I read critics that Arbitrum is a centralised sequencer. AFAIU, its possible to deploy a SORUs with a single rollup node. Does not that also make the latter a centralised sequencer?

Question from Greeneye44

3 Answers 3


The theory is that it's cheap and easy to spin up a rollup node and check proofs. So if there is any discrepancy, an honest party WILL pop up and refute.

I tend to subscribe to that view, it's a 14 day commitment period with a lot of money at stake. I think the incentives are aligned well enough to call this a decentralized sequencing system.

answer from whalesniper


The security assumption for rollups is that there is at least one honest party staked as a rollup operator and running a rollup node. If all rollup operators are compromised, then an attacker can do whatever they want, and steal all funds from the rollup. Of course, that's much easier to do if there's only one operator.

Note, this is about the result of computing operations on the rollup. The idea of optimistic rollups is that users must trust that there is someone honestly computing the resulting state and its hash, and defending that hash on the L1. Sequencers are about something else - namely, the order of operations. Thus, a centralized sequencer can't directly steal funds. If all operations must pass through the sequencer, then the sequencer censor everyone, holding all funds ransom until some sizable portion is payed to them; however, Arbitrum quite sensibly has a censorship-resistant delayed inbox mechanism that accepts operations from the L1. I think a lot of the criticism of Arbitrum on e.g. Twitter is confused on this point.

It seems to me there are three, closely-related drawbacks to a centralized sequencer:

  • single point of failure. If it goes down, everyone's stuck until it comes back online. They could bail via the L1 delayed inbox, but that would be slow and wicked expensive if everyone did it at once. However, in reality, such a critical system will be carefully protected from attacks and outages, and I think downtime will be rare.

  • Sequencer can censor transactions if it doesn't like you. However, if someone had convincing evidence that this was happening, there would be off-chain consequences imposed on the sequencer.

  • Sequencer has full MEV control. When a rollup gets big enough, this is probably a big deal, and probably the main reason we'd want to decentralize the sequencer somehow.

It's possible to write your Tezos kernel to respect a sequencer (centralized or decentralized). I'm building sequencer infrastructure for Tezos rollups in collaboration with the other core teams. This will be fully optional, and provide lower latency for applications where the security tradeoffs make sense.

answer from d4hines


A centralized sequencer gives the full control of inputs to the operator running the sequencer. With smart rollups, a DApp can have a censorship resistant way of retrieving its input operations: it suffices to consider messages posted in the so-called shared inbox maintained by the Layer 1. In that case the Tezos Layer 1 is the sequencer, and the resulting Smart Rollup is as decentralized and resistant to censorship as Layer 1.

More generally, the DApp developer has the freedom to implement a spectrum of solutions, from fully decentralized to fully centralized, which decide what is the sequence of inputs considered by the rollup. I suppose that’s in contrast with the current solutions offered by Arbitrum and Optimism, where a single, centralized sequencer node orders incoming operations.

answer from Nomadic Labs

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.