A Guide to Selecting the Right Data Availability Layer

The focus on scaling execution has brought a new wave of adoption for Layer 2s in the past couple of years. At the same time, more and more participants, faced with growth challenges due to constrained blockspace and prohibitive costs, have now come to recognize that a scalable data availability layer is crucial for effectively scaling blockchains. They have come to realize the need for a cost-effective base layer with expandable block space capable of supporting various types of rollups. 

Avail and several other teams are building scalable data availability solutions from the ground up, while others, such as Ethereum are attempting to increase the data availability capacity in existing blockchains. Regardless of the approach, one fact remains. The base layer developers choose today will define their competitive advantage in the coming years.

Avail is part of a growing modular ecosystem in the ongoing effort to increase data availability for blockchains. Other DA solutions—such as Celestia and EigenDA—are also working on this. Each solution is taking a different path toward blockchain scalability, including Ethereum, which is currently implementing Proto-Danksharding, known as EIP-4844, as a stepping stone towards their long-term goal of full Danksharding.

This post will evaluate the strengths and weaknesses of each. We will highlight different design choices to keep readers informed. Armed with the knowledge brought forward by this comparison, we hope readers can find a DA layer that's most suitable for them.

Let’s start with an overview before diving into each category:

Network Security  

A network’s security and its resiliency are among the first things that come to mind when considering a base layer. The following are the key factors when examining the strength of a network.

Consensus Mechanisms 

In consensus mechanisms, there is a fundamental dilemma between liveness and safety. Liveness ensures transactions are processed swiftly and the network remains operational, while safety guarantees that transactions are accurate and secure. Different blockchain systems make different choices to strike the right balance for their unique use cases.

Avail uses the BABE and GRANDPA consensus mechanisms inherited from the Polkadot SDK. BABE serves as the block production engine and prioritizes liveness by coordinating with validator nodes to identify new block producers. GRANDPA functions as the finality gadget and it enables the simultaneous finalization of all blocks leading up to a specific one as soon as over two-thirds of validators attest to a chain containing that block. This hybrid ledger equips Avail with network resilience and enables it to withstand temporary network partitions and a substantial number of node failures.

Avail’s design choice is similar to Casper and LMD GHOST used in Ethereum. LMD GHOST is Ethereum's block production engine that relies on probabilistic finality like BABE, while Casper FFG, like GRANDPA, is a finality gadget that provides the guarantee of finality.

Celestia’s design choice to use Tendermint enables them to finalize blocks as they are generated. However, the tradeoff of such a choice is the risk of the chain being halted when more than one third of their operators or validators are down. It is also important to note that block finality does not guarantee data availability. A fraud-proof-based design, as used by Celestia, means users need to wait for DA guarantees even in cases where blocks achieve instant finality.

Data Availability Committees or DACs are entities responsible for either supplying or certifying data availability. They use a cryptographic signature to indicate that one, or a supermajority of the committee members agree that the data is available. EigenDA is an off-chain DAC which Ethereum validators have the option to join. DAC members provide smart contract-verified attestations and rely on a separate external service for data ordering.

Decentralization

There are two critical factors to consider when it comes to security of a network: the total staked amount and the distribution of that stake. The degree of decentralization, which is synonymous with how evenly the staked amount is spread, directly influences a network's security. The cost of potential attack is used in evaluating how secure a network is. That is because an adversary trying to attack the network needs to compromise more nodes to capture the same stake if the stake is evenly distributed across a larger set of validators. 

Avail inherits Nominated Proof of Stake (NPoS) from Polkadot which enables it to support up to 1,000 validators. NPoS has an effective reward distribution that mitigates the risk of stake centralization due to its sequential Phragmén method, a multi-winner election method.

In addition, Avail is the only DA layer that can sample from its light client P2P network instead of relying on full nodes for data in the event of a network disruption or bottleneck. This exclusive feature sets Avail apart from all current and planned Data Availability solutions, providing a robust fail-safe mechanism and enhancing the resilience of Avail's Data Availability network.

Celestia adopts Tendermint as its consensus protocol whose validator set is up to a couple of hundred.  

While Ethereum as a monolithic blockchain serves as the gold standard for security with more than 900,000 validator nodes, the level of distribution of the network is inadequately reflected in the number. 

By contrast, Data Availability Committees generally consist of a handful of nodes responsible for confirming data availability to a blockchain. 

It is important to note that re-staking doesn’t borrow the security from Ethereum. Rather, its security relies on the total amount of restaked Ether on its platform. In other words, restaking provides no benefits on its security other than using a small portion of the existing stake locked in Ethereum. 

As a DAC, EigenDA, built on Ethereum, aggregates signatures from its full nodes. Its smart contract-verified attestations fall short in providing a similar DA guarantee level to data availability sampling. EigenLayer utilizes restaking that involves taking locked Ethereum to support its network and its move has also drawn criticism on the risks of re-using validators and overloading consensus.  

Execution Environment Overhead

Monolithic blockchains with smart contracts have introduced groundbreaking innovations over the past decade. Nevertheless, even the cutting-edge technology of its time, as exemplified by Ethereum, where data availability, execution, and settlement are combined as one, has imposed significant scalability limitations. These constraints have spurred the rise of layer-2s that move execution off-chain and prompted the development of improvement initiatives like EIP-4844, also known as Proto-danksharding, and Danksharding. 

Enshrined smart contracts define the state and play the role of a bridge to the rollups. In this approach, Ethereum serves as the authority for validating the accuracy of the rollups.

Avail splits execution and settlement from the base layer and enables rollups to post data directly to Avail. What makes this modular approach powerful is that rollups building on top can easily verify the state by using Avail’s P2P light client network, if used to propagate execution proofs, and have the flexibility to upgrade their rollups, rather than depending on smart contracts and the base layer to define the state. This new approach empowers developers with a base layer that can scale as per demand, giving them the option for bridging and choosing any execution supported layer for settlement.

Celestia has a similar approach with Avail. The only difference is that its light clients cannot yet support the network when full nodes are down.

EigenDA also doesn’t have an enshrined settlement layer. 

Growth Potential

In addition to the security and resiliency of a DA layer, the ability to accommodate increasing demand for rollups and blockchains building on top is paramount for their success. Let's take a look at some of the pivotal factors to consider. 

Validity Proofs

When discussing validity proofs, it's essential to understand the tradeoffs of fraud proofs and validity proofs in DA layers. KZG commitments used by Avail, a type of validity proof for ensuring DA, reduce memory, bandwidth, and storage requirements and provide succinctness, meaning proofs have fixed sizes regardless of polynomial degree. This makes KZG commitments ideal for Zero Knowledge-based blockchains, where efficiency, privacy, and scalability are essential.

In addition, Avail's light clients can promptly access and sample data and ensure correct block encoding and provide data availability guarantees upon the finalization of new blocks, unlike fraud-proofs which require a period of waiting for the conclusion of a challenge period. The combination of KZG commitments and Avail’s light clients expedites the verification process on Avail, allowing rollups or sovereign chains building on top to capitalize on its swift verification process and creating scalability and flexibility for blockchain design for years to come. This approach to verification is a pivotal factor that sets Avail apart from its peers such as Celestia.

Celestia utilizes secure hash functions, which is much faster than KZG commitment generation. The tradeoff here is that they have to rely on fraud proofs to confirm the accuracy of erasure coding, which introduces potential delays in ensuring data availability guarantees. 

Celestia’s light nodes cannot definitively confirm whether the data is available or if a pending fraud proof is yet to be received. In other words, the use of fraud proofs reduces the ability of the network's light nodes to definitively confirm data availability after sampling due to an obligatory challenge period as part of optimistic verification.

EigenDA will use KZG commitments and download only small amounts of data instead of complete blobs and employ validity proofs. Their approach is to use erasure coding to split data into smaller chunks and requires operators to download and store only a single chunk, which is a fraction of the full data blob size.

As for Ethereum, while the current iteration is not utilizing validity proofs, EIP-4844 and full Danksharding will adopt validity proofs when implemented.

Ability to Scale

The limitation such as prohibitive costs and slow transactions on Ethereum has prompted a surge of L2s. They have emerged as the execution layer for Ethereum, driving demand for blockspace higher. Publishing data to Ethereum today accounts for an estimate of 70 percent to 90 percent of the total cost for rollups. Expanding blockspace would result in additional costs for validators and applications developed on Ethereum.

Base layers such as Avail and Celestia are designed to solve this problem. Optimized for data availability, they have the capability to dynamically scale block size as demand increases. By incorporating light clients with Data Availability Sampling (DAS), they have the advantage of expanding Data Availability block sizes in response to rising demand for their network. This means that as block space increases, applications built on top remain unaffected because light clients within these networks can perform DAS without the need to download entire blocks. This unique capability sets them apart from monolithic blockchains.

Ethereum has the biggest community with a market capitalization of $191 billion as of Sept. 2023. While the protocols building on Ethereum enjoy economies of scale, they also face prohibitive transaction costs due to limited blockspace in the past years. The number of users and transactions have peaked amid the growth of rollups that have become the best choice for execution. As blockchain technology becomes more prevalent, the need for blockspace will only increase going forward.

While DACs can scale given their simplistic centralized approach, several rollups use DACs as a temporary measure until they come up with a decentralized DA solution.

Data Availability Sampling

Avail and Celestia both support light clients with data availability sampling (DAS), allowing light clients to provide trust-minimized security. The main differences as covered earlier, are how verification is conducted, and how Avail’s light client P2P network can replace full nodes to support the network in case of disruptions or bottlenecks. 

By comparison, Ethereum post EIP-4844, won't come equipped with DAS. This means its light clients won't have this upgraded, trust-minimized security feature. To complicate things further, Ethereum’s DA solution includes housing its smart contract environment. With full danksharding, DAS will be implemented to scale blobspace which is predicted to be in a few years from now.

EigenDA's security is built upon trust in a handful of full nodes or another entity due to the absence of Data Availability Sampling (DAS). The protocol's integrity relies on a supermajority within the committee being honest, and at least one additional entity holding a data copy, akin to an optimistic construction. Although the dual quorum approach increases security compared to a single quorum, it falls short of the ideal scenario where independent verification through DAS would be possible.

Cost

Ethereum is the most expensive solution in relation to congestion and demand. Even with EIP-4844 Ethereum will still be costly as it provides only a one-time increase in blockspace. DACs are the cheapest, but this comes at the cost of adopting a more centralized approach. 

By not having an execution layer, Avail and Celestia will be able to keep operational costs low. They can also easily grow blockspace, while Ethereum today cannot without DAS.

As for EigenDA, it has said it will introduce a flexible cost model for both variable and fixed fees but its actual cost is yet to be shared.

Performance Highlights

Now that we have examined growth potential, we will have a look at performance of these blockchains.

Block Time

See the table above for the time required to build a block for each.

Measuring a blockchain's performance by the time required to build a block offers limited insight because this metric only addresses one aspect of the process from block confirmation to verification completion. Even with a consensus mechanism that provides instant finality, DA verification can take time when using a fraud-proof-based approach. 

Ethereum uses Casper to finalize blocks every 64-95 slots, which means the finality for Ethereum blocks is roughly 12-15 minutes.

EigenLayer is not a blockchain but a set of smart contracts that run on Ethereum. Meaning it inherits the same finality time as Ethereum. So, if a user sends a transaction to a rollup, the rollup will need to forward the data for that transaction to EigenLayer to prove the data is available. However, the transaction will only be considered complete once the Ethereum block is finalized, even though the rollup has already accepted the transaction, resulting in a delay. Ways to circumvent the issue by providing faster DA guarantees with crypto-economic measures have been in discussion. 

Block Space

As rollups become the execution layer of the future, the need for blockspace will only increase going forward. DA layers like Avail and Celestia will be able to accommodate demand because of their modular design, whereas the growth for blockspace for Ethereum will be limited. Avail’s Kate testnet has configured the block size to be 2MB, which gets duplicated and erasure coded to 4MB. What sets Avail apart is its ability to increase the block size using efficient client-side verification techniques. Through internal benchmarking, Avail has tested block sizes up to 128MB without difficulty. Celestia is also able to increase block size as demand for block space increases with DAS.

EigenDA will expand throughput through decoupling DA and consensus, erasure encoding, and direct unicast. However, it comes at the cost of the rollups building on top not being able to inherit the censorship resistance of the base layer.

In Conclusion

Choosing a robust base layer to build on can be challenging. We hope this post helps our readers understand more about the pros and cons of different design choices and select the DA layer that works for you.

To stay updated, follow us on Twitter and subscribe to our newsletter. If any of the points raised in this post require further clarity, feel free to reach the team on the Avail Forum.