To understand sequencers it’s first important to revisit the topic of what a layer 2 network is vs a layer 1.
Taking Ethereum as an example layer 1, there’s a block (referred to as a slot) every 12 seconds which can fit up to 30million gas worth of transaction related computation. When there is more demand for transactions than block space, transactions are waiting around in a pool to be picked up, and so other transactions pay increasingly higher fees to try to secure space in front of them in the next blocks. This congestion has led to periods of inflated transaction fees on Ethereum and was a catalyst for the introduction of layer 2s which aim to offer parallel rails of blockspace with low fees and quick transaction inclusion time.
You can therefore think of layer 2s as a parallel highway for transactions, and crucially they have a connection back to the layer 1 with a compressed version of transactions being committed to the layer 1 periodically. This is where the term “rollup” comes from — because the layer 2 transactions are ‘rolled up’ and stored on the layer 1. There’s different implementations of this per layer 2 type but for this piece the central question is who’s doing this checkpointing back to the layer 1 and how does it select which transactions to process next?
This is where the sequencer comes in.
The sequencer is responsible for choosing which transactions to put in the next layer 2 batch, verifying them and committing a compressed version of them back to the layer 1. For this work it earns a fee.
However at the moment, all sequencers on popular layer 2 are being run by the foundation/labs/development team of the blockchain and therefore operate as a centralisation point. The risks with this are that;
- The sequencer would have the ability not to process certain transactions, whether that’s because they’re linked to illicit activity or for any other potentially capricious reason
- Regulators could focus actions against the centraliser as it has ultimate control for the ‘processing’ of transactions. It’s a short hop to imagine someone like OFAC sanctioning a sequencer if it’s found that the related blockchain is seeing high amounts of illicit activity
- They operate as a single point of failure and so any bug or attack on a sequencer could bring the layer 2’s network to a halt
- As transaction ordering is a key component in MEV, a corrupt and controllable sequencer could order transaction for their own or a colluders benefit
Due to these risks as well as the general approach that decentralisation = good, most layer 2s claim to be on a path to sequencer decentralisation.
One potential option is shared sequencers whereby rollups/layer 2s make use of a sequencer that’s ordering and committing transactions for not just one but multiple layer 2s. This approach is taken by Espresso who’s sequencer can be leveraged by both Optimistic and ZK Rollups and which uses their HotShot consensus protocol to order transactions and compress transactions. It’s currently in testnet only and there are open questions about how revenue would be split in a shared sequencer model — especially where the transaction volume and revenue may be disproportionately coming from one or two networks within the group.
Another approach is the ability to open up the sequencer set and allow either whitelisted group of sequencers to be selected across (Proof of Authority), to provide a fully decentralised model where anyone can opt to run a sequencer and the sequencer per block/epoch/session is selected via a consensus mechanism or governance approach (Based rollups). A further approach is using Distributed Validator Technology (DVT) which would share the responsibilities of a single sequencer across a number of node operators that would each use their fractional share of the sequencer key to vote on an order and the validity of the transaction compression.
Further innovations will likely be explored and especially those which can help achieve cross chain atomicity — the processing and committing of transactions to multiple blockchains in such a way that if one transaction fails then all connected transactions fail. The use case for this would be a multi-asset transaction which requires a number of trades across different DEXs on different blockchains. Likewise, as different chains take different approaches to MEV it may be that sequencers need to help protect users against MEV on some chains and exploit MEV for their benefit on others.
However with Coinbase earning $3.4m in fees as the solo sequencer since the Aug launch of its layer 2 Base, and monthly sequencer fees in the hundreds of thousands for other layer 2s, it’s clear that there’s a tension between the goal of decentralisation across all parts of the blockchain’s infrastructure stack and a notable revenue generation stream for the protocol creation and/or development team. Whether the layer 2s move forward apace with decentralisation and this revenue sharing, especially in this continuing bear market, is therefore a question mark.