Shared Sequencing
TLDR: Traditional sequencer models are largely vulnerable to censorship and MEV-type attacks. Shared sequencing aims to start tackling these and wider L2 scaling concerns such as rollup interoperability, however it is constrained by some key technical considerations. Startups such as Espresso and Astria are currently implementing this technology for use on Ethereum L2s.
Shared Sequencers vs Normal Sequencers
Sequencers are responsible for receiving, organising, and grouping transactions to be sent from Layer 2 (L2) rollups to Layer 1 (L1) base chains. Their importance in L1 <-> L2 communication garners them great influence in how transactions originating from L2s are ordered, raising concerns over centralised attack vectors. Shared sequencers aim to use a common, decentralised sequencer amongst multiple rollups to alleviate these concerns, as well as tackle some wider issues surrounding rollup fragmentation and interoperability.
In the centralised sequencing model (Figure 1), transactions submitted via an L2 rollup are submitted to a sequencer node, where they are ordered and pre-confirmed to the L2’s block space. Upon completion, the sequencer bundles these transactions into a single one to be committed to the L1 base chain. Once this bundled transaction is confirmed, verifier L2 nodes similarly record the transaction on their copies of the L2 chain. Finally, verifier nodes compare their new version of the state to that of the sequencer by checking that their state roots (essentially, a hash of the block) match. In doing so, they confirm the new state of the L2 chain and finalise the transactions.
Because sequencers choose the order of transactions as they get sent from the L2, a malicious sequencer could send in its own transactions alongside legitimate transactions to extract value from any price movements, similar to L1 frontrunning and sandwich attacks. This could generate profit at the expense of these legitimate users, reducing yields on their transactions. Also, certain centralised sequencers are potentially vulnerable to downtime/technological failures and performance issues as a result of localised hosting. Additionally, a malicious sequencer may choose not to process transactions from a certain wallet address, which opens up the possibility of entities abusing that leverage.
While the architecture of shared sequencing solutions vary, the core tenet pertains to the decentralisation and comingling of the sequencing process across separate L2 rollups. All of the aforementioned issues can be mitigated through the decentralisation that shared sequencers offer, limiting the leverage available for any single actor to perform such attacks.
It is important to note that many large L2s with centralised sequencing have balances in place to prevent censorship abuse from taking place. For example, the Arbitrum sequencer code has a `forceInclusion` method that forces a transaction to be accepted by the sequencer and relayed, although here the sequencer can still perform the previously discussed sandwich attacks through determining how the forced transaction is sequenced. Shared sequencing still offers an improvement here, by removing the possibility of censorship in conjunction with preventing the strategic sequencing of transactions for profit.
L2s face many scalability issues that stem from the siloed nature of each L2’s operations. Specifically, the fragmentation of liquidity amongst disjointed L2s and the inability to perform atomic cross-L2 transactions. Liquidity fragmentation causes poorer trading conditions on these L2s as traders are faced with higher price volatility and price variation between L2s. Utilising a shared sequencer would help mitigate this by allowing for seamless asset bridging across these disjoint platforms through the implementation of atomic cross-rollup transactions, reducing the friction associated with moving tokens across rollups.
In the case of a cross-rollup token swap without a shared sequencer, the initiator would put themself at a disadvantage as the other party would be able to decide whether or not they want to commit to the transaction depending on how market conditions evolve after the initiator has finalised their transaction. A shared sequencer would mitigate this asymmetry by coupling the inclusion of both rollups’ individual transactions such that if either were to not be included into the rollup the other would also be omitted. As a result of the ability to decrease friction in cross-rollup transactions, this would also facilitate much greater interoperability between different L2s, allowing users to easily and reliably perform transactions leveraging the state of another L2. Although solutions involving escrow contracts currently exist, no solutions manage to achieve this functionality seamlessly and without considerable time delays. As such, the proposed shared sequencing solution is significantly more effective and user friendly.
Landscape
Shared sequencing is a relatively new technological phenomena that has garnered significant interest in line with the recent explosion of roll-up Layer-2 solution development, like optimistic and zero-knowledge rollups. Currently, there are 5 noteworthy shared sequencer networks under development/on the market; Espresso Systems, Radius, AltLayer, Astria, and NodeKit. Below is a fundraising comparison between the four:
Having opened their public Doppio Testnet, Espresso Systems is perhaps the most well-known and well-funded shared sequencing company in the cryptosphere. Their product, the Espresso Sequencer, operates within the HotShot consensus protocol. The current GTM of the Espresso Sequencer focuses on integrating with rollups using Ethereum as their layer 1, and it is already integrated with the likes of Injective. However, there’s no fundamental restriction to Ethereum alone; theoretically, HotShot consensus could checkpoint on any smart contract platform and integrate with rollups targeting that platform. This could include non-EVM layer 1s like Solana or even layer 2s like Arbitrum.
Radius is quite unique when it comes to trustless sequencing solutions; they are introducing a zero-knowledge solution known as the Practical Verifiable Delay Encryption (PVDE) for implementation in an encrypted mempool. PVDE is an inherent feature of Radius’ sequencing layer, which is designed as a scaling layer for rollups, the architecture of which is shown below. In this approach, data surrounding submitted rollup transactions in the mempool that are waiting to be sequenced is encrypted, preventing the discovery of harmful maximum extractable value (MEV) opportunities and eliminating the need for trust in sequencers.
AltLayer has gained significant traction in the investor space as a decentralised and modular Rollup-as-a-Service (RaaS) protocol. Their stack offers developers the opportunity to launch rollups with highly customisable stacks, and even offers a no-code option as well as an SDK. Rollups built using AltLayer are incredibly versatile in stack design as shown below, and can currently choose from using AltLayer’s own shared sequencer network, run by sequencer nodes on an interlayer named the Beacon Layer, or from the Espresso or Radius shared sequencing solutions. Any AltLayer community member can apply to operate a sequencer node on the Beacon Layer; it is important to consider that they can abuse their leverage over transaction ordering, but are disincentivised to do so as they would incur significant losses in posted collateral in the event of any misbehaviour.
Astria is a ‘middleware’ shared sequencer network being developed by Settler Labs on Celestia (an L1). Astria is used by the Astria EVM; their flagship rollup and the first EVM layer on Celestia. As a ‘lazy’ sequencer, Astria guarantees atomic inclusion of transactions ordered in a block, but does not execute the state transition function for a given rollup. Astria thus only provides soft commitments for transaction sequencing on Celestia. For finalisation, Astria-sequenced soft-committed blocks are cross-referenced by a rollup-sovereign ‘Conductor’ against Celestia blocks for rollup-specific transactions; if the Conductor receives a Celestia block previously seen from Astria, the rollup is informed that the blocks are finalised. Rollups built using Astria’s ‘one-click’ developer kit can choose whether to finalise transactions using firm Celestia-level confirmations at a block time of 11s, or whether to benefit from fast pre-confirmations using Astria-level ‘soft’ commitments at a block time of 1s.
NodeKit is a newer player in the shared sequencing space, having come out of stealth development in mid-July to raise their ongoing seed round. NodeKit’s SEQ is a custom L1 blockchain with a built-in decentralised shared sequencer. It allows rollups layering on top of the EVM-compatible SEQ L1 to benefit from a decentralised sequencer and increased cross-rollup interoperability, with NodeKit’s SDK making it simple for anyone to launch a rollup on SEQ.
As is evident, there are a variety of different technical architectures employed by startups focused on shared sequencing. While the end experience of all products will be almost indistinguishable, it is likely that integration speed and flexibility in rollup stack architecture will become key catalysts for developer adoption, where AltLayer, Radius, and Astria especially show strength.
Drawbacks
Shared sequencing allows for the simultaneous ordering of transactions across multiple rollups for commitment to a base chain’s block space. This may sound exciting as a new frontier for cross-rollup DeFi composability, however the reality is that the atomic inclusion of transactions across rollups does not guarantee atomic execution [4]. While bundled infallible transactions such as withdrawals or transfers can be atomically executed between rollups, complex fallible transactions common in applications such as DeFi can still fail in execution.
This makes true inter-rollup DeFi composability infeasible, as shared sequencing cannot guarantee transaction execution between multiple rollups if 1 or more fallible transactions are included. In fact, most user-facing transactions like swaps or smart contract interactions contain at least one fallible transaction, removing any hope of cross-rollup DeFi composability and rendering atomic inclusion championed by shared sequencers almost useless (aside from for shared sequencing of infallible interactions like transfers).
Conclusion
Sequencers occupy a role of systemic importance in rollup architecture, but suffer from key composability, MEV, and security concerns. Although shared sequencing brings atomic inclusion of transactions across different rollups, full atomic composability across rollups requires atomic execution.
In light of this, the development of shared sequencers, although a necessary development in blockchain architecture, represents more of an incremental development in the direction of L2 interoperability through enabling data synchronisation and helping in achieving ‘consistent state’ across rollups, as opposed to the transformative advancement it is often portrayed as. Nonetheless, there are some promising startups developing shared sequencers with the aim of bridging liquidity across rollups and reducing the reliance on centralised points of failure.
