EIP-4844 economics and rollup strategies

In a recent paper, 2310.01155.pdf (arxiv.org), we study economics of the proposal and how rollups might react to it.

By modeling the cost of a rollup as a sum of two parts, blob posting costs and delay costs, we obtain that small rollups may opt for posting in a regular blockspace market, while large rollups will post in the new data market. Then, we proceed with joint blob posting strategy for two rollups and its effect on the equilibrium price. Results are intuitive: if both rollups were using blob data market, the price goes down, otherwise it goes up, as the demand for the blob market is increasing. Towards the end, we discuss how joint blob posting cost can be shared between (two) rollups, as a function of how much data each rollup contributes to. For this reason, we employ a Nash bargaining solution and obtain that the larger rollup always pays more than half of the posting cost, but less than it would pay proportionally. On the other hand, small rollup improves its per-transaction cost more than large rollup does, in this cost sharing setting.

Any feedback is welcome.


Glad to see resources on this topic!

I have a couple of thoughts I’d like to share and I’m curious about your perspectives.

Don’t you think that initially the blobs will be mostly empty? Given that rollups might not have the significant user base to fully utilize them at the outset, they could operate at extremely low cost, with only a minor amount of execution gas at layer one. Since we’re not at the stage of having 3 blobs per block (yet), it’s conceivable that it will be practically free for users until there’s enough traction to fill that blob space and cause congestion, which would subsequently increase the price.

Another question is whether dapps will have the option to dictate if the transactions of “their” users will be incorporated into a blob in every single block (even if it’s not fully packed) or not. This could allow them to choose between prioritizing enhanced security or opting for reduced costs, akin to what e.g Starknet envisions with volition – although the approach might differ.

I don’t think the blobs will be initially empty. If the blob price is very low, every rollup will want to buy a blob in every block, because that costs almost zero. That demand would drive up the price until enough rollups slowed down their posting. The result would be an equilibrium with the blob usage equal to the target of 3 per block.

(Also note that if the blob price is very low then non-Ethereum applications will find blob posting to be a cheap way to buy 18 days of storage. Those spammy users would fill the space until they bid up the price to equilibrium.)

The batch posting mechanism needs to post transactions in the order that they execute on-chain, so there’s no obvious way to post more delay-sensitive transactions faster than other transactions. To accelerate posting of a transaction, we also need to accelerate posting of the transactions before that one. That’s possible but the incentive issues are more complex.

The reason I mentioned that blobs could be initially empty is because rollups today use something on the order of 1 kb/s. However, after 4844, the Ethereum data rate should be close to 30 kb/s. But yes, I see your point now.

Thanks for the replies!