistic Rollups

The most widely-used layer 2 networks today are optimistic rollups. These are designed, as the name suggests, in an optimistic fashion and assume that any transaction that takes place on the layer 2 itself is valid until proven otherwise. Optimistic rollups reduce the amount of data required to be stored on Ethereum’s base layer by moving computation and state storage off-chain. The reduction in reliance on Ethereum for the major muscle movements of transacting and storing means that optimistic rollups can offer drastic improvements in transaction costs.

Optimistic rollups assume that the off-chain transactions that occur on the layer 2 are valid by default, meaning that cryptographic proofs are not published for the batches of transactions posted on-chain. One can think of optimistic transactions as “innocent until proven guilty.”

Instead, these rollups rely on a combination of incentives and fraud provers to detect instances in which transactions are not valid. The node responsible for recording transactions within the layer 2 blockchain and then aggregating and relaying them back to Ethereum is known as a sequencer.

An allotted period of time, typically one week, serves as what is known as a “challenge period” in which fraud proofs may be submitted to prove any invalid transactions. If invalid transactions are deemed to have occurred, then the protocol rewrites the improper transactions and the entity responsible for the inclusion of the invalid transaction, the sequencer, receives a penalty. Settlement onto Ethereum is finalized after the challenge period has ended.

While optimistic rollups are already operating with scale today, they’re still in their infancy and continued work to decentralize and improve user experience is being actively worked on.

Today, the vast majority of general purpose Ethereum layer 2 networks operate as optimistic rollups. The two largest optimistic layer 2 networks are Arbitrum and Optimism. These networks hold a collective $5 billion in value and have begun to rival the total transaction count of Ethereum’s mainnet when combined. Sending ether using these networks typically costs under $0.20 per transaction compared to over $1 on Ethereum which can cost a few dollars per transfer. Executing more complex transactions, such as swapping tokens, makes the differences in cost more extreme.⁵ Each ecosystem is home to over 50 decentralized applications (dApps) with their largest apps being GMX (Arbitrum), a decentralized spot and perpetuals exchange, and Synthetix (Optimism), a derivatives protocol that allows for synthetic assets to be created and traded.

Zero-Knowledge Rollups

ZK-rollups offer an emerging alternative approach to optimistic rollups. Similarly, transactions are executed off-chain and batches of information are relayed back to Ethereum. However, differences emerge in the types of data provided about the transactions taking place off-chain. ZK-rollups do not follow the “innocent until proven guilty” model. Instead, these rollups require users to submit validity proofs, which provide cryptographic proof that transactions are indeed valid. This means that once a ZK-rollup settles transactions on Ethereum, then all transactions are deemed valid.

The ”Zero-Knowledge” in ZK-rollups comes from a mathematical concept known as a zero-knowledge proof, which refers to a method of verifying if a set of information is true without revealing the actual information itself. A popular example of how zero-knowledge proofs work comes from a scenario involving two different color billiard balls and a color-blind friend. How might one prove to a color-blind friend that they can distinguish between two different colored billiard balls? One can simply identify their colors initially, have their friend switch the balls (or not) behind their back, and then ask the prover to identify the colors again. The odds of correctly identifying in any single trial are 50%, but the odds of guessing correctly, rather than provably knowing, decrease with each consecutive independent trial. In this example one could provably verify the differences between the colors without their friend knowing the colors given enough trials. This example is similar to how ZK-rollups use zero-knowledge proofs to minimize the information set required to prove a series of valid transactions.

ZK-rollups are believed to offer advantages in terms of speed and security over optimistic rollups. Additionally, the differences in fraud proofing means that ZK-rollups allow for value to be rapidly ported back to Ethereum, compared to optimistic rollups, which make users wait one week. The tradeoff for these advantages come with additional complexity, which is why the most notable existing ZK-reliant rollups are application-specific only, such as dYdX.

Multiple teams across the digital asset industry, including Polygon, Matter Labs, and Scroll are competing to be the first to launch a general purpose Ethereum Virtual Machine (EVM) compatible zk-rollup (zkEVM), which could mean that the product that zero-knowledge supporters are waiting for may not be all that far away from launch.⁶ Some expect the first zkEVM to be launched in 2023. Many, including Vitalik Buterin himself, consider zkEVMs to be the ultimate scaling technology once fully operational with the ability to offer functionality on par with Ethereum’s mainnet, while settling transactions significantly faster and cheaper.⁷

Conclusion

In 2022, Ethereum’s transition from Proof-of-work to Proof-of-stake, known as The Merge, took center stage. With the Shanghai/Capella upgrades complete, now allowing withdrawals of staked ether, developer focus has shifted towards scaling and improving the user experience. Decentralized applications have shown a willingness to duplicate their offerings to layer 2 environments to offer better scalability and user experience that, in some cases, could allow them to compete with centralized applications on core financial primitives, such as transferring, trading and borrowing assets.