https://preview.redd.it/b80c05tnb9e51.jpg?width=2550&format=pjpg&auto=webp&s=850282c1a3962466ed44f73886dae1c8872d0f31
Submitted for consideration to The Great Reddit Scaling Bake-Off
Baked by the pastry chefs at Offchain Labs
Please send questions or comments to [[email protected] ](mailto:[email protected])
1. Overview
We're excited to submit Arbitrum Rollup for consideration to The Great Reddit Scaling Bake-Off. Arbitrum Rollup is the only Ethereum scaling solution that supports arbitrary smart contracts without compromising on Ethereum's security or adding points of centralization. For Reddit, this means that Arbitrum can not only scale the minting and transfer of Community Points, but it can foster a creative ecosystem built around Reddit Community Points enabling points to be used in a wide variety of third party applications. That's right -- you can have your cake and eat it too!
Arbitrum Rollup isn't just Ethereum-style. Its Layer 2 transactions are byte-for-byte identical to Ethereum, which means Ethereum users can continue to use their existing addresses and wallets, and Ethereum developers can continue to use their favorite toolchains and development environments out-of-the-box with Arbitrum. Coupling Arbitrum’s tooling-compatibility with its trustless asset interoperability, Reddit not only can scale but can onboard the entire Ethereum community at no cost by giving them the same experience they already know and love (well, certainly know).
To benchmark how Arbitrum can scale Reddit Community Points, we launched the Reddit contracts on an Arbitrum Rollup chain. Since Arbitrum provides full Solidity support, we didn't have to rewrite the Reddit contracts or try to mimic their functionality using an unfamiliar paradigm. Nope, none of that. We launched the Reddit contracts unmodified on Arbitrum Rollup complete with support for minting and distributing points. Like every Arbitrum Rollup chain, the chain included a bridge interface in which users can transfer Community Points or any other asset between the L1 and L2 chains. Arbitrum Rollup chains also support dynamic contract loading, which would allow third-party developers to launch custom ecosystem apps that integrate with Community Points on the very same chain that runs the Reddit contracts.
1.1 Why Ethereum
Perhaps the most exciting benefit of distributing Community Points using a blockchain is the ability to seamlessly port points to other applications and use them in a wide variety of contexts. Applications may include simple transfers such as a restaurant that allows Redditors to spend points on drinks. Or it may include complex smart contracts -- such as placing Community Points as a wager for a multiparty game or as collateral in a financial contract.
The common denominator between all of the fun uses of Reddit points is that it needs a thriving ecosystem of both users and developers, and the Ethereum blockchain is perhaps the only smart contract platform with significant adoption today. While many Layer 1 blockchains boast lower cost or higher throughput than the Ethereum blockchain, more often than not, these attributes mask the reality of little usage, weaker security, or both.
Perhaps another platform with significant usage will rise in the future. But today, Ethereum captures the mindshare of the blockchain community, and for Community Points to provide the most utility, the Ethereum blockchain is the natural choice.
1.2 Why Arbitrum
While Ethereum's ecosystem is unmatched, the reality is that fees are high and capacity is too low to support the scale of Reddit Community Points. Enter Arbitrum. Arbitrum Rollup provides all of the ecosystem benefits of Ethereum, but with orders of magnitude more capacity and at a fraction of the cost of native Ethereum smart contracts. And most of all, we don't change the experience from users. They continue to use the same wallets, addresses, languages, and tools.
Arbitrum Rollup is not the only solution that can scale payments, but it is the only developed solution that can scale both payments and arbitrary smart contracts trustlessly, which means that third party users can build highly scalable add-on apps that can be used without withdrawing money from the Rollup chain. If you believe that Reddit users will want to use their Community Points in smart contracts--and we believe they will--then it makes the most sense to choose a single scaling solution that can support the entire ecosystem, eliminating friction for users.
We view being able to run smart contracts in the same scaling solution as fundamentally critical since if there's significant demand in running smart contracts from Reddit's ecosystem, this would be a load on Ethereum and would itself require a scaling solution. Moreover, having different scaling solutions for the minting/distribution/spending of points and for third party apps would be burdensome for users as they'd have to constantly shuffle their Points back and forth.
2. Arbitrum at a glance
Arbitrum Rollup has a unique value proposition as it offers a combination of features that no other scaling solution achieves. Here we highlight its core attributes.
Decentralized. Arbitrum Rollup is as decentralized as Ethereum. Unlike some other Layer 2 scaling projects, Arbitrum Rollup doesn't have any centralized components or centralized operators who can censor users or delay transactions. Even in non-custodial systems, centralized components provide a risk as the operators are generally incentivized to increase their profit by extracting rent from users often in ways that severely degrade user experience. Even if centralized operators are altruistic, centralized components are subject to hacking, coercion, and potential liability.
Massive Scaling. Arbitrum achieves order of magnitude scaling over Ethereum's L1 smart contracts. Our software currently supports 453 transactions-per-second for basic transactions (at 1616 Ethereum gas per tx). We have a lot of room left to optimize (e.g. aggregating signatures), and over the next several months capacity will increase significantly. As described in detail below, Arbitrum can easily support and surpass Reddit's anticipated initial load, and its capacity will continue to improve as Reddit's capacity needs grow.
Low cost. The cost of running Arbitrum Rollup is quite low compared to L1 Ethereum and other scaling solutions such as those based on zero-knowledge proofs. Layer 2 fees are low, fixed, and predictable and should not be overly burdensome for Reddit to cover. Nobody needs to use special equipment or high-end machines. Arbitrum requires validators, which is a permissionless role that can be run on any reasonable on-line machine. Although anybody can act as a validator, in order to protect against a “tragedy of the commons” and make sure reputable validators are participating, we support a notion of “invited validators” that are compensated for their costs. In general, users pay (low) fees to cover the invited validators’ costs, but we imagine that Reddit may cover this cost for its users. See more on the costs and validator options below.
Ethereum Developer Experience. Not only does Arbitrum support EVM smart contracts, but the developer experience is identical to that of L1 Ethereum contracts and fully compatible with Ethereum tooling. Developers can port existing Solidity apps or write new ones using their favorite and familiar toolchains (e.g. Truffle, Buidler). There are no new languages or coding paradigms to learn.
Ethereum wallet compatibility. Just as in Ethereum, Arbitrum users need only hold keys, but do not have to store any coin history or additional data to protect or access their funds. Since Arbitrum transactions are semantically identical to Ethereum L1 transactions, existing Ethereum users can use their existing Ethereum keys with their existing wallet software such as Metamask.
Token interoperability. Users can easily transfer their ETH, ERC-20 and ERC-721 tokens between Ethereum and the Arbitrum Rollup chain. As we explain in detail below, it is possible to mint tokens in L2 that can subsequently be withdrawn and recognized by the L1 token contract.
Fast finality. Transactions complete with the same finality time as Ethereum L1 (and it's possible to get faster finality guarantees by trading away trust assumptions; see the Arbitrum Rollup whitepaper for details).
Non-custodial. Arbitrum Rollup is a non-custodial scaling solution, so users control their funds/points and neither Reddit nor anyone else can ever access or revoke points held by users.
Censorship Resistant. Since it's completely decentralized, and the Arbitrum protocol guarantees progress trustlessly, Arbitrum Rollup is just as censorship-proof as Ethereum.
Block explorer. The Arbitrum Rollup block explorer allows users to view and analyze transactions on the Rollup chain.
Limitations
Although this is a bake-off, we're not going to sugar coat anything. Arbitrum Rollup, like any Optimistic Rollup protocol, does have one limitation, and that's the delay on withdrawals.
As for the concrete length of the delay, we've done a good deal of internal modeling and have blogged about this as well. Our current modeling suggests a 3-hour delay is sufficient (but as discussed in the linked post there is a tradeoff space between the length of the challenge period and the size of the validators’ deposit).
Note that this doesn't mean that the chain is delayed for three hours. Arbitrum Rollup supports pipelining of execution, which means that validators can keep building new states even while previous ones are “in the pipeline” for confirmation. As the challenge delays expire for each update, a new state will be confirmed (read more about this here).
So activity and progress on the chain are not delayed by the challenge period. The only thing that's delayed is the consummation of withdrawals. Recall though that any single honest validator knows immediately (at the speed of L1 finality) which state updates are correct and can guarantee that they will eventually be confirmed, so once a valid withdrawal has been requested on-chain, every honest party knows that the withdrawal will definitely happen. There's a natural place here for a liquidity market in which a validator (or someone who trusts a validator) can provide withdrawal loans for a small interest fee. This is a no-risk business for them as they know which withdrawals will be confirmed (and can force their confirmation trustlessly no matter what anyone else does) but are just waiting for on-chain finality.
3. The recipe: How Arbitrum Rollup works
For a description of the technical components of Arbitrum Rollup and how they interact to create a highly scalable protocol with a developer experience that is identical to Ethereum, please refer to the following documents:
Arbitrum Rollup Whitepaper
Arbitrum academic paper (describes a previous version of Arbitrum)
4. Developer docs and APIs
For full details about how to set up and interact with an Arbitrum Rollup chain or validator, please refer to our developer docs, which can be found at https://developer.offchainlabs.com/.
Note that the Arbitrum version described on that site is older and will soon be replaced by the version we are entering in Reddit Bake-Off, which is still undergoing internal testing before public release.
5. Who are the validators?
As with any Layer 2 protocol, advancing the protocol correctly requires at least one validator (sometimes called block producers) that is honest and available. A natural question is: who are the validators?
Recall that the validator set for an Arbitrum chain is open and permissionless; anyone can start or stop validating at will. (A useful analogy is to full nodes on an L1 chain.) But we understand that even though anyone can participate, Reddit may want to guarantee that highly reputable nodes are validating their chain. Reddit may choose to validate the chain themselves and/or hire third-party validators.To this end, we have begun building a marketplace for validator-for-hire services so that dapp developers can outsource validation services to reputable nodes with high up-time. We've announced a partnership in which Chainlink nodes will provide Arbitrum validation services, and we expect to announce more partnerships shortly with other blockchain infrastructure providers.
Although there is no requirement that validators are paid, Arbitrum’s economic model tracks validators’ costs (e.g. amount of computation and storage) and can charge small fees on user transactions, using a gas-type system, to cover those costs. Alternatively, a single party such as Reddit can agree to cover the costs of invited validators.
6. Reddit Contract Support
Since Arbitrum contracts and transactions are byte-for-byte compatible with Ethereum, supporting the Reddit contracts is as simple as launching them on an Arbitrum chain.
Minting. Arbitrum Rollup supports hybrid L1/L2 tokens which can be minted in L2 and then withdrawn onto the L1. An L1 contract at address A can make a special call to the EthBridge which deploys a "buddy contract" to the same address A on an Arbitrum chain. Since it's deployed at the same address, users can know that the L2 contract is the authorized "buddy" of the L1 contract on the Arbitrum chain.
For minting, the L1 contract is a standard ERC-20 contract which mints and burns tokens when requested by the L2 contract. It is paired with an ERC-20 contract in L2 which mints tokens based on whatever programmer provided minting facility is desired and burns tokens when they are withdrawn from the rollup chain. Given this base infrastructure, Arbitrum can support any smart contract based method for minting tokens in L2, and indeed we directly support Reddit's signature/claim based minting in L2.
Batch minting. What's better than a mint cookie? A whole batch! In addition to supporting Reddit’s current minting/claiming scheme, we built a second minting design, which we believe outperforms the signature/claim system in many scenarios.
In the current system, Reddit periodically issues signed statements to users, who then take those statements to the blockchain to claim their tokens. An alternative approach would have Reddit directly submit the list of users/amounts to the blockchain and distribute the tokens to the users without the signature/claim process.
To optimize the cost efficiency of this approach, we designed an application-specific compression scheme to minimize the size of the batch distribution list. We analyzed the data from Reddit's previous distributions and found that the data is highly compressible since token amounts are small and repeated, and addresses appear multiple times. Our function groups transactions by size, and replaces previously-seen addresses with a shorter index value. We wrote client code to compress the data, wrote a Solidity decompressing function, and integrated that function into Reddit’s contract running on Arbitrum.
When we ran the compression function on the previous Reddit distribution data, we found that we could compress batched minting data down to to 11.8 bytes per minting event (averaged over a 6-month trace of Reddit’s historical token grants)compared with roughly 174 bytes of on-chain data needed for the signature claim approach to minting (roughly 43 for an RLP-encoded null transaction + 65 for Reddit's signature + 65 for the user's signature + roughly 8 for the number of Points) .
The relative benefit of the two approaches with respect to on-chain call data cost depends on the percentage of users that will actually claim their tokens on chain. With the above figures, batch minting will be cheaper if roughly 5% of users redeem their claims. We stress that our compression scheme is not Arbitrum-specific and would be beneficial in any general-purpose smart contract platform.
8. Benchmarks and costs
In this section, we give the full costs of operating the Reddit contracts on an Arbitrum Rollup chain including the L1 gas costs for the Rollup chain, the costs of computation and storage for the L2 validators as well as the capital lockup requirements for staking.
Arbitrum Rollup is still on testnet, so we did not run mainnet benchmarks. Instead, we measured the L1 gas cost and L2 workload for Reddit operations on Arbitrum and calculated the total cost assuming current Ethereum gas prices. As noted below in detail, our measurements do not assume that Arbitrum is consuming the entire capacity of Ethereum. We will present the details of our model now, but for full transparency you can also play around with it yourself and adjust the parameters, by copying the spreadsheet found here.
Our cost model is based on measurements of Reddit’s contracts, running unmodified (except for the addition of a batch minting function) on Arbitrum Rollup on top of Ethereum.
On the distribution of transactions and frequency of assertions. Reddit's instructions specify the following minimum parameters that submissions should support:
Over a 5 day period, your scaling PoC should be able to handle:

• 100,000 point claims (minting & distributing points)
• 25,000 subscriptions
• 75,000 one-off points burning
• 100,000 transfers

We provide the full costs of operating an Arbitrum Rollup chain with this usage under the assumption that tokens are minted or granted to users in batches, but other transactions are uniformly distributed over the 5 day period. Unlike some other submissions, we do not make unrealistic assumptions that all operations can be submitted in enormous batches. We assume that batch minting is done in batches that use only a few percent on an L1 block’s gas, and that other operations come in evenly over time and are submitted in batches, with one batch every five minutes to keep latency reasonable. (Users are probably already waiting for L1 finality, which takes at least that long to achieve.)
We note that assuming that there are only 300,000 transactions that arrive uniformly over the 5 day period will make our benchmark numbers lower, but we believe that this will reflect the true cost of running the system. To see why, say that batches are submitted every five minutes (20 L1 blocks) and there's a fixed overhead of c bytes of calldata per batch, the cost of which will get amortized over all transactions executed in that batch. Assume that each individual transaction adds a marginal cost of t. Lastly assume the capacity of the scaling system is high enough that it can support all of Reddit's 300,000 transactions within a single 20-block batch (i.e. that there is more than c + 300,000*t byes of calldata available in 20 blocks).
Consider what happens if c, the per-batch overhead, is large (which it is in some systems, but not in Arbitrum). In the scenario that transactions actually arrive at the system's capacity and each batch is full, then c gets amortized over 300,000 transactions. But if we assume that the system is not running at capacity--and only receives 300,000 transactions arriving uniformly over 5 days-- then each 20-block assertion will contain about 200 transactions, and thus each transaction will pay a nontrivial cost due to c.
We are aware that other proposals presented scaling numbers assuming that 300,000 transactions arrived at maximum capacity and was executed in a single mega-transaction, but according to our estimates, for at least one such report, this led to a reported gas price that was 2-3 orders of magnitude lower than it would have been assuming uniform arrival. We make more realistic batching assumptions, and we believe Arbitrum compares well when batch sizes are realistic.
Our model. Our cost model includes several sources of cost:

• L1 gas costs: This is the cost of posting transactions as calldata on the L1 chain, as well as the overhead associated with each batch of transactions, and the L1 cost of settling transactions in the Arbitrum protocol.
• Validator’s staking costs: In normal operation, one validator will need to be staked. The stake is assumed to be 0.2% of the total value of the chain (which is assumed to be $1 per user who is eligible to claim points). The cost of staking is the interest that could be earned on the money if it were not staked.
• Validator computation and storage: Every validator must do computation to track the chain’s processing of transactions, and must maintain storage to keep track of the contracts’ EVM storage. The cost of computation and storage are estimated based on measurements, with the dollar cost of resources based on Amazon Web Services pricing.

It’s clear from our modeling that the predominant cost is for L1 calldata. This will probably be true for any plausible rollup-based system.
Our model also shows that Arbitrum can scale to workloads much larger than Reddit’s nominal workload, without exhausting L1 or L2 resources. The scaling bottleneck will ultimately be calldata on the L1 chain. We believe that cost could be reduced substantially if necessary by clever encoding of data. (In our design any compression / decompression of L2 transaction calldata would be done by client software and L2 programs, never by an L1 contract.)
9. Status of Arbitrum Rollup
Arbitrum Rollup is live on Ethereum testnet. All of the code written to date including everything included in the Reddit demo is open source and permissively licensed under the Apache V2 license. The first testnet version of Arbitrum Rollup was released on testnet in February. Our current internal version, which we used to benchmark the Reddit contracts, will be released soon and will be a major upgrade.
Both the Arbitrum design as well as the implementation are heavily audited by independent third parties. The Arbitrum academic paper was published at USENIX Security, a top-tier peer-reviewed academic venue. For the Arbitrum software, we have engaged Trail of Bits for a security audit, which is currently ongoing, and we are committed to have a clean report before launching on Ethereum mainnet.
10. Reddit Universe Arbitrum Rollup Chain
The benchmarks described in this document were all measured using the latest internal build of our software. When we release the new software upgrade publicly we will launch a Reddit Universe Arbitrum Rollup chain as a public demo, which will contain the Reddit contracts as well as a Uniswap instance and a Connext Hub, demonstrating how Community Points can be integrated into third party apps. We will also allow members of the public to dynamically launch ecosystem contracts. We at Offchain Labs will cover the validating costs for the Reddit Universe public demo.
If the folks at Reddit would like to evaluate our software prior to our public demo, please email us at [email protected] and we'd be more than happy to provide early access.
11. Even more scaling: Arbitrum Sidechains
Rollups are an excellent approach to scaling, and we are excited about Arbitrum Rollup which far surpasses Reddit's scaling needs. But looking forward to Reddit's eventual goal of supporting hundreds of millions of users, there will likely come a time when Reddit needs more scaling than any Rollup protocol can provide.
While Rollups greatly reduce costs, they don't break the linear barrier. That is, all transactions have an on-chain footprint (because all calldata must be posted on-chain), albeit a far smaller one than on native Ethereum, and the L1 limitations end up being the bottleneck for capacity and cost. Since Ethereum has limited capacity, this linear use of on-chain resources means that costs will eventually increase superlinearly with traffic.
The good news is that we at Offchain Labs have a solution in our roadmap that can satisfy this extreme-scaling setting as well: Arbitrum AnyTrust Sidechains. Arbitrum Sidechains are similar to Arbitrum Rollup, but deviate in that they name a permissioned set of validators. When a chain’s validators agree off-chain, they can greatly reduce the on-chain footprint of the protocol and require almost no data to be put on-chain. When validators can't reach unanimous agreement off-chain, the protocol reverts to Arbitrum Rollup. Technically, Arbitrum Sidechains can be viewed as a hybrid between state channels and Rollup, switching back and forth as necessary, and combining the performance and cost that state channels can achieve in the optimistic case, with the robustness of Rollup in other cases. The core technical challenge is how to switch seamlessly between modes and how to guarantee that security is maintained throughout.
Arbitrum Sidechains break through this linear barrier, while still maintaining a high level of security and decentralization. Arbitrum Sidechains provide the AnyTrust guarantee, which says that as long as any one validator is honest and available (even if you don't know which one will be), the L2 chain is guaranteed to execute correctly according to its code and guaranteed to make progress. Unlike in a state channel, offchain progress does not require unanimous consent, and liveness is preserved as long as there is a single honest validator.
Note that the trust model for Arbitrum Sidechains is much stronger than for typical BFT-style chains which introduce a consensus "voting" protocols among a small permissioned group of validators. BFT-based protocols require a supermajority (more than 2/3) of validators to agree. In Arbitrum Sidechains, by contrast, all you need is a single honest validator to achieve guaranteed correctness and progress. Notice that in Arbitrum adding validators strictly increases security since the AnyTrust guarantee provides correctness as long as any one validator is honest and available. By contrast, in BFT-style protocols, adding nodes can be dangerous as a coalition of dishonest nodes can break the protocol.
Like Arbitrum Rollup, the developer and user experiences for Arbitrum Sidechains will be identical to that of Ethereum. Reddit would be able to choose a large and diverse set of validators, and all that they would need to guarantee to break through the scaling barrier is that a single one of them will remain honest.
We hope to have Arbitrum Sidechains in production in early 2021, and thus when Reddit reaches the scale that surpasses the capacity of Rollups, Arbitrum Sidechains will be waiting and ready to help.
While the idea to switch between channels and Rollup to get the best of both worlds is conceptually simple, getting the details right and making sure that the switch does not introduce any attack vectors is highly non-trivial and has been the subject of years of our research (indeed, we were working on this design for years before the term Rollup was even coined).
12. How Arbitrum compares
We include a comparison to several other categories as well as specific projects when appropriate. and explain why we believe that Arbitrum is best suited for Reddit's purposes. We focus our attention on other Ethereum projects.
Payment only Rollups. Compared to Arbitrum Rollup, ZK-Rollups and other Rollups that only support token transfers have several disadvantages:

• As outlined throughout the proposal, we believe that the entire draw of Ethereum is in its rich smart contracts support which is simply not achievable with today's zero-knowledge proof technology. Indeed, scaling with a ZK-Rollup will add friction to the deployment of smart contracts that interact with Community Points as users will have to withdraw their coins from the ZK-Rollup and transfer them to a smart contract system (like Arbitrum). The community will be best served if Reddit builds on a platform that has built-in, frictionless smart-contract support.
• All other Rollup protocols of which we are aware employ a centralized operator. While it's true that users retain custody of their coins, the centralized operator can often profit from censoring, reordering, or delaying transactions. A common misconception is that since they're non-custodial protocols, a centralized sequencer does not pose a risk but this is incorrect as the sequencer can wreak havoc or shake down users for side payments without directly stealing funds.
• Sidechain type protocols can eliminate some of these issues, but they are not trustless. Instead, they require trust in some quorum of a committee, often requiring two-third of the committee to be honest, compared to rollup protocols like Arbitrum that require only a single honest party. In addition, not all sidechain type protocols have committees that are diverse, or even non-centralized, in practice.
• Plasma-style protocols have a centralized operator and do not support general smart contracts.

13. Concluding Remarks
While it's ultimately up to the judges’ palate, we believe that Arbitrum Rollup is the bakeoff choice that Reddit kneads. We far surpass Reddit's specified workload requirement at present, have much room to optimize Arbitrum Rollup in the near term, and have a clear path to get Reddit to hundreds of millions of users. Furthermore, we are the only project that gives developers and users the identical interface as the Ethereum blockchain and is fully interoperable and tooling-compatible, and we do this all without any new trust assumptions or centralized components.
But no matter how the cookie crumbles, we're glad to have participated in this bake-off and we thank you for your consideration.
About Offchain Labs
Offchain Labs, Inc. is a venture-funded New York company that spun out of Princeton University research, and is building the Arbitrum platform to usher in the next generation of scalable, interoperable, and compatible smart contracts. Offchain Labs is backed by Pantera Capital, Compound VC, Coinbase Ventures, and others.
Leadership Team
Ed Felten
Ed Felten is Co-founder and Chief Scientist at Offchain Labs. He is on leave from Princeton University, where he is the Robert E. Kahn Professor of Computer Science and Public Affairs. From 2015 to 2017 he served at the White House as Deputy United States Chief Technology Officer and senior advisor to the President. He is an ACM Fellow and member of the National Academy of Engineering. Outside of work, he is an avid runner, cook, and L.A. Dodgers fan.
Steven Goldfeder
Steven Goldfeder is Co-founder and Chief Executive Officer at Offchain Labs. He holds a PhD from Princeton University, where he worked at the intersection of cryptography and cryptocurrencies including threshold cryptography, zero-knowledge proof systems, and post-quantum signatures. He is a co-author of Bitcoin and Cryptocurrency Technologies, the leading textbook on cryptocurrencies, and he has previously worked at Google and Microsoft Research, where he co-invented the Picnic signature algorithm. When not working, you can find Steven spending time with his family, taking a nature walk, or twisting balloons.
Harry Kalodner
Harry Kalodner is Co-founder and Chief Technology Officer at Offchain Labs where he leads the engineering team. Before the company he attended Princeton as a Ph.D candidate where his research explored economics, anonymity, and incentive compatibility of cryptocurrencies, and he also has worked at Apple. When not up at 3:00am writing code, Harry occasionally sleeps.