Rollup Economics: We Overestimated the Impact of EIP-4844 on Scalability

Written by: 0xfan, Smarti Lab

Compiled by: Peng SUN, Foresight News

TL; DR:

1. We evaluate the potential reduction in Gas fees, TPS (transactions per second), and the capacity to accommodate Rollups after the implementation of EIP-4844 using two computational methods.
2. It is estimated that EIP-4844 could accommodate more Calldata, ranging from 38 to 192 times, with Calldata sizes of 10KB and 2KB respectively. As more Calldata can be accommodated in the same block, the cost per unit of Calldata will also decrease accordingly.
3. Assuming a uniform Calldata size of 2KB for each Rollup, EIP-4844 can accommodate a maximum of 384 Rollups.
4. Under normal conditions (i.e., when blocks reach their target size), Ethereum will achieve 175 TPS through EIP-4844, with a maximum of 350.
5. Contrary to popular belief, relying solely on EIP-4844 is not sufficient for Ethereum to significantly improve scalability.
6. Utilizing alternative DA layers (such as Celestia) or DACs (like zkPorter), increasing the compression rate of L2 transaction data, and raising the proportion of zk Rollups will all have significant impacts on further enhancing Ethereum's scalability.

Proto-danksharding (i.e., EIP-4844) proposes to implement most of the logic and rules that may be used in Danksharding in the future. Currently, high storage costs on L1 lead to high transition fees for L2. To address this issue, EIP-4844 introduces a new data type called Blob, which is cheaper and larger than calldata, providing another way to store rollup data.

With the upcoming launch of EIP-4844, L2 sequencers may gain higher profits. This is because sequencers are responsible for batching transactions into L1 and paying data fees, and the L1 data fees paid by sequencers will be significantly reduced. Lower transaction fees could potentially generate more MEV by increasing the number of orders on L2.

The Cancun upgrade will include EIP-4844, but there is currently no exact timeline for the upgrade. The Ethereum Foundation research team has indicated that the Cancun upgrade may go live by the end of October. However, it is more likely to be launched around the first quarter of 2024.

So, to what extent can EIP-4844 reduce transaction fees? Currently, L2 transaction fees consist of two main components:

• Rollup Costs: The cost of packaging, submitting, and storing transactions on Ethereum.

• Execution Costs: The cost of running transactions on L2.

L2 Transaction Fee = Rollup Costs + Execution Costs = [ L1 Gas Price * (Calldata + Fixed Overhead) ] + [ L2 Gas Price * L2 Gas Used ]

Taking Optimism as an example, currently, nearly 80% of the total transaction fees come from L1 storage costs (i.e., calldata costs). We will temporarily ignore the impact of other fees and propose two methods to estimate how much L2 transaction fees may decrease after EIP-4844.

In EIP-4844, after the proposal is implemented, each Blob has a size of 128KB and consumes 131,072 Gas. Therefore, on average, each Blob data byte will consume 128 * 1024 / 131,072 = 1 Gas. In comparison, currently, storing a single calldata byte requires 16 Gas. This indicates that the storage cost of L2 transactions will decrease by 16 times.

However, this method only compares the storage costs per byte and does not consider the total Gas capacity of the block. Since the total amount of Gas that a single block can carry may change after EIP-4844, the storage costs for L2 transactions may decrease by more than 16 times.

The second method considers block size and examines how many times current calldata can fit under different block sizes. Based on current parameters, in the scenario of the target block size, a block can accommodate 3 Blobs (0.375MB) and a maximum of 6 Blobs (0.75MB). Considering that current calldata occupies about 2-10KB per block, after EIP-4844, it can accommodate 0.75 * 1024 / 2 = 384 times the calldata.

However, as block size increases from the target value to the maximum value, Gas prices grow exponentially. Therefore, under more common conditions (i.e., when blocks reach their target size), EIP-4844 can accommodate 38 to 192 times the calldata of 10KB and 2KB respectively. As the capacity for calldata within the block increases, the storage costs for calldata will also decrease accordingly. Thus, the costs for L2 transactions will also decrease correspondingly.

Additionally, assuming a uniform Calldata size of 2KB for each Rollup, EIP-4844 can accommodate a maximum of 384 Rollups. This does not reach the thousands of Rollups that many envision.

From this, we can also derive the TPS magnitude that Ethereum can achieve after EIP-4844. Currently, an average L2 transaction requires about 3000 Gas of calldata on L1. Considering that the Gas cost per byte of calldata is 16, this indicates that each L2 transaction on L1 is approximately 187 bytes.

After EIP-4844, with a target block size of 0.375 MB, Ethereum generates a block every 12 seconds. Therefore, the available space per second is 0.375 / 12 * 1024 = 32 KB, which can accommodate 32 * 1024 / 187 = 175 transactions. Thus, under normal conditions (i.e., when blocks reach their target size), Ethereum's TPS after the EIP-4844 upgrade should be 175, with a maximum of 350.

While higher TPS can improve efficiency, it is worth noting that even with the implementation of EIP-4844, Ethereum still cannot match Visa, which currently has a TPS of up to 1700. This gap may still lead to congestion in both L1 and L2 networks, especially in high-demand scenarios.

Therefore, relying solely on EIP-4844 is not sufficient for Ethereum to achieve greater scalability. We still need a more cost-effective and efficient data availability solution to store more calldata (such as DA layers like Celestia or DACs like zkPorter), which are still crucial for achieving scalability.

Finally, the compression rate of L2 transactions will directly affect the size of calldata stored in L1. The higher the compression rate, the lower the required L1 fees. As zk Rollups continue to develop, the amount of data that needs to be stored on L1 will decrease, which is more conducive to improving Ethereum's scalability. Unlike Optimistic Rollups, zk Rollups only need to store state changes rather than the entire transaction.

Conclusion

In this article, we used two different computational methods to evaluate the potential reductions in Gas fees, TPS (transactions per second), and the capacity to accommodate Rollups after the implementation of EIP-4844. The results indicate that assuming a uniform Calldata size of 2KB for each Rollup, EIP-4844 can support a maximum of less than 400 Rollups. This is far from the demand for thousands of Rollups that many expect. Utilizing alternative DA layers or DACs, increasing the compression rate of L2 transaction data, and raising the proportion of zk Rollups will all have significant impacts on further enhancing Ethereum's scalability.