Foresight Ventures: The Present, Obstacles, Competition, and Future of Rollup

Author: msfew, Foresight Ventures

This article focuses on the Ethereum Layer2 Rollup universe (only including Secured rollup), starting from simple and understandable core concepts and mechanism designs, exploring the pros and cons of current Rollups, and envisioning their potential routes and solutions for additional features such as decentralization, further scalability, composability, and privacy.

A Secured Rollup is a model like Arbitrum or Optimism, where settlement, consensus, and data availability rely on Ethereum or other L1s, and the Rollup is only responsible for execution.

If the settlement, consensus, and data availability in a Rollup use solutions other than L1, then it is not a pure Secured Rollup, and thus not a narrow sense Rollup.

0. The Essence and Principles of Rollup

a) The Essence of Rollup

There are two ways to scale a blockchain: one is to optimize the blockchain itself, and the other is to use the blockchain in a better way.

Rollup is the second approach; its true essence is simple: to use the blockchain faster, cheaper, and "trustworthily" to scale the blockchain (specifically referring to Ethereum).

Rollup is: On-chain smart contracts + Off-chain aggregators.

It's that simple. The combination of these two features defines Rollup and enriches its concept.

• On-chain smart contracts indicate that its trust model is based on a smart contract on Ethereum, leveraging Ethereum's security, rather than needing to establish a new trust consensus like Alt L1s. We can trust Arbitrum's protocol just like we trust Uniswap's protocol (which is fundamentally a smart contract).
• Off-chain aggregators mean that it executes and aggregates transactions off-chain, compressing a large number of transactions and ultimately posting them on the Ethereum mainnet, achieving faster and cheaper results.

The principle of Ethereum is that each node stores and executes every transaction submitted to it by users, making such a decentralized network very expensive.

The following text discusses the mechanisms of Optimistic Rollup, represented by Arbitrum and Optimism, and zk Rollup, represented by zkSync and StarkNet, for readers to understand simply.

b) The Principle of Optimistic Rollup

As the name suggests, Optimistic Rollup is optimistic, adopting a presumption of innocence, where everyone defaults to believing that each execution is correct, and security is ensured through fraud proofs that can challenge each batch state.

Users submit transactions in Arbitrum, and the sequencer of Arbitrum executes the transactions, submitting the state root and transaction data in batches to the smart contract on the Ethereum mainnet.

What if an error occurs during the execution of Optimistic Rollup?

Optimistic Rollup has a dispute verification period, meaning that the data on-chain will take, for example, a week to be finalized, during which anyone can challenge it and prove that the batch is incorrect.

c) The Principle of zk Rollup

As the name suggests, zk Rollup uses zk zero-knowledge proof technology (actually Validity Proof), adopting a presumption of guilt, where everyone defaults to questioning that each execution is incorrect, and security is ensured through the proofs contained in each batch.

Users submit transactions in zkSync, and the sequencer of zkSync executes and stores the transactions, submitting the state proofs and new states in batches to the smart contract on the Ethereum mainnet.

How does zk prove that its execution is not incorrect?

The Prover (most of the time the sequencer) generates an unforgeable proof for the execution of the transaction, proving that these new states and executions are correct. The sequencer compresses the proof and other data before submitting it to the smart contract on the Ethereum mainnet for verification.

How do they compress data to reduce gas fees?

Rollup can compress the byte size of each transaction, including Nonce and signatures. For zk Rollup, it can compress further by not putting data unrelated to state updates on-chain, because zk proofs can already prove the correctness of state updates, while Optimistic Rollup cannot ignore this part of the data due to the need for challengeability.

Why do both require complete state?

To prevent the sequencer from running away, we can rebuild the entire Layer2 using data on Ethereum. This is the concept of data availability.

d) Differences Between the Two Types of Rollup

If you still can't understand, you can refer to the diagram below.

The design of Optimistic Rollup is more like Arweave, designing mechanisms through economics and games, performing well but not guaranteeing 100% correctness.

The design of zk Rollup is more like Filecoin, designing mechanisms through cryptography and mathematics, guaranteeing 100% correctness but always incurring additional computational and time overhead.

The difference is that one uses a challenge period to ensure that errors can be corrected, while the other uses cryptography to ensure that errors cannot occur.

We will compare their endgame performance in detail later.

1. Rollup is Excellent

First, let's review the overall excellence of Rollup networks:

• Ecosystem: Arbitrum has spawned NFT and metaverse ecosystems; Optimism has issued tokens, established governance and community; StarkNet is brewing various GameFi innovations; zkSync has several new DeFi projects. Each Rollup stands out on its own.
• Scalability: TPS is indeed excellent; although it may not reach theoretical peaks, the rapid confirmation of L2 provides a fast and cheap Web3 experience.
• Experience: For developers and users, switching to Rollup is a simple matter, and the experience is almost indistinguishable from Ethereum. Various cross-chain bridges, wallets, Uniswap, and other infrastructure also serve users' smooth experience. This is also supported by a good ecosystem.
• Innovation: Rollup has become an excellent soil for Ethereum application innovation, and is more imaginative than Polygon's "large testnet" style network.

In addition, if you are interested in the comparison between subnet and Rollup, you can refer to our previous research and Toghrui's views.

More specifically, major Optimistic Rollup and zk Rollup have provided the following performances:

• Arbitrum: TVL is around $2.4 billion. The Nitro upgrade has been launched on the testnet, and will replace Arbitrum's existing VM with WASM-Geth, optimizing performance and adaptability. (We can discuss the various potentials of WASM in future articles.)
• Optimism: Tokens have been issued, and many "pioneering and incredibly optimistic experiments" have been conducted (various mishaps, hiding facts, and being indifferent, recently often questioned by disappointed users as "the waste prince of the Ethereum Foundation"). The Bedrock upgrade will make the VM MIPS-Geth, achieving EVM equivalence. Optimism's developers have also been talking about the concept of zkMIPS; we don't know if they will work in this direction in the future.
• StarkNet: The cross-chain bridge is not fully open yet. The specialized language Cairo and StarkNet's ecosystem development are doing well, and the gaming ecosystem is worth paying attention to. StarkEx's Validium model has also produced applications like dYdX and ImmutableX. Moreover, StarkNet can also serve as a settlement layer, building L3.
• zkSync: It was the first to create a relatively complete zkEVM. The 2.0 version's Volition model allows users to flexibly choose between zkPorter or Ethereum's data availability solution.
• Aztec: A privacy zk Rollup based on the UTXO model, soon to support a privacy DeFi ecosystem (but due to the large size of transaction data, it still needs additional optimization). Although it does not support general computation, it is one of the few truly zk Rollups.
• Fuel: V1 is an Optimistic Rollup and is very decentralized. V2 takes an unconventional route, not adapting to EVM, but supporting parallel tx processing through a high-performance VM based on a UTXO model, aiming to create the fastest execution layer.

In addition to Rollups based on Ethereum as the mainnet, there are also:

• everPay: A Layer2 based on the SCP paradigm on Arweave.
• Milkomeda: An EVM Rollup on Algorand.
• Orbis: A Validium model Rollup on Cardano.

The different designs of Rollup solutions have many unique features, truly serving countless users and bringing a better Crypto and blockchain experience.

2. Rollup Needs Improvement

After the praise in the previous chapter, we will point out some issues that Rollup needs to address in this section:

a) Centralization

The smooth user experience of Rollup networks is achieved at the expense of security, as many components of the network are centralized. Although there are three points that can "argue" this issue:

1. In zk Rollup, zk proofs are mathematically unforgeable.
2. Users can rebuild all states through L1's DA.
3. The trust model itself is a centralized smart contract.

However, the centralization of Rollup still leads to issues such as censorship and extremely low security. Currently, its centralization mainly manifests in:

• Centralized sequencer: Most Optimistic Rollup and zk Rollup sequencers exist in a centralized form. This means that transaction execution, ordering, and block production rely on a centralized server. Previously, Arbitrum experienced a network outage due to hardware issues with the sequencer…
• Centralized upgrades: Most Rollup networks' upgrades are centralized, meaning that the cost of effectively blacking out the entire Rollup is negligible compared to the billions of TVL on the network. For example, zkSync is managed by a security council for upgrades. Reflecting on the previous Ronin hack incident, relying on a few people's multi-signatures for billions in funds is absolutely unsafe.
• Closed-source components: Some key components of certain Rollups are closed-source. For instance, almost all key components of StarkNet are not open-source, remaining centralized in development and lacking an open-source license. This is not very Web3.

In my view, as a network or protocol, it must ultimately be decentralized and open-source; otherwise, it contradicts the spirit of Crypto and Web3.

b) Authentic Performance

The super high performance currently exhibited by Rollup is "fake"; the real finality is not that excellent, and there is still room for performance improvement. This is a double-edged sword.

When users submit transactions on L2, they immediately receive feedback on whether the transaction was successful or failed, but this instantaneous user experience is only the finality on L2.

During the period when data has not been submitted to L1 or the Prover is still in the proof or arbitration window, this data has not achieved L1 finality. Therefore, the excellent user experience of Rollup is actually because users borrowed some time through Rollup, allowing them to first obtain a soft confirmation.

Due to the above design, the following issues arise:

• Fake Finality: Data must be on a real blockchain like Ethereum to achieve true finality; so how fast is the real finality of Rollup? In other words (and of course, if we really want to calculate TPS, the number of txs contained in a batch of data should also be considered), how often do they submit data to Ethereum? Here is a real analysis based on on-chain data.

• Overstated TPS: I personally believe that TPS is an extremely poor metric that should be eliminated. Various Rollups or public chains calculate an extremely high or at least higher TPS than competitors through countless preset conditions and confusing concepts to showcase their strength. However, in actual use, their performance is still not that "Web3." The truly ideal Web3 experience I envision needs to be completely seamless, and the entire network must also be decentralized.

c) Economic Mechanism

Using Optimism's token as an example, we can gain insight into the business and economic mechanisms, token models, and MEV of Layer2 Rollup.

Business Model: Rollup's profit = Layer2 fees - Layer1 block space fees (the security cost of storing and verifying proofs and other data on L1) - node server and other computing costs. Among them, Layer2 fees = gas fees + MEV + some premiums charged by Layer2. However, users cannot participate as nodes in this business model; only Rollup developers can profit. Below are the earnings and costs of Arbitrum and Optimism in Layer2 fees and Layer1 block space fees.

Token Mechanism: Although Optimism's network is good, its token economy is still relatively weak. OP has three fatal issues: 1. There is no utility, such as being unable to pay network gas or PoS staking (of course, these two are not good Rollup utilities), only used for governance. 2. It cannot capture the value of network earnings and block space; how much the sequencer earns is unrelated to token holders. 3. The circulating supply is very low, with significant selling pressure, and incidents during market making have exacerbated this situation. These issues are not just problems for Optimism; other Layer2 Rollups are likely to face similar challenges.

MEV: Since Rollups currently have centralized sequencers, these earnings can be captured by the sequencer. MEV is a double-edged sword. On the positive side, Rollup networks can embrace the value that MEV brings to the network, making it an important part of the economic mechanism and miner incentives after decentralization; on the negative side, Rollup networks need to eliminate any existence of MEV, as it is often seen as an additional tax on users. Arbitrum chooses to "ignore" MEV, adopting a first-come-first-served mechanism, but this places higher demands on node performance and lowers the cost of DDoS attacks.

The open network's business model, improving the token mechanism, and addressing MEV will be three major challenges for Rollup networks on their path to token issuance and decentralization.

3. Endgame Performance Comparison Between Optimistic Rollup and zk Rollup

I don't want to compare the TPS and other paper performance of Optimistic and zk Rollup camps like other articles do (if you're curious, you can read more reference materials at the end), but rather compare them based on their essence and characteristics, as well as their long-term endgame performance:

The detailed interpretation and points of interest regarding this diagram are as follows:

1. Difficulty of achieving EVM equivalence: In fact, whether from EVM compatibility, node configuration requirements, or optimization, the overall implementation difficulty of zk Rollup is higher than that of Optimistic Rollup. This also serves as a form of additional proof overhead… The additional proof overhead not only burdens the network's performance but also poses a significant obstacle to the technical implementation of EVM compatibility.
2. Easier to meet privacy and other needs: Since zk is already used, developing a truly private zk-zk Rollup will be much simpler than zk-Op Rollup. Of course, there are zk-Op Rollup projects like ZKOPRU that have already gone live.
3. Optimization direction: For Optimistic Rollup, parallel transaction processing is a straightforward and effective optimization, while zk Rollup using the same approach (the next version of StarkNet's sequencer can process in parallel) will still ultimately lose to Optimistic Rollup in the long run (again due to the additional proof overhead). Therefore, zk Rollup will adopt a Fractal L3 scaling alternative, which means continuing to build Rollups on top of Rollups. The recursive superpower can not only allow L2 to verify the correctness of L3's proofs but also enable L1 to verify the correctness of L2's verification of L3.
4. Specific optimization direction: These are derived from roadmaps and envisioned design plans, and there is currently no practical implementation. I personally believe it will take a long time to complete these upgrades. Some of these upgrades may have side effects, such as introducing a new mechanism that requires additional trust assumptions.
   Although this diagram concludes that the performance of Optimistic Rollup will be higher than that of zk Rollup in the long run, the trust guaranteed by mathematics is stronger. I believe the endgame performance of Rollup will be Optimistic, but the true overall endgame will be zk.

The design is perfect, but the engineering implementation difficulty remains enormous. The above comparison is still akin to comparing TPS; it is more theoretical. We should start from reality and observe a Rollup network's security, decentralization, and actual performance and ecosystem.

4. The Future of Rollup

Having compared Optimistic Rollup and zk Rollup, we have also comprehensively understood their future optimization plans for performance. So, what will the future development of Rollup look like from other perspectives?

a) Decentralization

From the perspective of decentralization, we focus on the issues of sequencer and contract upgrades. In the future, a Rollup must be decentralized, which requires the following decentralized designs:

• Decentralized sequencer: The mechanism of Optimistic Rollup's sequencer will basically be various modified PoS (leader election, MEV auction, rotation mechanism); and zk Rollup will have mechanisms similar to Hermez's Proof of Efficiency or Proof of Validity Proof or Tendermint PoS. Of course, if the network is a PoS mechanism, it will require additional capital costs.
• Decentralized contract upgrades: Contract upgrades should not be managed by multi-signatures but rather by decentralized governance (involving tokens) or completely relinquishing contract upgrade rights (like Uniswap V2 upgrading to V3, which is a non-mandatory upgrade).
• Multi-language clients: Like Ethereum, multi-language clients are also a necessary condition for decentralization, ensuring 100% uptime of the network.

Here is a small piece of information: by comparing the documentation of Arbitrum and Optimism, one can sense their slightly different attitudes towards decentralization.

b) Composability

The composability of Rollup refers to the ability of smart contracts on different Rollups to directly read and write to each other.

For Rollups on Ethereum, Ethereum as the settlement layer is the shared bridge for Rollups. In the future, the composability between Rollups is likely to form a call through L2 ⇒ L1 ⇒ L2 (for example, the dAMM designed by StarkNet).

However, true composability is difficult to achieve due to the inability to verify each other because of different mechanisms between Rollups, requiring more exploration.

c) Functionality

In terms of functional expansion, zk Rollup has more potential. The concept of L3 makes App-specific Rollups particularly attractive; Immutable X has also announced that it will deploy protocols on StarkNet, allowing it to exist as an L3.

The appeal of L3 lies in:

• Transforming applications into protocols, opening their interfaces and services, achieving greater composability.
• L1 as the hour hand, L2 as the minute hand, and L3 as the second hand. Further enhancing the overall network performance without affecting the pulse of L1.

I envision the most interesting functionalities of L3 to be:

• Privacy L3
• Pure payment L3
• NFT marketplace L3
• Game engine L3
• DEX and DeFi L3

However, the challenge for L3 is how to handle DA. We may adopt the following different models:

• Secured Rollup: Still counts as Rollup, placing DA on L1 like L2, but may involve more complex exit mechanisms.
• Validium/Optimistic Chain: Placing it on L2 or additional DA layers like Celestia, requiring extra trust assumptions.

5. Conclusion

This is our understanding of what Rollup requires.

In addition, Rollup has created countless new topics: L3, modular blockchains, data availability solutions, UX improvements brought by account abstraction, new tech stacks, new programming languages, new developers, new auditing institutions…

Everything about Rollup is so new, so shiny, yet so immature…

How to "better use" blockchain like Rollup is a question worth exploring.