Ethereum is an engineering marvel: the first true “world computer” with a global, shared state that can be read from and written to in a permissionless manner by anyone. Despite its enormous achievements so far, Ethereum still has a long road ahead before it is usable by the masses.
One of the main hindrances to mass adoption is how easily the Ethereum network can become “congested” when many transactions are competing to be included in the next block, skyrocketing the fee that users need to pay if they wish to have their transactions confirmed within a reasonable amount of time. During the dog-coin mania in May 2021, average Ethereum network fees spiked above $60 USD per transaction, as can be seen in the chart below. Even today, in a “bear market”, the average fee paid in USD hovers around one US dollar simply because the USD price per ETH token has risen so much since inception.
Source: https://ycharts.com/indicators/ethereum_average_transaction_fee, as of 25.04.2023
To avoid pricing out future new entrants, transaction fees need to be driven down to as close to zero as possible without compromising the network’s security. Out of the many options proposed and tested to reduce costs, “rollups” have emerged as the preferred solution for Ethereum because they offer significant improvements in terms of scalability, while still maintaining the security and decentralization of the Ethereum network.
Rollups function by executing transactions off-chain and then rolling up the data from these transactions into a compressed bundle that is submitted to the Ethereum network for finalization, hence the name “rollup”. This compression greatly reduces the amount of computational work required to process the same number of transactions, while still benefiting from the security provided by the Ethereum network. For this reason, Ethereum rollups are often referred to as “Layer 2” or “L2” solutions that run on top of Ethereum’s “Layer 1” or “L1”.
At this stage in Ethereum’s development, rollups are a more secure and decentralized scaling method than sharding, which relies on splitting a single blockchain into smaller “shards” that communicate with each other. Sharding helps to scale a blockchain network because its nodes no longer need to process all of the transactions nor store the history of the entire combined blockchain, only the transactions and history of their specific shard. The shards then communicate with each other to update the global, shared state.
Currently, the main roadblock to implementing sharding is that it splits up the validator set into n smaller sets that are easier to individually attack. If even a single shard is corrupted, the integrity of the entire blockchain can be damaged. Think of the Roman testudo formation: the legionaries could move faster if they separated, but the security of the entire legion is compromised in doing so.
One of the biggest advantages of Ethereum rollups is that they can host decentralized applications (dApps) that require high-security guarantees but are impractical to run on Ethereum L1 due to excessive transaction fees and/or limited throughput. Several financial applications fall into this category.
Uniswap and other decentralized exchanges (DEXs) built on the liquidity pool model drive nearly all on-chain trading activity because their design is optimal when on-chain transactions are expensive and slow. Now that Ethereum rollups can ameliorate these bottlenecks while deriving security from L1, one should expect to see orderbook-based DEXs begin to grow in volume share.
Asset securitization is another financial application enabled by Ethereum rollups; bundling thousands of illiquid loans or NFTs into fungible securities has been a fruitless endeavor on Ethereum L1 due to high transaction costs, but becomes feasible on L2.
Now that we know what Ethereum rollups are and why they are important, let’s explore the two major types that are being developed and the differences between them:
Rollups currently come in two main flavors: Optimistic and Zero-Knowledge.
Optimistic rollups (ORs) assume that the transaction data they send to Ethereum L1 is valid by default; this is the “optimistic” assumption from which these rollups derive their moniker. OR “operators” are the entities responsible for batching L2 transactions together into blocks and submitting them to L1. These blocks are named “assertions” because their validity can be disputed by anyone watching during the “challenge period”, currently a minimum of seven days, beginning from when the block is submitted. OR operators are required to provide a bond, similar to a proof-of-stake system, which can be slashed if they are found to have posted an invalid block. If at least one node of an OR is honest, fraudulent transactions can be detected and punished.
Zero-knowledge rollups (ZKRs) follow loosely the same procedure of bundling transactions off-chain before sending them to L1s for finalization that ORs do, with the key difference being that no optimistic assumptions about the validity of the transactions within the bundle are necessary; they are provably valid at the time of their submission, thanks to the use of Zero-Knowledge Proofs (ZKPs). This means that no challenge period is required and funds can be withdrawn from the rollup without delay.
Furthermore, ZKPs can prove the validity of all transactions within a ZKR block without revealing all of the details of each individual transaction therein, significantly reducing the amount of data that needs to be processed on L1 and allowing for greater user privacy, compared to ORs.
The main downside to ZKRs is that ZKPs are not easily made compatible with Ethereum Virtual Machine (EVM) operations and ZKP production can be very computationally expensive. This is why, compared to ZKRs, ORs are simpler to operate and more compatible with the EVM and existing Ethereum tools. This relative simplicity explains why Optimistic rollups were earlier to reach production and have commanded most of the Ethereum L2 activity up to now.
However, recent research breakthroughs are quickly driving down the operational costs of ZKRs and improving their EVM-compatibility. Over time, the increased capital efficiency and privacy of ZKRs should lead them to slowly gain market share over ORs as research continues to drive down their operational cost.
In future blog posts, we will cover the three most prominent ORs (Arbitrum, Optimism, and Base) and the three most prominent ZKRs (zkSync, Starknet, and Polygon Hermez) in greater detail. Their features and trade-offs will be compared and discussed in detail. We hope you follow along and enjoy!