What are Rollups? Rollups are non-custodial sidechains that solve Plasma's data availability problem. Rollup blocks include enough data for anyone to reconstruct & challenge their state. In Plasma, only a subset of validators can challenge blocks sent to the main chain.
Rollups come in two flavours:
- Optimistic rollups - ORs depend on a dispute game run by active validators.
- ZkRollups - ZKRs rely on cryptographic zero-knowledge (ZK) proofs, which are instantly validated by a smart contract.
"Rollup scaling implementations are relatively new entrants to the market. They aim to solve the data availability problem by keeping the transaction data stored on the Ethereum’s blockchain in a bundled/compressed format. The main goal of this transaction grouping is to reduce validation costs. The grouped transactions are not validated directly by the Ethereum consensus mechanism. Instead, a different mechanism is used to prove the validity of these transactions."
- From this deep dive (27-1-2020 done by Dragonfly) into why Plasma died, and how zk-Rollup came out of it:
"But zk-Rollup came with its own set of tradeoffs. Namely, validity proofs are computationally expensive to generate (details here). These zk-SNARKS are produced every block and can take upwards of 10 minutes to generate while costing up to 350,000 gas per verification (post Istanbul). For reference, that’s about 3.5% of an entire block (was 8% pre Istanbul).
Additionally, it is currently not possible to deploy general smart contracts on zk-Rollup sidechains. Proposals are under development for specialized zero-knowledge VMs that would enable this, such as zkVM and ZEXE, but they still require lots of specialized knowledge to interact with them. For the most part, zk-Rollups limit general programmability.
By mid-2019, these new developments had re-energized the Ethereum research community. zk-Rollup seemed to solve many of the problems that had plagued the layer-2 narrative. Companies such as Matter Labs (one of our portfolio companies) and LoopRing began actively developing zk-Rollups, and both have testnet implementations live today. With optimizations, Matter Labs believes that it can achieve upwards of 2,000 TPS on its ZK Sync network.
Additionally, Starkware (also a portfolio company) is building a variation on zk-Rollup they call StarkExchange. StarkExchange uses a STARK to prove the validity of sidechain transactions, but delegates the problem of data hosting off-chain (if the sidechain ever halts, exits are guaranteed through on-chain checkpointing). They are implementing a DEX in partnership with DeversiFi with this design and will be launching on mainnet in the near future."
- Settles transactions as L1 calldata. Assumes every transaction is valid by default as it does not compute the transaction. Uses fraud proofs to dispute transactions, hence the seven-day withdrawal period.
- The following comes from this Twitter thread by Token Terminal (24-6-2020):
What are Optimistic rollups? ORs resemble Plasma sidechains, but tradeoff some of their scalability to run fully general smart contracts (whereas ZKRs are currently better suited for simple payment transactions). ORs can run anything Ethereum can.
"A major project in the Optimistic Rollups (OR) domain is Optimism. Optimism is building the Optimistic Virtual Machine (OVM) which aims to scale arbitrary Ethereum transactions on L2. The team recently ran a demo to scale DeFi protocol in collaboration with Synthetix. The results from the demo were encouraging. Another effort along the lines of OR is IDEX's own scaling solution. The IDEX team ran a demo for the solution in October of last year."
- Rollup diff compression. Using that compression, optimistic rollup Fuel benchmarks 2500 tps for token transfers.
Difference with Plasma
"The main difference between Plasma and Optimistic Rollups (OR’s) is that OR’s can run virtual machines - so-called Optimistic Virtual Machines (OVMs) - which allows for the execution of smart contracts on layer 2."
"Where these concerns arise is from the perceived position of the sequencer in the Optimistic Ethereum network. It is important to note that the sequencer is not necessary to access Optimistic Ethereum - it is there to provide significant UX improvements. Most people also assume a single sequencer means a single point of failure, which would be bad, but the failure case is just the shitty L1 UX we already have. Obviously falling back to L1 blocktimes is not ideal, but there are numerous potential solutions including backup sequencers that can step in if the active sequencer fails or is attacked. All of these concerns are only temporary, as the goal is still to move towards a network of sequencers as quickly as possible."
- Settle on L1 as cryptographic proofs. Utilizes zero-knowledge cryptography (SNARKs or STARKs). The state can only be updated with validity proofs, securing transactions cryptographically.
- The following comes from this Twitter thread by Token Terminal (24-6-2020):
What are ZkRollups? In a ZKR, validators generate ZK proofs of all state transitions. ZK proofs are instantly verified by the Rollup contract on Ethereum. This means that it is extremely difficult to manipulate a block or commit a block with an invalid state."
"zk Rollups, the other Rollup scaling option, aims for larger scalability of Ethereum L2 in terms of L2 transactions per second (TPS). However, the technology is still limited to specific use cases, e.g., payments or exchanges, and is bottlenecked by the complexity of the generation of zk proofs. In this area, there are two competing projects: Loopring and Matterlab’s zkSync. Loopring initially used zk Rollups to implement an L2 DEX and have since expanded into the L2 payments arena, with the launch of Loopring Pay. In contrast, zkSync first focused on L2 payments. The team recently launched the product."
- From Bankless (16-4-2020):
"Optimistic Rollups are a layer 2 sidechain construction which seeks to alleviate the burden on Ethereum’s main chain. The basic idea is that instead of validating all transactions on the rollup sidechain, the mainchain simply publishes them and "optimistically" assumes they're valid unless explicitly challenged.
The core benefit to employing rollups is lower gas costs for users, which translates to more transactions per second (several hundred TPS at least) for the network as a whole. Cheaper transactions also means that certain applications that would otherwise be too gas intensive to even be possible, like privacy solutions that employ complex cryptography, now become feasible. Thus, while rollup itself doesn't inherently offer privacy benefits, it's a fitting substrate on which to build privacy preserving technology. Likewise, rollup doesn't inherently increase transaction latency (speed), but gives a nice environment on which channels, which can offer virtually instant transactions, can be built.
The way Optimistic Rollups handle data gives the constructions some nice simplicity, especially relative to alternative layer 2 protocols. Crucially, this translates to UX for end users that could (and should!) feel nearly identical to that of using layer 1. Likewise, for developers and protocol designers, much of the tools and thought-models they are accustomed to will still be available, though as we'll see, tackling some challenges around composability may require some additional work.
First, a user deposits some funds (“funds” could be Ether, ERC20s, ERC721s, etc.) onto a rollup chain; this first step mirrors the UX of many layer 1 dApps, in which the user must initially transfer funds into a contract before using it. At this point, a user can issue transactions on the application as they usually would; if the rollup chain is designed to prioritize censorship resistance, getting one’s transaction included doesn't depend on any more trust, reputation, or goodwill—and doesn’t invite any more potential censorship or front-running—than publishing an L1 transaction.
When a user wants to bring their assets back to layer 1, they issue a special withdrawal transaction. Here we see a potential difference: recall that Optimistic Rollup’s security model depends on the ability for parties to issue challenges; thus, we need to allow a grace period for any parties monitoring the activity to (potentially) prove fraudulence. This means that once a withdrawal is requested, the user must wait before the funds are accessible again on layer 1; the economic security this mechanism gives is a function of both the amount of stake required to produce rollup blocks and the length of this waiting period (Ed Felten argues here that 3 hours is sufficient).
With that said, the hope/expectation is that users will rarely, if ever, have to actually wait through this grace period. Third parties can offer to purchase ownership of your pending withdrawal by sending you the equivalent value on layer 1, minus some fee, enabling you to access your funds without having to wait. Thus, as long as such liquidity providers are indeed available, and with an interface that abstracts the complexity away, even withdrawals will hit UX parity with layer 1."
- From ethresea.rch (28-7-2020):
"Current scalability solutions move the users into another domain where their transactions are validated. If they want to leave that domain they may have to pay a high cost. There are ways to minimize 1 these costs.
There seems to be some urgency to scale interactions between users and smart contracts given the current gas market fluctuations.
Here we propose a temporary scalability improvement of 3x on layer 1 by putting many users in a smart contract “super” wallet and batching their various uniswap interactions. This solution can realistically be build audited and deployed in a few months. It is composable with smart contracts that allow batched user interactions for example uniswap where matching pool operations are very similar no matter how many users funds are included. Something like DAI where users have a specific smart CDP that they own is not so well suited.
This will work in the short term. In the medium term a zkrollup with similar properties can be built. Which will allow this to scale orders of magnitude more.
Rollups intertwined with the consensus and data availability layer. Slow to update, considered in a "final" state with governance required to upgrade.
- A breakdown in security differences can be read in this post by Vitalik Buterin (1-2-2022). He goes into different zkEVMs here(4-8-2022).
- From this blog by PlotX (25-1-2021):
"There are two types of rollups: (i) ZK Rollups, and (ii) Optimistic Rollups. ZK Rollups, although faster and more efficient than Optimistic Rollups, do not provide an easy way for the existing smart contracts to migrate to layer 2. Optimistic Rollups, on the other hand, run an EVM compatible virtual machine called Optimistic Virtual Machine (OVM) which allows for executing the same smart contracts that can be executed on Ethereum."
"There are 2 schools of thought when understanding layer-2 solution trade-offs:
- Zero-knowledge proof is the superior solution while optimistic rollups has critical flaws.
- The recursive zero-knowledge proof tech required for generalized smart contract support is very difficult and even if achieved, proof sizes would be too big and costly making optimistic rollups the winning layer-2 solution.
- It’s too new and risky: STARKs and PLONKs are recent innovations while SNARKs have been around since 2013
- Generic turing complete smart contracts require recursion, something ZkSync wasn’t expected to be capable of (though Starkware has recently rolled this out)
- Proof sizes will be big and expensive
- Use new programming languages which needs to be transpiling from Solidity code (something not yet available)
What those in the ZKP camp say about optimistic rollups:
- 1 rollup can support a max of $10m -$50m in value before it can be profitably exploited
- Very slow withdrawal times
- Inferior security compared with layer-1 Ethereum
- Rolling back state would be unworkable in practice
Although ZKP solutions might be technically better, Optimism has the first mover advantage combined with simple onboarding and backing.
As the 2 leading DeFi giants launch Optimism in the new year it will be key to watch how much usage and liquidity they attract to the layer-2 instances. If the migration is fast then this would amount to a hard to disrupt network effect. If the migration of leading layer-1 projects to layer-2 does not quite happen, then this may give ZkSync time to roll general smart contract support with transpiling and attract some early adopters.
Starkware’s ZKP based StarkEx platform is ready now. It is using their Turing complete Cairo language which is capable of general smart contract support and has been adopted and is in production use by DeversiFi and dYdX."
"If an attacker can 51% attack the smaller chain, how much damage can they do?
If the smaller chain is a "rollup", they can delay transactions and perhaps even force users to pay L1 fees, but they can't steal anything. This is because there is an on-chain mechanism (whether fraud proofs or data availability proofs) that can actually check that the events on the smaller chain leading up to the attacker's withdrawal are all valid.
This is what we mean by "shared security". If you are hodling assets and doing stuff on the smaller chain, are you as secure as if you were doing it on the larger chain, or are you less secure? On a rollup or a plasma, you are just as secure. On an independent L1 or a sidechain, you are much less secure. On a validium, you're somewhere in the middle."