Transaction sequencing has become a growing issue in the Layer-2 (“L2”) world. The primary role of an L2 rollup is to provide a secure venue for cheaper transactions. L2 rollups provide execution layers for users and then submit their transaction data to the parent Layer-1 (“L1”), i.e., Ethereum in the case of Arbitrum, Optimism, zkSync, etc.
Sequencers are the entities that have been given the right to order these transactions into groups. A sequencer receives unordered transactions from users, processes them into groups off-chain, and generates a compressed batch of ordered transactions. The transactions can then be put into blocks and sent to the parent L1.
Rollups do not actually need a sequencer; it is a design choice for a better user experience in the form of cheaper fees and quicker transaction confirmations. For instance, like how most rollups use the Ethereum base layer for data availability, they can also use it for sequencing. However, Ethereum’s base layer is likely to be relatively inefficient and expensive. This has meant that every major L2 rollup project has, so far, found it more convenient, cheaper, and user-friendly to run a centralized sequencer.
As the sequencer controls the ordering of transactions, it has the power to censor user transactions (although complete censorship is unlikely as users can submit transactions directly to the L1). The sequencer can also extract the maximal extractable value (“MEV”), which could be economically harmful to the user base. Furthermore, liveness can be a major issue, i.e., if the sole, centralized sequencer goes down, then the entire rollup gets affected.
The solution to the problem is shared, decentralized sequencers. Shared sequencers essentially provide decentralization-as-a-service to rollups. In addition to solving the issues of censorship, MEV extraction, and liveness, shared sequencers also introduce cross-rollup composability, unlocking all sorts of new possibilities.
Espresso, Astria, and Radius are working on innovative shared sequencing solutions with various unique features in their respective architectures. While Espresso seeks to leverage EigenLayer to bootstrap its network, Astria maintains close ties with the modular data availability network Celestia. Radius brings its unique encrypted mempool to the conversation.