While crafting this article, I stumbled upon a discourse between Andre Cronje and Mert, the CEO of Helius Labs, delving into a debate over the practicality of using the ‘transactions per second’ (TPS) metric as a gauge for blockchain scalability. Andre argued that TPS isn't precisely a logical metric for assessing network efficiency, as it revolves more around the current activity on the network. He advocated for transaction capacity per second as a more sensible measurement and time-to-finality, which is the time frame required for a transaction to be considered valid and unalterable, as the ideal metric.

It remains unclear if Andre’s position on TPS as a metric is a direct dig at L2 rollups, albeit, it surely fits the script given that rollups, in their strategy to scale the base layer, marginalized the time-to-finality metric, influenced by block confirmation mechanisms and the choice of approach to consensus. Rollups focus more on transaction throughput, leaving the consensus mechanism to the base layer.

The ideal scenario would involve a solution with a comprehensive approach to scaling applications, rollups, and Ethereum, i.e: prioritizing security whilst enhancing throughput and diminishing latency. This could possibly be achieved through a solution wherein data availability would need to be separated from the base layer. For better understanding, let’s paint a picture using DropBox. 

POV: It’s 2023; you find yourself grappling with a 2013-made laptop that incessantly lags due to a surplus of stored files. Frustration mounts and the urge to smash the laptop arises as tasks take an exasperatingly long time to execute. While the instinctive solution would be to delete files, imagine these are important files that can't be easily discarded.

Enter DropBox: your one-stop solution to help you safely store your files, and conveniently access them without hassle, while freeing your laptop to optimize performance. Now, you can click on a button and have it open almost immediately or respond faster, as there are now fewer files stored on the laptop and more of them stored off it, on DropBox. 

While Celestia might not exactly be Dropbox for the blockchain, it’s however, close in terms of value offering and should help you grasp what it essentially does as we explain within the next few headlines and paragraphs — this time with a wee bit of jargon. The goal remains to provide you with a simplified explanation for clear comprehension while presenting a bullish case for Celestia.

Thanks to “0xngmi” for helping me with the “DropBox” analogy. 

Why we modular 

Celestia operates as a modular layer or modular network, quite different from its monolithic counterparts such as Ethereum and her L2s. The thing is, these monolithic chains handle both their consensus process, settlements, and data storage and processing on the same layer. The result? ridiculous gas fees, slower transactions, and poor user experience. 

Modular blockchains on the other hand, offer a unique solution to this by splitting the network into different parts, allowing different functions on different modules on the network level. With this, the load on the primary chain is reduced. 

Celestia as a modular layer

Going by the explanation above, Celestia functions as a modular layer that handles the consensus and data availability functions of the network on its own layer while the execution is handled on the base layer. 

Prior to modular infrastructure like Celestia, other attempts have been made to scale Ethereum such as Plasma which Vitalik rightly describes as a “blockchain scaling solution that allows all data and computation, except for deposits, withdrawals and Merkle roots, to be kept off-chain” in a recent article. 

This sort of mirrors the concept of side chains which in retrospect has a near semblance of what Celestia ideally is. On the other hand, L2 Rollups (Optimistic rollups and Zero-knowledge rollups) are also an attempt to declutter the computational load of the main network by processing transactions off-chain and submitting the essentials to the mainnet. Both present credible solutions to scale Mainnet and EVM applications but in my opinion, they fall short in a lot of ways as opposed to the Celestia model.

To comprehensively understand Celestia's operation, the primary emphasis should rest on recognizing Celestia's role as a data-availability layer. Let’s delve into this concept by exploring the intricacies of what data availability (DA) involves.

Data availability (DA)

At first glance, Data availability might appear centered around the decentralized storage or processing of transaction data on the blockchain. However, it goes beyond that. The primary concern is validating and securing data submitted to the blockchain, ensuring its integrity, and guarding against malicious activities. Therefore, Data availability focuses on both the accessibility of data by all nodes in the network state and providing proof of the integrity of transaction data. The faster a network can do this, the faster a network achieves time-to-finality.

Monolithic chains face three primary challenges related to their approach to data availability; necessitating full nodes to download, verify, and store all the data. The most obvious issue is that demanding full nodes to download, verify, and store all data will only result in storage problems as the network expands (attracting more transactions). 

This storage challenge poses a risk to decentralization, as it can potentially expose the network to vector attacks, given that running full nodes becomes expensive due to storage requirements. Another challenge Monolithic networks face is that this DA method hampers the network’s throughput (speed of transaction processing) due to the complexities associated with full nodes indulging in the verification/validity of transaction data. Lastly, bandwidth issues no thanks to the volume of data full nodes will necessarily need to slurp in this method. 

The Celestia approach to DA

You don’t need to be a wizard to realize that one of the ways to solve monolithic DA is to simply reduce the load on the mainnet by moving the function to another layer. For example, a career photographer might need another memory storage device to support the storage on the main device (camera). Likewise, there are two approaches to solving this DA issue which requires the storage of data offchain. 

The first is Data Availability Committees (DACs) – A group of nodes that verify availability and validate transaction data on rollups (L2s) before they are sent to mainnet. And secondly, Data Availability Networks (DAN) like Celestia that utilize light clients before full nodes.

Now, this is where it gets interesting and frustrating at the same time – the technical jargon associated with how Celestia functions as a data availability network is a pain in the bosom. So, we’ll try our best to give this the blocmates spin. 

Why Celestia is getting it right

There are two standout features of Celestia’s DA method. The first feature is that the network uses a concept called Data Availability Sampling (DAS) to validate the transaction data, ensuring that the data isn’t malicious (remember we mentioned this as one of the functions of data availability on the blockchain). 

The way this feature works is by allowing light clients (a more efficient network participant in comparison to nodes such as non-custodial wallets) to randomly sample the data as many times as possible on different parts of the block data until it reaches a validity point of 99%. 

The process by which the network achieves random sampling is through the rearrangement of data into larger puzzles using a scheme called 2D Reed-Solomon encoding. Consequently, the light clients query full nodes for corresponding data, randomly picking a few pieces of puzzles. 

The alignment of full nodes with the feedback from queried data determines the rate at which they correspond, validating the queried data. The network's security improves with an increased number of light clients, as full nodes gain the capability to reconstruct data randomly sampled by the light clients or lighter nodes.

The second notable feature is something called Namespaced Merkle Trees (NMTs). This feature allows applications using Celestia as a DA layer to download their data without having to worry about Data from other application clients. In terms of organization and verification of data, Celestia utilizes NMTs to create unique identifiers for each node in the Merkle tree. 

How Celestia can scale Ethereum L2s (The TIA bull case) 

Celestia plans to scale ETH L2s (Optimistic and ZK rollups) using the concept of celestiums, where a celestium represents an ETH L2 that utilizes Celestia for data availability. This is a result of how rollups function — using calldata to post transaction data to Mainnet for settlement. The inefficiency of this approach in terms of how capital-intensive it is to continuously post these calldata even as transactions increase with more users, gave rise to the need for celestiums as a scaling solution. 

Celestiums will function through the Quantum gravity bridge contract which will enable L2 dApps to utilize Celestia for data availability by posting their transaction data to Celestia which will be placed into blocks by Celestia PoS network participants otherwise called validators. Then after, the data will be relayed as DA attestation from Celestia back to Ethereum. 

Celestia’s consensus method will enable slashing for an incorrect DA attestation, as a deterrent against fraudulent proofs, largely enhancing improved network security. This will be achieved through the method we discussed earlier I.e through the use of data availability sampling that allows light nodes to randomly sample transaction data as many times as possible, querying the full nodes in smaller puzzles for correspondence. 

You might wonder why an L2 would require a scaling solution when fees are comparatively low. However, it's important to note that these rollups aren't designed to handle the challenges of a bullish market or widespread adoption, as they may encounter considerable delays, increased throughput, and a surge in fees. 

If you benefited from the Arbitrum airdrop earlier this year, you’ll remember that it serves as a case in point to the submission above, given that the network experienced significant downtime thanks to the volume of transactions by users attempting to claim their airdrop tokens.

With a jar of common sense, one can recognize the significant use case for the $TIA token, Celestia network's native token. Validators earn rewards from fees paid by rollups or developers by staking the TIA token for validation which would become super attractive as more L2s adopt Celestia as a DA layer likewise applications. 

Unlike ETH L2s lacking native tokens for consensus, the TIA token serves as the consensus token for the DA layer of Celestia, while ETH remains in use for execution and settlement. However, this could change as TIA can easily be adopted as the native token for a rollup or L2. 

Thoughts 

L2 rollups, not just applications, are beginning to look towards Celestia as the DA layer for processing transactions. For instance, Movement, a new EVM L2, has only recently announced that it’ll be launching pretty soon using Celestia as its DA network.

Celestia's strength lies in the ability of its light nodes to scale both Celestia and its clients as more users join the network. This ensures that with each app integration as a Celestia client, blocks are created for validation by TIA Stakers, keeping the network busy, staking lucrative, and the entire system secure, all while maintaining decentralization.

It’s even more exciting when you realize that as at the time of this article, the gravity bridge contract is not yet live, and there are more exciting things to come such as Cevmos which will further improve the connection between EVM chains and Cosmos. What this means invariably, is that there’s so much potential and favorable upside for TIA from here.