Which Rollup to Choose for Your Project: Arbitrum, Optimism, Base, ZK EVM, and Others

This article will focus on rollups — the most popular category of Layer 2 (L2) solutions for Ethereum, such as Optimism, Arbitrum, zkSync, Base, Blast, and others. While we will use the terms «rollup» and «L2» interchangeably for convenience, it's important to note that they are not the same thing.

L2 refers to any solution that optimizes interaction with a specific blockchain (L1) — saving time and money, enabling dapps that L1 doesn't support for technical reasons, and so on. For example, the Lightning Network is an L2 for Bitcoin that optimizes payments; Polygon, by the way, was also initially launched as an L2 for Ethereum.

A rollup is a special type of L2 chain that «rolls up» batches of transactions and sends them in this compressed form to Ethereum (or another L1). Their advantage lies in relatively high speed and low fees compared to L1.

In a nutshell, here's how rollups work:

• Rollups process and confirm user transactions like any other blockchain. When enough transactions are gathered, information about them is compiled into a «rollup,» which is then sent (like a package) to the Ethereum network and recorded on its blockchain.
• Thus, transactions from the rollup in one form or another make their way onto the main blockchain, although otherwise the rollup operates entirely independently of L1, having its own ecosystem of dapps and so forth.

The main challenge in developing rollups lies in proving that all transactions in the «rollup» are valid (after all, L1 validators didn't check them). The solution depends on the type of rollup.

1. Optimistic rollups simply assume that all transactions sent to Ethereum are valid. However, interested participants have a week from the time the «rollup» is received to challenge any transaction in it. For this reason, transactions in optimistic rollups are finalized on L1 only after 7 days, although for regular users, this is only a problem when sending cryptocurrency from the rollup to Ethereum (you will receive your ETH on L1 only after a week). Most popular rollups fall into this category: Arbitrum, Optimism, Base, and others.

2. ZK-rollups generate special cryptographic proofs that allow L1 to verify the validity of transactions without checking them. Hence the term «zero knowledge»: the verifying system doesn't need to know anything about what it's verifying — the proofs are enough. At the same time, the content of the transaction is finalized immediately upon receipt on L1, without any seven-day windows.

ZK-rollups include L2 solutions such as zkSync and Polygon zkEVM. Some experts, including Vitalik Buterin, believe that the future lies in this technology, although generating ZK-proofs is a laborious and costly process.

Often, projects justify their choice of a particular blockchain by citing its low fees. However, it's not enough to simply compare chains based on current fee sizes; it's important to understand what influences them and what fee dynamics to expect.

Different structures: on ZK, transaction costs decrease as the number of users grows

The transaction fee in any rollup consists of two parts:

1) the cost of processing on the L2 itself;
2) the cost of recording on L1 (Ethereum).

When choosing between optimistic and ZK rollups, it's necessary to consider that the fee of the L2 itself (i.e., the first part) is formed differently for them.

On optimistic rollups, as well as in most blockchains, a surge in network activity can lead to a fee increase, as users compete for a place in the transaction processing queue. Imagine a tourist bus where passengers are willing to pay the driver extra to secure a seat. Often, these passengers are trading bots (and we'll soon see how because of them, the transaction price on Base skyrocketed tenfold).

Now imagine a tourist bus where the rental cost is fixed: the more people in the group, the less each one pays. This is similar to ZK rollups — generating a proof requires a fixed amount of resources, regardless of the number of transactions in the rollup — whether it's one or a thousand. Therefore, the cost of using a ZK rollup for end-users decreases as the number of users increases (as paradoxical as it may sound).

Additionally, ZK rollups usually group transactions by protocols — for example, a rollup may contain only swaps or only NFT transactions. Why might this pose a problem for a project? Suppose your protocol has few users at the moment, and their transactions end up in partially empty rollups; then, you'll need to consider subsidizing fees for users, or else they'll be disappointed by the high cost of using the protocol. Many dapps on ZK do this, although the problem was partially solved with the Cancun fork (see below).

The currency for fees and the advantage of the Paymaster solution in zkSync

The currency for gas fees in all rollups for Ethereum is ETH. For example, while Optimism has its token $OP and Arbitrum has $ARB, gas fees are paid in ETH.

zkSync stands out among rollups by allowing users to pay gas fees in any ERC-20 tokens. This is made possible by the Paymaster solution, which can also be used to pay gas fees by the protocol for end-users. Each protocol must customize a Paymaster contract using tutorials and examples. Paymaster is an important advantage to consider when choosing a rollup to launch a project.

Why rollup fees dropped by 60-90% (and then rose again)

Here's a chart of fee changes on the Optimism network before and after the Ethereum Dencun upgrade (Deneb + Cancun):

From 100-120 ETH per day at the beginning of March, the total volume of fees dropped to 20 ETH per day. Why?

The reason lies in the Cancun upgrade, which introduced a new mechanism for registering rollup «bundles» on L1. Previously, this was done using the «calldata» field, where recording each kilobyte of data was very expensive. More than 90% of the entire rollup fee was attributed to recording on L1.

Now, in the Ethereum network, a new type of object has appeared — blob. A blob can be large (up to 125kb) and consists of a body (where transaction data is recorded) and a header. The EVM does not see what's inside the blob (thus, it doesn't spend resources on reading), but it can verify that its content is indeed available on the network. There's no longer a need to write anything to calldata.

The effect of the upgrade is especially noticeable in the graph of fees sent from the rollup (Optimism) to L1:

So, does this mean the problem of high fees is solved? Not quite. Take a look at the graph of the average fee on the Base rollup (purple line):

As you can see, the fee size quickly returned to the level before the upgrade ($2 at the peak). The reason is the activity of bots trading meme coins. The issue is not that the blockchain is overloaded with transactions, but rather that bots intentionally pay more to have their transactions processed faster than others — this allows them to profit from arbitrage or frontrunning.

Conclusion: while the Cancun upgrade minimized the cost of recording rollup transactions on Ethereum, the overall fee can still periodically be very high — particularly due to bot activity. When choosing a rollup, study the fee graph — fortunately, this data is available on Dune Analytics.

For

1. Familiar Tools: Solidity, MetaMask, etc. While fast and efficient next-generation L1s often use alternative programming languages (Move on Aptos, Rust on Solana, etc.), the development process on a rollup is the same as on Ethereum.

2. Easy Migration to Other Rollups: This advantage follows from point 1; if you've built a protocol for Arbitrum, you can launch it on Optimism, Base, etc., with minimal code rewriting. This isn't feasible with an application, say, on Solana.

3. Cost Savings on Development: It's easier to find good Solidity developers, and their service rates are more competitive than specialists in Move and other «exotic» smart contract languages.

Against

1. Average (or Below Average) Time to Finality: This is the time required for a transaction to be finally confirmed on the blockchain. From the end-user perspective, this parameter is more important than TPS; for example, if the block time is 3 minutes and finalization occurs after 6 blocks, the time to finality would be 18 minutes, regardless of TPS.

Some L1 blockchains have a time to finality of less than 1 second, such as Aptos and Avalanche, while rollup users may have to wait up to 20 minutes.

2. Risk of High Fees: As we've discussed, a surge in activity on a rollup like Base (due to DEX bots, for example) can lead to a fee increase up to $2. This is often impossible on many L1s either because transactions can't be accelerated by paying more (e.g., NEAR) or because fees are so minimal that even a several-fold increase won't be a problem for users (Solana, MultiversX, Aptos, Sui, etc.).

3. Limitations Associated with Solidity and EVM: Like any revolutionary and experimental technology, Solidity and Ethereum Virtual Machine are vulnerable and don't always efficiently utilize resources. Partly for this reason, projects on Ethereum are often targeted by hackers. In contrast, next-generation blockchains often use more secure and economical solutions.

We have compiled the most important indicators of the largest rollups into a table. Below, we will discuss why it is important to consider each indicator and how to prioritize them. All data is current as of May 13, 2024. Maximum values among all rollups (the leader in a particular indicator) are highlighted in green, while minimum values are highlighted in red.

*Arbitrum has another chain, Arbitrum Nova, which is technologically more advanced but has not yet gained significant popularity.

TVL (Total Value Locked)

There are various methods of calculating TVL, or the total value of funds locked in protocols within an ecosystem. The most popular approach, used by the analytics platform DefiLlama, is to count the funds «locked» in all DeFi protocols. Funds invested in the NFT market and games are not considered. Conversely, the L2Beat resource, an authority in analytics specifically for rollups, calculates TVL as the value of all tokens circulating in the ecosystem, including those transferred through bridges.

Keep in mind that TVL is subject to sharp fluctuations, especially if a new protocol suddenly gains viral popularity. Therefore, it is worth paying attention to the TVL chart and the dominance indicator of the largest protocol (dapp) (available on DefiLlama).

For example, the new rollup, Blast, reached a TVL of $1 billion in just 35 days. Does this mean that Blast is the future?

Actual TPS (transactions per second)

Comparing chains based on their potential TPS is quite meaningless. What difference does it make whether a blockchain can process 5000 or 1000 transactions per second if the actual demand is 5-20 operations per second? It will take years before the actual number of transactions on L2 networks approaches the maximum, if it ever happens at all. However, actual TPS allows you to assess the demand of the Web3 community — and therefore the interest in your future project.

However, keep in mind that a high actual TPS may indicate a high level of activity by DEX bots. A good example is Base with a TPS of over 30. As mentioned above, Base owes this high figure to bots trading meme coins.

DEX trading volume

This indicator should be given special attention if you are concerned about the future popularity of your project's token, as well as its stability and chances of listing on a major exchange.

The total trading volume on all DEXs in the ecosystem indirectly reflects the overall popularity of tokens in that ecosystem and the level of user activity (although it should be noted that some of the activity is generated by trading bots).

By launching a project on a chain with low trading volumes, you risk that:

A single major swap or manipulation on DEX could cause a price spike, which could negatively affect the project's reputation;
It will be difficult for you to attract the attention of a major exchange to the token (listing is primarily ensured for tokens that are already actively traded on DEXs).

Grants

Many rollups (Optimism, Arbitrum, Blast) have grant programs, but they should not be seen as free money giveaways, but rather as hurdles to overcome.

• Each rollup has its own requirements. The application and accompanying documents will need to be prepared very carefully, demonstrating the project's competitive advantages, the team's previous experience, etc.;
• KYC/KYB will be required, and the identities of the founders will need to be disclosed;
• In case of success, funds will be allocated in stages upon achieving the goals set out in the application — therefore, these goals will need to be achieved in a timely manner.

In general, it is somewhat easier to obtain funding through the Gold Points program of the Blast blockchain than through the official programs of Optimism Collective or Arbitrum. But it's best not to see a grant as a key factor when choosing a rollup and certainly not to include it in your financial model in advance.

Technology

Does the rollup provide a convenient environment for protocol creation? Is its technology interesting enough to ensure ecosystem growth for years to come? Or perhaps the popularity of this rollup is driven by hype or the first-mover advantage, rather than the quality of its technology?

All these questions need to be answered when choosing a rollup.

When launching a project on a rollup, one must be prepared for the role of rollups themselves to evolve in parallel with Ethereum — but how exactly is unknown.

Just a couple of years ago, it was planned that Ethereum would transition to sharding technology and rollups would largely become unnecessary. However, Ethereum ultimately implemented the so-called proto-sharding technology — a completely different approach that actually facilitates the operation of rollups and gives them a central role in the ecosystem.

At the same time, Vitalik Buterin has already discussed moving certain ZK rollup functions from L2 to L1 and creating a kind of centralized native zkEVM. However, he writes that rollups will retain several important functions, such as fast transaction pre-confirmation, protection against MEV bots, and enhancements to the EVM.

Furthermore, it is still unknown which type of rollup will dominate in the future — optimistic, ZK, or even hybrid rollups.

In other words, even if today you manage to find the optimal answer to the question «Which L2 to choose for the project?», there's no guarantee that this answer will remain correct in a year or two. If you plan to develop the project for many years, be prepared to refine its architecture and, if necessary, even move from one rollup to another.