The Dencun hard fork, which is anticipated to launch in March 2024, will see multiple Ethereum Improvement Proposals (EIPs) go live on the Ethereum mainnet.
Among them, EIP-4844, also known as proto-danksharding, has been gaining particular interest.
In today’s Ethereum ecosystem, layer-2 scaling solutions like rollups play a crucial role in improving transaction throughput and reducing costs. These rollups “inherit” security from Ethereum by having the execution layer nodes store a temporary copy of the network’s ledger and history. This process allows for the verification of transaction accuracy.
Read more: Rollups saved Ethereum users a boatload of gas fees: Report
However, the current method of achieving this — through writing data as “calldata” — is prohibitively expensive, costing around $1000 per megabyte. This pricing model poses significant challenges for layer-2 solutions, particularly during periods of high demand, due to the substantial costs involved.
EIP-4844 will introduce blob transactions as an alternative to calldata. These blob transactions are a temporary data storage mechanism that are designed to address data availability needs for Ethereum rollup solutions. Blobs are essentially large packets of data that can be processed and stored more efficiently than the current method allows.
Read More: Ethereum’s next upgrade to focus on blobs
In an interview with Blockworks, Edward Felten, the co-founder of Offchain Labs, explains that 4844 will provide a lower level of data availability service but one that still meets the needs of rollups.
“It’s not available to the execution layer of Ethereum, and it is kept by Ethereum nodes for only 18 days,” Felten said. “That’s enough for rollups.”
There are currently two ways in which data is used, Felten noted. One of them is to prove the results of what happened on the rollup.
During this period of time, it is important to ensure that the data itself is robustly available to everyone so that different participants can participate in ensuring the transactions are accurately recorded.
“The initial 18 days is long enough for any realistic proving mechanism to do its work and be done,” he said.
Read more: Glitch in upgrade of Ethereum Goerli testnet won’t delay Dencun rollout
Following the 18-day proving period, users on that particular rollup may still care about the data of their rollup, but this becomes the rollup’s responsibility to provide and keep track of the information, Felten said.
Will blobs make transactions cheaper?
Blob pricing is designed so that when usage is high, the price will go up, and when usage is low, the price will go down.
Felten notes that it is difficult to estimate exactly how much 4844 will reduce the cost of posting data for rollup solutions due to two unknown factors.
“One thing that’s likely going to happen if transactions get cheaper is users will do more of them,” Felten said. “We don’t really know how much more traffic will come on to layer-2s and how many more transactions there will be as a result of transactions getting cheaper.”
This means that if the price of transactions were to get cheaper, that may generate more demand for data posting and blob space, which will, in turn, push the price of blobs back up.
Another unknown factor is exactly how rollups themselves will use these blobs.
“A rollup that’s trying to optimize cost and finality delay for their users, we think we’ll post smaller blobs more often if the price of blobs is low,” Felten said. “If the price of blobs goes way down, that will cause rollups to use more blobs, and that will tend to put upward pressure on the price.”
Felten notes that it will be important for rollups to consider exactly how long it should wait before it posts its data as a batch to Ethereum.
Currently on Arbitrum with calldata, the network accumulates about 100-120 kilobytes of data in a batch before posting it on Ethereum. But unlike calldata, where layer-2s must pay per byte, with blobs, rollups must buy the entire blob, whether or not they use it or not, Felten explains.
“What that means is the tradeoff is a little bit different because if you only have half a blob worth of data and you post it, you’re buying the full blob and only using half of it,” Felten said.
“If you wait longer, you’re going to post more data in your blobs and that saves costs, but on the other hand, your user transactions will take longer to get to Ethereum, and so they will reach finality later,” Felten adds.
Data posting strategies
A potential way to reduce the costs of posting data would be through shared posting. That is where two different rollups may choose to combine their data and post it to Ethereum mainnet.
If rollups were to share data posting in order to save costs, a problem that they must resolve would be to determine what the cost split would be, Akaki Mamageishvili, a researcher at Offchain Labs, told Blockworks.
“Faster rollups, or busier rollups, will save less on the delay costs, while less busy rollups, or slower rollups, will save much more on delay costs, so cost sharing should not be proportional,” Mamageishvili said.
He adds that rollups will need to determine which other layer-2s they would like to “team up” with, and notes that it is likely that larger rollups will likely not need to team up with others, whereas smaller rollups may need to find other smaller rollups to share costs and improve delays.
Data availability options
Ethereum itself is not the only blockchain to which rollups can post their data. Data availability solutions such as Celestia and Eigenlayer are also hoping to address data posting issues through data availability sampling.
Read More: Celestia, the first modular data availability network, launches on mainnet
Each layer-2 solution is able to choose exactly where they want to post their data. In the case of Arbitrum One and Arbitrum Nova, there is currently a DAO vote to ensure that Arbitrum chains can include the capability to leverage EIP-4844 immediately following the Dencun upgrade.
“In my view, it’s a decision for each chain to make about cost versus trust, because you are putting your trust in the data availability system that you’re using. Each chain can make it’s own decisions, I don’t think it’s a one-size-fits-all.”
