DA Layer "Three Kingdoms Battle": Analysis of the advantages and differentiated development directions of Celestia, Avail, and EigenDA. Who will dominate the DA market?

Written by: c4lvin, Four Pillars

Compiled by: Glendon, Techub News

Key points:

The current Data Availability (DA) layers (Celestia, Avail, EigenDA) will long split the market in different directions, similar to the "Three Kingdoms" in ancient times, each focusing on different areas of advantage.

In the short term, the three DA layers are competing to increase throughput, and the outcome of this performance battle may determine which project will dominate the DA market in the long run.

This article will explore three major DA layer projects: Celestia, Avail, and EigenDA. Despite the technical similarities, they are developing in different ways.

The current comparison of DA does not reflect the future roadmap.

Celestia, Avail, and EigenDA have many commonalities in their business areas. Starting with Celestia, followed by Avail and EigenDA, they emerged almost simultaneously, and their current target markets are also highly overlapping. Moreover, shortly after the rise of DA, Ethereum's Dencun upgrade introduced blobspace (an extremely cheap data storage space), which somewhat undermined the narrative value of DA.

Currently, Celestia, Avail, and EigenDA seem to share the market in different overlapping areas, which has led to many technical comparative analyses on "how to choose a DA layer:"

Although many researchers have thoroughly explained the technical status of these projects, in the long run, these three projects will head in different directions. Ultimately, it is expected that each DA layer will find solutions that match its characteristics, just as the territories were divided among Wei, Shu, and Wu during the Three Kingdoms period.

This article will explore the development goals of these three major DA layer projects and their future differentiated development paths.

Background: Data Availability

What is data availability?

Data availability refers to the proof that certain specific data exists on the network. Why is such proof needed?

In the consensus process of blockchain, new blocks are usually propagated from the leader to the nodes, and the nodes need to verify whether the blocks transmitted by the leader are consistent with the blocks actually submitted to the network. Without independent verification, the consensus may be manipulated by a leader hiding malicious transactions.

This logic also applies to L2. The sequencer ensures data visibility by propagating data to full nodes, in which case a process is needed to verify whether the received blocks match the actual blocks submitted to the network.

Source: rollup.wtf

The simplest verification method is to directly check the state of the chain. However, as shown in the above image, submitting data to networks like Ethereum is very inefficient in terms of cost and speed. The emergence of the DA layer is precisely to optimize this process.

What factors should projects considering adopting the DA protocol pay attention to?

First is security. Since the DA layer introduces additional trust anchors, if the DA layer experiences any delays or fails to provide verification, the Optimium lacking an independent data availability mechanism may encounter data unavailability issues. To address this problem, the DA layer will adopt a separate consensus process or introduce Data Availability Sampling (DAS) technology, allowing users to verify data integrity through light clients.

Secondly, DA must provide sufficient performance. With Ethereum significantly improving its DA efficiency through the Dencun upgrade, which added blobspace, the DA layer must offer performance far exceeding that of Ethereum in terms of cost and throughput to attract project adoption.

What types of DA protocols currently exist?

Mainstream options include Ethereum, Celestia, Avail, and EigenDA. Not all chains need to use external DA. According to L2 BEAT data, over half of the Optimium/Validium blockchains use a multi-signature-based DA, referred to as "DAC (Data Availability Committee)"; in this case, data availability is considered quite centralized.

Comparison of DA Protocols

Source: Avail Blog

The following is a comparison presented in the "Guide to Choosing the Right Data Availability Layer:"

EigenDA: The highest throughput. However, it has been criticized for not achieving this key drawback due to economic security. Recently, it has begun to provide economic security guarantees through updates to the penalty mechanism (Slashing). Furthermore, due to its choice to adopt a DAC structure rather than an independent network, it has lower security compared to other consensus-based DA protocols.

Avail: The slowest block time is 20 seconds, but the final confirmation speed is quite fast, only 40 seconds. It supports approximately 1000 validators and allows client verification through DAS to enhance security.

Celestia: Provides a short block time of 6 seconds and high throughput. It guarantees single-slot finality (SSF), which makes compatibility between rollups using Celestia very good. However, for rollups that do not adopt single-slot finality, approximately 10 minutes of challenge period is needed to ensure finality.

In summary, EigenDA offers very high throughput, Avail provides a higher degree of decentralization compared to other DAs, and Celestia offers high scalability for rollups within its ecosystem, all of which are distinctions from Ethereum.

DA Status

As the DA market is expected to be segmented into multiple fields in the future, observing projects that have already integrated protocols reflects the current market landscape better than looking at the prices of related protocol tokens.

EigenDA

The most notable feature of EigenDA's ecosystem partners is that most RaaS (Rollup as a Service) projects, such as AltLayer, Caldera, Conduit, and Gelato, have chosen it. This indicates that projects outsourcing Rollup operations may deliberately choose DA with low data submission costs and without the need to independently run light clients, as they have largely given up on decentralization.

In addition, high-performance chains such as Fluent, SOON, and MegaETH are also worth paying attention to. These chains require a significantly higher volume of real-time submitted data compared to other chains and pursue extreme performance efficiency, making the choice of EigenDA, which currently has the best throughput, a logical one. According to data from L2 BEAT, Mantle and Celo, which have the highest TVL in Validium/Optimium, are also using EigenDA, and these two chains alone account for about 40% of the total TVL in this category (approximately $3.06 billion). As potential chains like SOON and MegaETH go live, this proportion may further increase.

Celestia

Eclipse currently accounts for over 90% of Celestia's data usage. This may be because Eclipse aims to achieve ultra-high performance through GigaCompute, with data upload volumes far exceeding those of other chains, and there was also significant on-chain activity before its TGE.

On the other hand, the Celestia ecosystem has supported various Rollups. Besides the general L2 Manta, almost all the others are specialized Rollups (or application chains). This seems to relate to the trust assumptions of finality mentioned above: because Rollups within the Celestia ecosystem can overcome their relatively long 10-minute DA finality challenge time through single-slot finality (SSF), they achieve both decentralization and performance advantages in the application chain ecosystem. This also explains why the Initia ecosystem chose to use Celestia.

Avail

Apart from DA, Avail also operates various stacks, including the interoperability stack Nexus and the multi-asset consensus layer Fusion, which means it has many partner projects, specifically regarding DA partners:

Application Chains: Fuse, Ternoa, Arcana, OpenLayer, Darwinia, Neova, Stackr

Bitcoin Ecosystem: Yala, Zulu, BVM, (Starkware)

Ethereum L2: Sophon, Lens

Re-staking: Symbiotic

Similar to Celestia, Avail's main ecological partners are also application chains, rather than high-performance L2. However, its uniqueness lies in attempting to create synergies with projects joining its ecosystem by combining DA with the Fusion and Nexus stacks.

Nexus: Integrates and verifies the state of all chains within the ecosystem through zero-knowledge proofs.

Fusion: Providing economic security through staking supported by ETH, BTC, SOL, and all ERC20 tokens.

Avail has quickly captured the Bitcoin L2 ecosystem with its provision of multi-asset consensus, including Bitcoin, and the highest security structurally. Recently, it introduced a re-staking framework (Avail DA + Fusion) through collaboration with Symbiotic. This indicates that Avail's ultimate direction is drastically different from other DAs.

Short-term vision of the DA layer

EigenDA: Pursuing Higher Throughput

EigenLayer completed its slashing mechanism upgrade on April 18 to address long-standing criticisms regarding its lack of economic security. It is expected that many new projects that were previously hesitant to adopt EigenDA due to the absence of a slashing mechanism will begin to integrate.

EigenDA has been criticized for the delayed introduction of the DAS and penalty mechanisms, and the team has committed to achieving decentralization in the long term. However, the short-term upgrades of the protocol clearly show its development direction, including the upcoming Blazar (EigenDA V2) upgrade.

Blazar Upgrade

Public information regarding the Blazar upgrade is relatively limited, but EigenDA aims to enhance latency and throughput through V2, potentially changing the project's infrastructure.

Source: EigenDA Docs

The above image shows the architecture of EigenDA V1 and V2, which adds a new component called "Relay" specifically designed for storing Blob blocks or for high-speed distribution, significantly enhancing the data transmission speed and download performance of DA nodes.

In version V1, the Disperser sends both the blob headers and blob blocks to the DA nodes simultaneously, which leads to excessive network load. Version V2 separates the transmission of blob headers and blob blocks, allowing DA nodes to observe the blob headers and request data as needed. This architecture enhances efficiency while minimizing DDoS risk by reducing data transmission load.

In addition, by removing batch bridging and internalizing Blob confirmations in the Rollup logic, EigenDA aims to reduce Rollup confirmation delays from several minutes to a few seconds.

Despite the multiple updates in V2, it is clear from the Blazar upgrade that EigenDA's mid-term roadmap focuses on enhancing data transfer speeds. Even though it currently offers an impressive throughput of 15 MB/s, it seems intent on dominating the market by further improving performance.

Celestia: Towards the DA Finale

Source: Celestia Blog

Celestia disclosed its roadmap in the blog update of September 2024, as shown in the image above. Its goals can be simplified as follows:

Expand the block size to GB level to support all blockchains;

Optimize light node performance to enable DA verification on all devices.

The roadmap indicates that Celestia is advancing multiple development directions in parallel. Its focus is not limited to a specific area but strives to achieve scalability in various aspects such as throughput, verifiability, and interoperability. This may slow down the completion speed of the roadmap, but it reflects the long-term vision of building a "complete DA" to cover all blockchains.

Celestia's project progress sharing is relatively transparent, holding a developer call (Live dev call) every two weeks, and regularly updating development progress on GitHub. Through this method, we can indirectly confirm Celestia's development roadmap and priorities.

So far, the mainnet has gone through two upgrades, namely Lemongrass and Ginger. The Lemongrass upgrade includes several updates aimed at enhancing the compatibility of the IBC ecosystem, such as cross-chain accounts and packet forwarding modules; while the Ginger upgrade focuses on scalability, reducing block confirmation time to 6 seconds and doubling throughput. The protocol recently completed its third network upgrade "Lotus", which is an update related to token inflation and staking rewards, rather than a major technical upgrade.

It is worth noting that Celestia launched the Mamo-1 testnet on April 14, aiming to significantly enhance performance by increasing the block size to 128 MB and boosting throughput to 21.33 MB/s (more than 16 times the current level). Additionally, Celestia has hinted via tweets that a major update is expected on May 16, which may announce the launch of the Mamo-1 mainnet.

Overall, Celestia continues to share development progress, confirming that it is concurrently advancing multidimensional upgrades. However, with the recent emergence of numerous high-performance L2 solutions, its short-term direction seems to have shifted towards aggressive performance enhancements, as DA throughput is becoming increasingly important.

Avail: Full-stack DA, but scalability first

The above diagram showcases the core concept of Avail: it places greater emphasis on security compared to other DAs, aiming to achieve highly compatible interactions within the ecosystem through the DA, Nexus, and Fusion stack.

However, in the short term, in order to ensure competitiveness with other DA layers, Avail also seems to be prioritizing throughput enhancement. Avail released a series of scalability-related roadmap updates in March and April of this year. The first was the TurboDA protocol disclosed in March, which significantly reduces the final confirmation time for DA to 250 milliseconds. TurboDA is more akin to a layered solution rather than a direct upgrade to AvailDA: when data is submitted to TurboDA, it first provides pre-confirmation like existing rollups, and then completes the final confirmation in AvailDA after 2 blocks. Although this approach may not be entirely secure, its goal is to provide Rollups within the ecosystem with interoperability speeds similar to Celestia SSF levels. (Note: Technical details of TurboDA have not yet been found.)

The second item is the upgrade of the block size to 10 GB level disclosed in April, aimed at maximizing interoperability between rollups within the ecosystem and reducing the block time from the current 20 seconds to 600 milliseconds. To achieve this goal, AvailDA proposed the following plan:

Optimize by shortening the process of generating data and commitments;

Selectively transmit block data based on the Blob header;

Verify transactions unrelated to DA separately through zero-knowledge proof mechanisms, thereby separating the DA verification process from DA. This is fundamentally similar to the direction suggested by other DAs, and if one wants to significantly increase the block size, this seems to be an inevitable direction.

Avail's measures may give the impression that its development direction has changed, as it is known for prioritizing security and interoperability, yet it has recently been continuously proposing performance-related updates. However, the ultimate goal of the performance upgrades is to enhance interoperability between the Rollups joining the ecosystem, so we can conclude that its fundamental direction has not changed.

The future of the DA market

The main competitors in the current DA field, EigenDA, Celestia, and Avail, are clearly targeting the market. In the short term, all three are expected to achieve satisfactory levels of performance and security through high throughput based on large blocks and DAS.

At the same time, the DA market is expected to experience rapid expansion. With the emergence of ultra-high-performance L2s such as MegaETH and Eclipse, the volume of data they need to process far exceeds that of existing Rollups. Processing this data through Ethereum Blobs is too costly, so they have to use high-throughput DA protocols. Additionally, emerging data demand scenarios (such as 0G for chain-level AI computing, Bitcoin L2, etc.) will also give rise to a new DA market.

However, the current DA ecosystem still needs to overcome the following obstacles to expand the market:

First, some projects are building their own DA layers. A typical example is 0 G, which is building and using its own DA layer due to the performance limitations of Celestia and EigenDA and the peculiarities of AI computing. In addition, high-TVL projects such as Metis and Fraxtal are also building and using self-developed solutions, MEMO and FraxtalDA, respectively. This suggests that a DA solution that is implemented autonomously and under control is more beneficial to the project itself in terms of cost and compatibility. On the other hand, most validiums/optimiums currently manage data through multisig-based off-chain storage (DAC), and they have not yet joined the DA ecosystem, which means that the existing DA layer lacks sufficient interoperability incentives or efficiencies to attract these projects.

Secondly, the development of Ethereum DA. Ethereum aims to continuously enhance the availability of Blobs to meet network demands and keep it in alignment with Ethereum L2. The Pectra upgrade has doubled the number of Blobs, and the subsequent Fusaka upgrade will further increase the number of Blobs.

In the long run, by introducing Danksharding, DAS, reducing finality, and shortening block times, Ethereum will be closer to the performance level currently supported by DA protocols. Therefore, from the perspective of DA protocols, it will become increasingly difficult to integrate existing Ethereum Rollups over time. They will have to face the pressure of maintaining a lead over Ethereum's DA developments. This is also the reason why Celestia and Avail focus on their respective Rollup ecosystems; future DA projects will need to provide not only simple DA throughput and cost-effectiveness but also additional advantages such as strong interoperability.

Conclusion: The DA market is facing the "Battle of Red Cliffs"

Returning to the title, the expansion methods of each DA layer are akin to the structure of the Three Kingdoms:

EigenDA is similar to "Wei State" in that it occupies a dominant position in a performance-centric market due to its overwhelming throughput far exceeding that of other DA protocols, as well as the security foundation built through re-staking. However, due to the DAC structure and lack of a forfeiture mechanism, EigenDA still has security vulnerabilities, which pose the biggest challenge it faces in the short term.

Celestia is similar to "Wu State" in that it provides fragmented liquidity and interoperability for application chains (or purpose-based Rollups) based on the convenience of Rollup interoperability and modular structure. Currently, it is focusing on compatibility with various computing environment Rollups by connecting to Initia and Eclipse.

Avail is similar to "Shuguo", featuring a decentralized level that supports 1000 validators, which is the highest among DA projects, and upholds the concept of valuing ecosystem synergies. Its uniqueness lies in its emphasis on building a DA ecosystem based on foundations like Nexus and Fusion, and it is currently capturing markets that other DAs have yet to explore, such as Bitcoin L2.

These three DA layers have distinct characteristics, but short-term upgrades indicate that they may erupt into significant conflicts. EigenDA, Celestia, and Avail are all working to improve throughput, especially aiming for substantial increases based on a 10GB block size. Just as the Battle of Chibi largely determined the fate of the Three Kingdoms, which project can occupy the DA market in the long run will also depend on the direction of performance competition among the DA layers in the short term. During the Three Kingdoms period, Zhuge Liang's strategies greatly changed the outcome of the wars, and in the "Three Kingdoms War" of DA layers, it is worth noting who will play this role, and what unique advantages each DA protocol can offer beyond performance.