The DA(C) Problem

In decentralized networks, we value the phrase "Don't trust, verify". As a validator in the network, you want to be confident that the information you receive is correct, which you do by reconstructing and comparing the state (e.g. blockchain state) from the data you have received (e.g. transactions).

Data Availability is the confidence a participant in a network has that all data that is required to verify newly published data is available to all network participants. If it isn't, it would mean a data withholding attack is taking place, jeopardizing the decentralization and liveliness of the network. On blockchains, data availability simply means downloading a block from other peers in the network. If the block's data is available, you can download and verify it. If it isn't, the block won't be added to the network. For Layer 2 scaling solutions, which do not have the same block production mechanism, the problem of data publishing is much more complex.

Solving data availability for L2s

At the core of data availability lays data redundancy. The original data is stored in many different nodes in the network, so that you can be confident the data was available for many different participants. This becomes hard to verify without downloading all the data over and over again. To mitigate this, redundancy is implemented by erasure encoding the original data and dispersing its segments to different nodes in the network. The network of nodes responsible for processing these erasure encoded segments form the Data Availability Committee (DAC). The ultimate goal of these DACs is to provide Layer2s with a cheap and efficient way to verify its own data availability without having to download all data in the DAC.

In a DAC, the erasure encoded chunks are processed by the different nodes, which together create a DAC certificate that attests the blob's data is fully available. This process, however, requires you to trust the DAC, which makes the DAC permissioned in nature. Examples of such DACs are Arbitrum AnyTrust, which is a DAC consisting of reputable parties like P2P, Consensys, and Google Cloud.

To bring permissionless and scalable data availability layers to life, two directions have emerged: Modular Blockchains and Middleware DAC. Essentially, they're both providing data availability through a committee, but they do so in very different ways.

Modular Blockchains

Modular blockchains differentiate themselves from monolithic blockchains by seperating execution, settlement, consensus, and data availability. For this documentation, we'll only focus on data availability and doing so from the user's perspective.

In a modular blockchain, a user submits a blob-carrying transaction to the blockchain. Validators verify the transaction and create data availability attestations for erasure encoded chunks that are dispersed through the P2P networking layer. These attestations are, in most modular blockchains such as Avail and Celestia, send to smart contracts for Layer 2s to access. In the blockchain itself, however, the chunks are served through data availability sampling executed by clients.

The issue with modular blockchains is that they can only handle a very small amount of data: currently 2MB every 15-20 seconds. The plan is to scale these throughputs up to 4MB in the short term and around 30MB on the long run. While this is more than enough for rollups, other decentralized networks such as AI, DePIN, gaming, DeFi, and high-throughput L2s will require arguable much more. The main limitation in throughput is to maintain efficiency in data availability sampling and saving bandwidth in the consensus process of the blockchain.

Middleware DAC

The case that Middleware DA layers, such as EigenDA and HyveDA make, is that DA is not a problem that necessitates consensus. A network of nodes could simply provide attestations that they have processed the data and utilize succinct cryptography to make these attestations verifiable. The issue remains, however, that DACs are permissioned, which is an extra security assumption that the user should not be having to make.

As a solution, we can simply replace the permissioned network of DA nodes with a Delegated Proof of Stake network of DA nodes where the stake is slashable. DA nodes will be incentivized to create DA attestations and will be slashed when they act wrongfully. Important information about the attestations will be submitted on-chain, so that it can be verified and utilized in smart contracts, such as rollup contracts.

The great attribute of permissionless DACs is that the nodes do not need to reach consensus to participate, making scaling much easier and allowing for much higher throughputs of data at much lower costs.

HyveDA: Fully Permission-less, High-Throughput DAC

HyveDA creates ultra-high-throughput data availability using a fully permissionless data availability committee built on Symbiotic shared security. The DA network removes any trust assumption on the committee members, while also mitigating attack-vectors on the network. The documentation in this section will take you through the architecture of our network, answering questions about the communication between nodes and considering adversarial behaviour.