Decentralizing Rollups
A brief introduction to creating a decentralised mechanism in rollups
Blockchain and cryptocurrencies were born out of the desire to establish/protect individual freedom and privacy from the potential threats of the digital age. One of the cornerstones of this desire is decentralisation. Dependence on a centre limits individual freedom and leads to the domination of an authority or a third party. Ethereum, as one of the most important representatives of this philosophy, establishes the financial privacy of the individual and allows these transactions to take place in a decentralised manner.
The current paradigm we have for scaling Ethereum is based on "rollups". Rollups can be summarised simply as transactions that take place off-chain and send proofs of the transactions to Ethereum within certain time periods. The proofs are sent by a machine called the "Sequencer".
The sequencer is a critical component of rollups. It organises and sequences the transactions, which makes the network run faster and cheaper. However, Sequencer is not a "trustless" machine, which we currently have to rely on. This is something we do not like about Sequencer in terms of decentralisation. This is where the decentralisation of the Sequencer becomes not only a technical issue, but also a very fundamental philosophical one.
In this paper, various Sequencer mechanisms are examined and analysed in detail.
Decentralising The Sequencer
As mentioned above, the Sequencer arranges transactions and sends them in batches to the smart contract on L1. However, it is possible for users to transfer the transactions they have executed on L2 to L1 on their own, although this method is unproductive and costly. Therefore, the Sequencer processes these actions in batches. In case a user needs to transfer his L2 balance back to L1, L1 does not have accurate information about L2, so it needs a proof along with the withdrawal request. If a user wants to withdraw, he is completely dependent on the Sequencer. The Sequencer has the right to censor a user's transactions without giving any reason. However, if a user believes that their transactions are being censored, they should be able to add their transactions to L1, just like the Sequencer.
Some of the mechanisms are;
Anarchy: Anarchy is the most straightforward method, where anyone can send transactions to L1, but it also has some disadvantages. Several individuals may attempt to perform this task simultaneously, but only one participant's transaction will be successful. This results in futile efforts and an avoidable consumption of gas.
Single/Centralised Sequencer: Today, many different rollups use a single sequencer. This is the most efficient method, but has drawbacks in terms of censorship and liveness.
- Censorship: As mentioned above, the sequencer prevents or delays the inclusion of transactions.
- Liveness: If the sequencer crashes, there is a backup sequencer to take its place. If the backup sequencer also stops, anyone should be able to join the system.
Backup sequencers can be selected by the administration. This creates a more reliable area in terms of censorship resistance (CR) and liveness. The single sequencer can thus continue as the main approach for a longer period of time.
Auction: Anyone can bid for the right to execute transactions for a certain period of time (which may be for a certain block). This auction can be conducted using methods such as the Dutch auction (a method where the opening price is quoted at the beginning and then the price slowly decreases) and the sealed-bid second-price auction (a method where bids are submitted secretly and the highest bidder pays the second highest price at the end of the period).
PoS-style Leader Election: Anyone who wants to join the system has to lock a number of tokens (this token can be the protocol's own token or ETH) into the rollup contract. The probability of winning the right to become a sequencer is directly proportional to the amount of tokens locked. Malicious sequencers are slashed and their locked tokens are confiscated.
Proof of Authority: Instead of a single sequencer, there are multiple distributed sequencers. These can be trusted individuals or companies (such as Arbitrum AnyTrust). The fact that these sequencers work in sequence can be an incentive for them to behave honestly. This method is generally safe in terms of CR and Liveness, but has its flaws.
Based Rollup: With this method, rollups can perform the task that the sequencer will do through Ethereum. (It is a method that creates space to implement the anarchy method mentioned above, without any drawbacks; for a detailed reading, see Justin Drake's article).
- L2 miners (as opposed to the traditional term mining, which refers to those who forward transactions from the L2 mempool) collect the transactions made in L2 into packets and forward them to the provers (those who forward L2 transactions to the L2 block);
- Provers receive L2 transactions from L2 miners and create a block;
- L1 miners (used in the same sense) include L2 blocks in L1 transactions.
Shared Sequencer
As mentioned above, traditional rollup decentralisation methods assume individual solutions for each rollup sequencer. However, it is also possible to design it as a distributed "sequencer network" instead of a single sequencer. There are several advantages to using a shared sequencer;
Modularity: Rollups can be easily integrated with a shared sequencer, essentially providing Sequencers-as-a-Service (SaaS).
Pooled Security: Consolidation improves security and streamlines conflict resolution.
Fast Transactions: Fast sub-L1 block-time confirmations are maintained.
Cross-chain Atomicity: Enables simultaneous multi-chain transactions.
However, using only the native L1 for multiple L2s has limitations in terms of data throughput and transaction speed. The question is whether specialised shared sequencers can outperform the traditional L1 sequencing approach. Espresso and Astria are working on this.
OK. Where is my token?
To implement the decentralisation solutions mentioned above, a token is essential. If there is an auction, a token is needed to execute the bids, if there is a PoS leader election, a token is needed to lock, if a based rollup is to be built, a token is needed.
For example, a central sequencer will finalise the transactions in L2 in a very short time (in fact, there is no finality in rollups, there is a pre-confirmation called "soft confirmation"). This "soft confirmation" decision is taken by the sequencer itself, in a centralised manner. The way to decide and implement it in a decentralised way is to create a local consensus on L2. To participate in a local consensus, a token is also required.
In addition to these critical needs, this token can also be used to pay transaction fees and to have a say in changes and improvements to the network.
Conclusion
Decentralisation is essential for financial freedom and privacy. Ethereum has made this idea its core vision. The current paradigm for scaling Ethereum is based on rollups. However, rollups have not yet achieved full decentralisation. There are several approaches to decentralising rollups. Anarchy, Single Sequencer, Auction, POS Leader Election, Proof of Authority, Based Rollup are some of them. With the Shared Sequencer architecture, it is also possible to create a distributed network of sequencers. This is an important approach to achieve decentralisation.
Resources:
Vitalik, (2021). An Incomplete Guide to Rollups
L2 MEV wat by Lisa from Taiko Labs
Starkware, (2022). A Decentralization and Governance Proposal for StarkNet
Based Rollups by Justin Drake
New Token Model for Layer 2 by Fuel Labs
And great article by Charbonneau, who summarises all aspects of rollups, which I have benefited from a lot; Rollups Aren’t Real by Charbonneau