Data availability: all you need to know about DA layers like EigenDA and Celestia - Part 2
Data availability layers & how they solve the DA problem | Main DA layers: Celestia, EigenDA, Polygon Avail & NEAR DA
Where blockchains cannot solve the data availability (DA) problem, specialized DA layers swoop in to provide rollups with a cheap and secure space to store their transaction data. In this Part 2 of our DA guide, we’ll take a detailed look at the key solutions like Celestia and EigenDA.
DA 101: A recap
In the first part of our guide to DA, we have learned that:
Data availability (DA) is all about making blockchain transaction data available for verification. Ethereum mainnet doesn’t face the DA problem, because each of its full nodes processes all the transactions in a new block, so it knows that they are legit.
The DA problem arises once light nodes enter the picture, because they don’t personally process all the transactions. Nevertheless, they can rely on Ethereum’s sync committee of 512 randomly selected full nodes for data on any transaction, so we can consider the DA problem solved.
DA is a crucial problem for Ethereum rollups, however. They need a way to make transaction data accessible for verification by nodes on the L1 (normally Ethereum), but at the same time not spend too much on writing that data to the L1.
In the past, rollups used calldata to record transaction data on Ethereum mainnet - a very expensive solution. Now, after the Cancun upgrade, rollup data is posted in compressed blobs that full nodes can’t read. So the DA problem becomes: how do Ethereum nodes make sure all the rollup transactions are legit?
The solution (sort of) for optimistic rollups is a challenge window (usually 7 days) - a period when any interested validator can challenge a transaction. ZK rollups use zero-knowledge proofs instead of a challenge window.
Another aspect of the problem is ensuring a rollup sequencer makes ALL of the rollup transaction data available (and doesn’t withhold any) - without having to post it all on the L1. This can be done with data availability sampling (DAS), where nodes download small chunks of data and check if they match.
In Part 2, we will delve into DA layers - special solutions (which can function as separate blockchains or not) that provide rollups with a place to store their data where it is always accessible to all interested parties, from L1 nodes to end users.
DA layers and rollups: an attractive solution to the DA problem
To make sure that all transaction data is always available to all interested parties without spending a fortune on Ethereum storage, rollups can use external DA layers - blockchains that specialize in storing and rollup data. This solution offers numerous advantages:
Low cost: up to thousands of times cheaper than storing calldata on Ethereum L1;
Support for innovative features, such as data availability sampling (DAS), Namespaced Merkle Trees, and fraud proofs;
Easier rollup launch: DA layers simplify the task of building a new rollup, as you don’t have to worry about the data availability side of things - there are now even Rollup-as-a-Service solutions;
Plug-and-Play: a rollup can switch between alternative DA layers.
Additional perks: a DA layer can offer benefits to the rollups that use it, such as XXX.
Note that not just about any rollup can use external DA layers. In fact, a rollup has to be modular - that is, support separate layers for execution, settlement, consensus, and DA. Any rollup built with the OP Stack - including Pontem’s own Lumio - satisfies these requirements, because this stack is modular by design. Other modular stack alternatives include Arbitrum Orbit and zkStack.
By contrast, an existing rollup like Arbitrum One cannot be integrated with an external DA layer, as it doesn’t support the separation between data availability and other functions by design.
Another requirement is that a DA layer needs to provide an interface for a specific rollup stack. For example, Celestia’s integrations for Arbitrum Orbit chains and OP chains are only available in testnet for now. So, for instance, our SuperLumio rollup that settles on Ethereum mainnet can technically support independent DA layers, but for now we use the good old calldata on Ethereum as we wait to be able to integrate EigenDA, a solution by our partner Eigen Labs.
Let’s now explore the four major DA layer providers: Celestia, EigenDA, Polygon Avail DA, and NEAR DA.
EigenDA
EigenDA by Eigen Labs is Pontem’s chosen data availability partner for Lumio implementations on EVM chains. There are ample reasons for this choice: advanced architecture, production readiness, and especially the opportunities to integrate Lumio into the wider EigenLayer ecosystem.
EigenLayer is a restaking layer: it allows users to reuse their staked ETH and liquid staking tokens based on ETH, delegating them to Operators and earning incentives. In turn, operators are recruited by projects to provide so-called Actively Validated Services (AVS) in return for rewards. These AVS projects benefit from the pooled security of staked ETH, so everybody wins: stakers, operators, and AVS protocols.
EigenDA is the first such AVS: it onboards operators to act as nodes, storing and verifying blobs of rollup data. In a nutshell, a rollup’s sequencer node sends batches of transactions to EigenDA and receives a DA certificate, which it posts to Ethereum as calldata. The certificate guarantees that rollup data is available for verification on the DA layer.
DA certificates are issued by a Data Availability Committee, or DAC - a centralized off-chain entity that attests (via a cryptographic signature) that data is available on the DA. In case anything goes wrong with EigenDA, the system will automatically fall back to Ethereum mainnet (via posting calldata).
EigenDA data blobs can now be as large as 16 MB - up from 2 MB previously, which is a huge improvement in terms of rollup gas costs, because rollups can now go longer between certificates (and calldata is expensive, as we know). According to EigenDA, large blobs can save up to $2000 a day!
To check that rollup blocks have been coded correctly, EigenDA uses validity proofs. We’ll talk in detail about validity proofs and KZG commitments in Part 2 of this article; suffice it to say that it’s a definitive way to show that transactions in a block are valid, so a rollup protocol that uses validity proofs doesn’t need a challenge window, and it’s more trustless than an optimistic rollup. However, such proofs are also more expensive to calculate: as always, there’s a tradeoff.
EigenDA is still in testnet, but rollup developers can already use it with the main rollup stacks: OP Stack and Arbitrum Nova (zkStack is coming soon). Importantly, both of these forks now support fault proofs - a permissionless way to challenge rollup transactions (though it doesn’t eliminate the 7-day challenge window).
Pricing for projects depends on the required data bandwidth or storage space: fees start from 0.015 ETH/Gb (around $35) or 70 ETH a year ($165,000) for a reserved 256 Kb/s channel. Yes, data availability doesn’t come cheap!
EigenDA is integrated with major Rollup-as-a-Service providers (platforms where you can launch a customizable rollup in a few clicks), such as Conduit (which Pontem used to launch SuperLumio), Caldera, and AltLayer.
Celestia ($TIA)
Celestia is perhaps the best-known DA layer. It’s built with the Cosmos SDK stack, so it’s technically part of the Cosmos family of chains, even though it can serve rollups from different ecosystems. Celestia itself is a proper independent blockchain, with Proof-of-Stake consensus, a native token (TIA), staking rewards, etc.
A rollup that settles on an EVM chain (like Ethereum) and uses Celestia for DA is called a Blobstream rollup. To prove that rollup data has been published on the DA layer, Celestia validators transfer so-called data commitments to the EVM chain. In case there is any doubt about the availability of a certain rollup block on Celestia, or if a transaction appears to be invalid, a full node on the rollup can issue a fraud proof, similarly to how it happens during a dispute on an optimistic rollup.
This is a very cost-efficient approach, but it forces Celestia to have a disputation window - just like on an optimistic rollup! - during which light nodes have to wait before they can consider a block valid. This is a major difference from EigenDA, which uses validity proofs.
Rollups that want to use Celestia need $TIA tokens to pay for data storage space, or blobspace. $TIA also plays an important security role: as Celestia validators have to stake $TIA, their stake can be slashed if they break the rules (withhold data).
Celestia supports data availability sampling (DAS): light nodes can download a number of random pieces from a block instead of having to download the full block. The bigger the blog, the more light nodes are required for DAS for each rollup. This is a more decentralized approach than the DA committee used by EigenDA.
Celestia’s mainnet is live, and it’s currently possible to deploy on Arbitrum Orbit (on mainnet with fault proof support), as well as on OP Stack in testnet.
Celestia is also integrated with Astria, a solution for deploying modular rollup that share the same network of decentralized sequencers, as well as with Rollkit, a framework for building sovereign rollups on various VMs.
Polygon Avail
Polygon’s family of products is getting bigger and bigger: there’s Polygon PoS mainnet, Polygon zkEVM, Polygon CDK rollup stack, Polygon Miden privacy rollup, and now also Polygon Avail, a whole ecosystem of solutions:
Avail DA, a direct competitor of Celestia and EigenDA;
Avail Nexus, a cross-chain composability protocol for rollups;
Avail Fusion, a security layer.
Avail DA is the first of these to become available on mainnet; it even has a token, AVAIL. Like EigenDA, Avail uses validity proofs rather than fraud proofs; and like Celestia, it features data sampling (rather than a committee).
What sets Avail apart from the competition is that it allows rollup nodes to sample data from the huge network of light nodes on Avail, whereas in other DA layers only full DA nodes can provide such samples. The advantage is that, if DA full nodes are down or start lagging for some reason, the DA layer will keep working instead of having to fall back on Ethereum mainnet.
Avail DA is already working with OP Stack, Polygon CDK, and Arbitrum Orbit. A zkSync integration is coming very soon, as well as one for StarkWare. The ecosystem already includes almost 10 application-specific chains built using Avail DA, including Fuse, Arcana, and Darwinia
Credit: Avail blog
NEAR DA
NEAR is a very interesting L1: extremely fast and cheap, non-EVM, and with sharding already implemented. The new NEAR DA layer is billed as being up to 30 times cheaper than Celestia.
The most interesting feature of NEAR DA is how it leverages NEAR’s sharding algorithm, essentially creating sharded data availability. On NEAR, different shards produce chunks of a block, adding a receipt to each chunk, and consensus focuses on the receipts rather than on checking the data in the chunks. The chunks together with the receipts are joined into complete blocks, and after around 60 hours the receipts are pruned (removed), though special archival nodes retain a record of the data.
NEAR DA uses an off-chain light client to verify that rollup data has been published. As blobs of rollup data are submitted to NEAR, they are also sharded and receipts are generated - and after a few days they are pruned. The light client provides users with blob inclusion proofs, while NEAR validators don’t get slowed down by the stream of rollup blobs.
This approach makes consensus on NEAR really fast and frees the DA layer from storing unnecessary data. 60 hours should be enough to query or challenge the rollup data stored on the DA layer, and the data throughput is much higher.
In the future, NEAR plans to add more shards, to the point where each protocol using NEAR DA will be able to have its own DA shard and won’t even need data availability sampling anymore.
While NEAR L1 is non-EVM compatible, NEAR DA will work perfectly with EVM-centric rollup stacks like OP Stack, Arbitrum Nitro, and Polygon CDK. For now, though, these integrations are at the Proof-of-Concept stage. It will be interesting to see if NEAR DA will be able to compete with Celestia, which, while apparently more expensive to use, is much better known.
Credit: NEAR blog
DA layers like EigenDA and Celestia are becoming very important players in the rollup space, while new solutions, including Polygon Avail and NEAR DA, are promising to make this niche even more interesting and competitive. However, our exploration of DA isn’t complete: in Part 3 of the guide, we’ll delve into technicalities like KZG commitments and DA storage costs, as well as see how Lumio on Solana solves the DA problem. Stay tuned!