Arbitrum paid about $1980/MB in DA fees last month. Meanwhile, Manta only paid $3.41/MB using Celestia DA. Modular DA layers can save a lot of gas fees for Rollups. Celestia, Avail or EigenDA... Who will win the DA war?
Why do we need modular DA?
In a monolithic blockchain like Ethereum, data availability (DA) is usually part of a single system design, and when block space utilization is high, data availability will become a bottleneck limiting its development. The higher the gas fee, the worse the user experience.
Ethereum has also been aware of the scalability problem for many years and has begun to explore various off-chain scalability solutions. Rollups, especially OP Rollups, have become a popular off-chain scalability solution in recent years, but they still face the problem of high gas fees when block space demand is high.
Another problem Rollups face is that they spend almost 70-90% of their revenue on calling data and storing Ethereum DA.
Modular data availability layer is one of the potential solutions to reduce DA costs, which can reduce costs by up to 99%.
This has been proven in Celestia and Manta. By choosing Celestia instead of Ethereum DA, Manta successfully saved nearly 99% of DA costs. Although Ethereum is also undergoing a major upgrade, it will introduce a separate block space for L2 call data.
It is estimated that the call data cost can be reduced by 5-10 times. But even if Proto Dank Sharding (EIP-4844) is compared with Celestia, Celestia still outperforms Blobspace by at least 50 times.
Now the question is, what are the design considerations of modular DAs like Celestia, Avail, and EigenLayer that enable them to reduce the call data publishing cost to such a low level?
There are many factors that affect the overall performance and cost of the DA layer, and crypto KOL hitesh.eth believes the following five are the most important:
Data Availability Sampling (DAS)
Consensus + DA guarantee
Decentralization
Scalability
cost
Data availability sampling
In Ethereum DA, nodes traverse all the data of a block to check whether the data is available, which not only takes a lot of time but also increases costs. In data availability sampling, light nodes can verify the availability of data without downloading the data.
Modular DAs like Celestia and Avail use techniques such as erasure coding to split data into small shards and allow light nodes to randomly select a subset of shards to download and verify.
Celestia also uses Namespaced Merkle trees (NMT), which allows Rollups to publish only relevant parts of the data to Celestia blocks, helping to reduce light node storage and verification costs. EigenDA does not use data availability sampling.
Consensus + Data Security
When talking about reducing the cost of data availability, consensus plays its own role. It consists of two parts: transactions should be added to blocks faster, and the accuracy of transactions should be securely guaranteed.
Therefore, the final block confirmation time plus the data guarantee time will play their respective roles in optimizing the performance of the DA layer. Celestia uses Tendermint, which is faster in terms of final block confirmation time, and also uses anti-fraud design to determine the accuracy of transactions.
On the other hand, Avail uses a hybrid consensus mechanism (Grandpa and Babe) inherited from the Polkadot SDK, which has a slower block finalization time compared to Tendermint. However, since Avail also uses KZG commitments for validity proof, it verifies transaction accuracy faster than Celestia.
EigenDA uses Ethereum’s Casper Proof-of-Stake consensus, which is slower than the other consensuses discussed above. EigenDA uses DAC for the verification process; a DAC is a committee composed of a group of validators who are responsible for the verification process.
Data Availability Committees (DACs) save verification time, but also bring centralization risks. Using DACs can also effectively reduce DA costs, as demonstrated by Metis and other validity solutions.
Decentralization
Decentralization is one of the most important aspects of design considerations. Decentralization refers to how many validators are securing the network and how the stake is distributed across the network. Celestia supports up to 100 validators.
Avail uses designated proof of stake and supports up to 1000 validators. It can remain operational even if most full nodes go down because they are sampled from the p2p network of light nodes. Avail is the only DA layer with a strong fail-safe mechanism.
EigenDA is an Active Validation Service (AVS) that shares security from Ethereum validators by locking and re-staking ETH on smart contracts. The security guarantee of EigenDA will depend on the amount of ETH re-staked by the number of validators.
Scalability
The modular layer should have the ability to dynamically scale block size as demand increases. Celestia and Avail use large blocks, data availability sampling, and light nodes to meet growing demand. The DAC used by EigenDA is also considered scalable.
cost
The cost of modular DAs depends on various factors discussed above. Celestia DA is now available and currently charges $3.41/MB to Manta; the cost will also vary based on the price of TIA. Avail and EigenDA are not yet available, and NEAR DA is the cheapest.