L0, L1, L2 Scaling
Disclaimer: This is not financial advice. Anything stated in this article is for informational purposes only, and should not be relied upon as a basis for investment decisions. Chris Keshian may maintain positions in any of the assets or projects discussed on this website.
To subscribe to my mailing list, input your email here.
Today, most of the blockchains you have heard of or interact with are known as Layer 1 blockchains (L1s). This means that they are the base layer blockchain that handles everything a blockchain is supposed to do - transaction, computation, and settlement. Decentralized Applications (DApps) can be built on Layer 1 blockchains, and likely most of the DApps you have interfaced with are L1 DApps.
The problem with most L1s is that throughput on blockchains is low, making them slow, costly, and unusable for many tasks. Bitcoin manages to perform 7 transactions per second, while Ethereum can perform 15-20 TPS. To put this in context, Visa can do 1,700 TPS. If blockchains realize their full potential, this transaction time must improve.
So how do we do this?
I categorize the available scaling solutions in three broad buckets:
Scale the Layer-1 (L1) directly
Share the load with Layer-2 networks (L2s) and create a modular stack
Create a system that permits multi-chain interoperability, with task-specific appchains
Let’s look at these one by one.
Scale the Layer-1 directly
Solutions that take this path attempt to change the rules of the L1 protocol directly. There are a few main ways to do this:
Increase the amount of data contained in each block, like in the Bitcoin block size limit controversy in 2017
Accelerate the rate at which blocks are confirmed, like alternate L1 models such as Solana’s Tower Consensus
Modify the consensus mechanism to decrease the computing power, as Ethereum did when it shifted from Proof of Work to Proof of Stake
Break the state of the entire blockchain into distinct datasets (i.e. Sharding - more on this below)
All of these methods increase the overall throughput of the main chain. The biggest current event regarding L1-direct scaling is ETH 2, which makes two fundamental updates to Ethereum:
Shifts the Ethereum network from PoW to PoS (#3 above). With the shift to PoS, instead of requiring miners to solve cryptographic algorithms using substantial computing power, PoS systems process and validate new blocks of transaction data based on participants staking collateral in the network.
Implements Sharding (#4 above). Sharding entails breaking the state of the entire blockchain network into distinct datasets called "shards" — a more manageable task than requiring all nodes to maintain the entire network. These network shards are simultaneously processed in parallel by the network, allowing for sequential work on numerous transactions.
Eth 2 has been delayed many times, and as such, the community has concurrently pursued alternative means of scaling. Thankfully, there is a second class of scaling solutions - Layer-2 scaling solutions.
Share the load with Layer-2 networks and create a modular stack
As discussed above, a Layer-1 network refers to a monolithic blockchain. A Layer-2 protocol is a third-party integration that can be used in conjunction with a Layer-1 blockchain. Scaling solutions that pursue this path attempt to shift a portion of a blockchain protocol’s transactional burden to the L2 architecture. The L2 protocol then handles the brunt of the network’s processing and only subsequently reports back to the main blockchain to finalize the results.
In more advanced forms of this approach, you may hear this referred to as a “modular stack”. In order to describe this, let’s first understand the components of blockchain actions:
Data Availability (DA) – Ensuring the transaction data behind rollup block headers has been published and made available so that anyone can recreate the state.
Execution – Computation taking the pre-state → run transactions → transition to the post-state.
Settlement – This varies based on the implementation, but tasks can include verifying/arbitrating proofs and coordinating cross-chain asset transfers/arbitrary messaging.
Consensus – At minimum agreement over the transactions and their ordering.
A monolithic L1 performs all of these actions, while a modular stack strips these tasks apart into separate components.
A new paradigm is emerging, with the goal of creating a completely modular blockchain stack. Projects like Celestia are leaning into this vision. Celestia provides the data availability and consensus layer, and fosters an architecture that permits other blockchains to service different layers of the stack - settlement and execution. In this model, pieces of the stack could be swapped for each other, permitting builders to mix and match components for their optimal function.
In the modular stack, the L2 is responsible for the execution component. By shifting the majority of data processing to the L2, the base layer becomes less congested, thereby increasing the scalability of the L1 chain.
The three most discussed L2 scaling solutions are:
State Channels
Side Chains
Rollups (Optimistic and ZK Rollups)
In this section we will focus on the two most popular L2 scaling approaches: Optimistic Rollups and ZK Rollups. Rollups work by moving computation tasks off the chain, while posting the compressed data on the Ethereum (or the alternative settlement chain) main network. Rollups can also feature off-chain data availability, which does not post data on Ethereum, thereby resulting in significant throughput improvements.
Rollups primarily involve smart contracts which “roll up” or bundle the transaction data and then move it off the chain for processing tasks. Network participants such as validators or sequences manage the data and submit collections of highly compressed transaction data to the main chain.
Optimistic Rollups and Zero-Knowledge Rollups differ in their methods for verifying each rollup.
Optimistic Rollups assume that all transactions are valid and submit batches of transactions without any computation. Optimistic Rollups also feature a “challenging period” during which anyone can challenge the authenticity of the transaction batch.
Zero-Knowledge Rollups create cryptographic proofs which help prove transaction validity. Every transaction batch features a personal ‘validity proof,’ which must be submitted on the main chain.
Here is a brief list of some of the companies pursuing these two approaches:
Optimistic Rollups
Optimism
Arbitrum
ZK Rollups
Starkware
Immutable X
Aztec
zkSync
Loopring
Optimistic Rollups
As Optimistic Rollups, Arbitrum and Optimism share many similarities:
Validators stake ETH and have an incentive to act honestly.
Rollups are optimistic because they are assumed to be valid at the time of the transaction.
Both have full nodes, accumulating layer-one transactions, and validator nodes monitoring the chain state.
Both relay only the calldata with the hashes of confirmed rollup blocks to the mainnet.
DApps on both chains can select their own validators. Since transactions are validated locally instead of by all nodes, nodes have to communicate less with each other, increasing transaction throughput.
Their main differences are:
Optimism uses single-round fraud proofs, Arbitrum uses multiple-round fraud proofs. Put simply, Optimism executes transaction proof in one round on the layer-one chain, Arbitrum does this in several rounds off-chain.
Optimism is EVM-compatible. So is Arbitrum, but Arbitrum also has its own Arbitrum Virtual Machine (AVM).
Optimism has a Solidity compiler, Arbitrum supports all EVM programming languages.
As you can see, the differences are mostly technical and of little interest to the everyday user. Arbitrum, however, has a significantly bigger ecosystem with a higher TVL and more active DApps.
ZK-Rollups
There are a couple of differences between Optimistic and ZK-rollup solutions that are important to note:
ZK-rollups need to submit cryptographic proof to the L1 to validate their transaction. This is more complex and thus costlier.
Optimistic rollups are always EVM-compatible; ZK-rollups are not.
Optimistic rollups have trustless visibility, but not all ZK-rollups do, so some transactions may not be traceable from start to finish.
Bridging is a better experience on ZK L2s because there is no seven-day waiting period to withdraw funds to layer one.
I believe ZK Rollups will ultimately capture the lionshare of the L2 market, though the development has been slower than Optimistic Rollup solutions. As such, Optimism and Arbitrum are both capturing market share quickly. ZK rollups will likely catch up since, from a technical perspective, they are more attractive, more efficient, and allow ten times as many transactions as Optimistic rollups.
Create a system that permits multi-chain interoperability, with task-specific appchains
The final bucket of scaling solutions is broader, and involves creating an ecosystem of independent chains that are all connected either through a Layer-0 substrate, a messaging protocol, or a network of bridges.
Layer-0 Ecosystems
A Layer-0 (L0) can be thought of as an interconnectivity network. L1 chains can be built on top of L0s. L0s allow for cross-chain interoperability amongst all L1s built within their ecosystem. This permits blockchains to communicate with each other by sending data and value seamlessly from chain to chain.
The two most prominent Layer-0 blockchain protocols are Cosmos and Polkadot:
Cosmos describes itself as “a decentralized network of independent parallel blockchains, each powered by Byzantine Fault-Tolerance consensus algorithms, which guarantees safety for up to a third of Byzantine, or malicious, actors.” Cosmos wants to create an “Internet of Blockchain” where blockchains built using its open-source tools Tendermint, Cosmos SDK, and IBC can communicate with each other. Many well-known projects like Binance and Polygon have created blockchains using Cosmos and are interoperable in the Cosmos ecosystem.
Polkadot describes itself as “a next-generation blockchain protocol that unites an entire network of purpose-built blockchains, allowing them to operate seamlessly together at scale.” Polkadot utilizes blockchain shards called parachains and parathreads, sovereign blockchains, which connect to and are secured by the Polkadot Relay Chain. Polkadot’s bridges allow these parachains and parathreads to connect to and communicate with external networks like Ethereum and Bitcoin. By using Polkadot’s Substrate framework, anyone can build an application-specific Polkadot parachain. Many well-known projects like Moonbeam, Efinity, and Acala have created blockchains using Polkadot and are interoperable through Polkadot’s ecosystem.
Bridges connecting multiple L0, L1, and L2 ecosystems
Another iteration of creating this fluid, multi-chain system is to connect many different chains together using bridges. Bridges are a means of moving data and value amongst multiple chains, whether that be from an L0 ecosystem to L1 ecosystem, or from an L2 execution environment to a L1 chain. This multichain, interoperable vision is prevalent today, and could perpetuate, with different chains and layers of the stack serving different functions or niche services within the broader blockchain ecosystem.
Conclusion
The blockchain scaling landscape is evolving rapidly, and there are diverging opinions on the end state. My guess is that all of these options will coexist for the foreseeable future, with smart teams working separately on each approach. Ultimately, I think the market will converge on the optimal solution, though I expect many years of experimentation before that point.
