First, we analyzed the essence of MEV attacks and the current state of the mainstream public chain DeFi ecosystem, as this directly affects the activity of on-chain MEV. Afterwards, we took Ethereum, Solana, Aptos, and Sui, four mainstream and relatively active public chains, as the main analysis targets to analyze the relationship between the MEV public chain architecture and the development of MEV. We found that the MEV system is actually closely related to the architecture and transaction ordering, which also directly affects the user experience.
Under the transaction ordering model based on Gas Fees, such as Sui and Ethereum, they face a vicious cycle of higher Gas Fees during active on-chain transactions (such as MEME transaction activity, significant price changes, NFT Mints, etc.), which leads to some users being unable to participate in the market at that time. Sui does not have the EIP-1559 smoothing mechanism, so the fluctuation of Sui's Gas Fees is more severe, while Ethereum's growth under this protocol is relatively smooth.
Under the deterministic function ordering model like Aptos, MEV is concentrated at the end, as the Leader nodes only have a complete view of the block after the ordering is completed, which makes the MEV on Aptos more complex and with fewer front-running attacks, resulting in more stable Gas Fees.
Solana, similar to Aptos, uses an FCFS deterministic model for ordering, so Searchers prioritize speed, which allows Searchers with better hardware to earn more profits. However, the reliance on speed will flood the entire network with a large number of bot-sent identical transactions to maximize the inclusion of their transactions in the block, leading to network crashes. Jito Labs introduced a pseudo-Mempool similar to Ethereum, supporting priority fees to get ahead of transactions, which also brings high volatility in Gas Fees and a large amount of spam transactions.
We can see that different architectures and transaction ordering models naturally give rise to corresponding MEV market conditions. These can be predicted based on transactions and architecture. For Ethereum itself, EIP-1559 is a patch to address the Gas Fees ordering priority mechanism (involving value redistribution and smoothing the Gas growth curve), but it still cannot solve the Sandwich attack and the poor user experience of high Gas. Therefore, the current solution for MEV is mainly to build a transparent and open market, but the poor user experience caused by Sandwich attacks still needs to be addressed. Meanwhile, the problems caused by architecture and derived protocols also need to be carefully considered, as Ethereum and Solana represent two different architectures with different issues that need to be addressed on a case-by-case basis.
At the same time, we also realize that apart from Ethereum, most public chains have relatively shallow research on MEV, mainly due to the lack of corresponding communities conducting extensive research and on-chain data support. There is still relatively broad research space here, especially for research on emerging public chains like DAG and new ordering models at the architectural level.
On-Chain DeFi Development
DEX Volume by chains
The main types of MEV are Sandwich, Arbitrage, and Liquidation, all of which are essentially related to DeFi, especially the first two, which are related to DEXes. Therefore, the larger the on-chain DEX trading volume, the more profit and opportunities can be earned, and the more intense the competition. In the DEX Volume of the past month, Ethereum and Solana far surpassed other public chains, which also means that these two public chains have the most active MEV.
Meanwhile, as Ethereum is taking L2 as its main scaling target, emerging Layer 2s like Arbitrum and Base also have relatively high trading volumes, but the MEV in this part is often earned by Sequencers. As Sequencers become more decentralized, the MEV issues and potential opportunities in Layer 2 will also gradually emerge.
Apart from Ethereum and Solana, excluding Layer 2, the Layer 1 with large trading volume is BSC, with a DEX trading volume of about $17.3 billion in the past month. BSC is a Fork of Ethereum, so its architecture is similar, and we will not go into further details. Next, we will introduce the architecture and MEV development of Ethereum, Solana, and the emerging DAG public chains like Sui and Aptos.
Ethereum
Ethereum MEV Architecture
In Ethereum, Flashbots introduced MEV-boost to transparentize the entire MEV process, and this architecture is called PBS (Proposer Builder Separate). We briefly introduce the entire PBS sealed first-price auction model. When users submit transactions through an RPC proxy, the RPC is equivalent to running a node and submitting the transactions to the public Mempool. Multiple Builders find the most suitable transactions to order and generate a profit-maximizing block (profit maximization refers to the maximum transaction fee Base+Priority+MEV), then multiple Builders interact with the Proposer through a Relayer, which is a bridge between multiple Builders and the Proposer. Builders submit bids to the Relayer, and the Relayer submits multiple block headers and corresponding bids to the Proposer, who generally accepts the highest bid. The Relayer will implement the MEVBboost specification, which is a technical specification proposed by Flashbot on how to standardize the competitive bidding between Builders and Proposers. In this process, all information is sealed, and the Relayer only submits block headers to the Proposer, so the Proposer has censorship resistance.
However, under the current PBS architecture, we have seen that since the introduction of the MEV-BOOST specification, this profit-maximizing sealed bidding auction mechanism has led Builders and Searchers to gradually move towards cooperation and trust. Whether it is Searchers or Builders, their interests are bound together after the centralization trend is also very obvious. Under POS, the centralization of Validators will also occur, and the entire MEV industry chain becomes highly centralized at various stages, and also introduces the problem of multi-party trust, with Searchers trusting Builders, and Builders and Proposers trusting Relayers. The centralization and trustification of MEV development is clearly at odds with Ethereum's ultimate vision of decentralization and de-trust.
MEV statistics after the ETH Merge, source: libMEV
In Ethereum, since the Merge, a total of $570 million in value has been extracted from the chain, of which Searchers have obtained 15.2% of the value, and the remaining 84.8% has returned to the ecosystem, with the vast majority going to Validators, i.e. POS stakers, and the rest to all token holders.
Categorized by type, source: libMEV
From the above figure, we can see that Sandwich, as the negative effect of MEV, accounts for about 66% of the overall transaction volume, which is the main on-chain activity that has the greatest impact on user UX. Searchers often have a higher profit margin on Arbitrage, around 18.4%, and this is a relatively beneficial MEV. We believe this phenomenon of high profit margins is due to the high volatility of cryptocurrencies.
The average MEV leakage of 15.2% is not unacceptable, but the main impact of MEV is on the user experience, especially under the EIP-1559 mechanism. In a highly volatile market, on-chain bots will be more active in seeking their own arbitrage opportunities, and Ethereum is ordered by Gas Fees, so everyone is competing for transaction fees, which will drive up the user's Gas cost. The EIP-1559 mechanism has not been particularly successful in suppressing the growth rate of Gas, leading to a sharp spike in Gas Fees in the short term.
MEV transaction profit distribution, source: Eigenphi
On Ethereum, the vast majority of the profits from each MEV transaction are concentrated around 0.9 u.
Solana
The architecture of Solana, source: Umbra Research
In Solana's block production mechanism, since the RPC interacts directly with the Leader and adopts the FCFS principle, it does not have a Mempool like Ethereum.
MEV-compatible architecture under the Jito client, source: Helius
The Jito Labs client currently occupies 50% of the client market share, so Jito Labs has built its own pseudo-Mempool, where users stay in the pseudo-Mempool for about 200 ms through the RPC. Jito Labs provides an off-chain bundle guarantee, ensuring that all transactions in the bundle are included in the block. Searchers can bid on the opportunity to sandwich pending transactions, and Searchers maximize their profits by bidding on the highest-priced Bundle, then the Block Engine finds the highest-bid Bundle and submits it to the Leader running the Jito Labs client.
This is the root cause of MEV, but MEV has its positive externalities and demand, and if Jito Labs does not do the pseudo-Mempool, other projects will also do it, so Jito Labs chooses to take over this market to make the entire MEV process more transparent and fair, reducing negative externalities. Of course, this demand for MEV bots makes users the weakest, as validators will charge fees, MEV bots will get arbitrage profits, but users will suffer higher slippage and potentially failed transactions.
Solana's fundamental design is FCFS, so there will not be a drastic increase in Gas Fees during peak network activity. After the Jito Labs client introduced the pseudo-Mempool, MEV bots often rush to grab transactions through speed, and due to the relatively fixed and low fees, MEV bots often recklessly send a large number of identical transactions, which indirectly leads to DDOS attacks, and because the client chooses the QUIC protocol to send transactions for speed, the client often maintains transaction channels between multiple users, and when the client cannot handle a large number of transactions, the client will disconnect some connections on its own, and users often lack economic benefits, the client will collude with MEV to disconnect users, which also leads to the inability of users to send transactions during peak times, which is a potential censorship attack.
Searcher profit distribution, source: Four Pillars
In this architecture, speed is the top priority. This will also lead to the concentration of Searcher profits, as Searchers with better hardware can earn more profits.
Proportion of Priority Fees and Jito Tips, source: Jito
After Jito Labs introduced the pseudo-Mempool, it also supports Priority Fees, which allows block builders to bid on Bundles. In the client built by Jito Labs' pseudo-Mempool, Priority Fees will be burned, which is about half of the Gas Fees.
Proportion of Jito client's SOL staking: source: Jito
Currently, MEV on Solana is still relatively simple compared to Ethereum, where Gas Price is the main sorting method, while Solana uses the FCFS algorithm. Due to the low cost, Searchers naturally tend to send a large number of identical transactions to the network to ensure they are successfully included, while speed is not the top priority on Ethereum, which can also lead to large fluctuations in Gas Fees. Both of these solutions will lead to poor user experience during peak periods, one being the inability to send transactions and the other bearing high fees.
It is worth noting that due to the existence of the pseudo-Mempool, the sandwich attacks brought by MEV have led to a very poor user experience, so the foundation and Jito have cooperated to shut down the Mempool, and have also audited these validators, forcibly excluding the validators participating in the sandwich attacks from the validator network.
DAG
Aptos and Sui are similar, both based on DAG. In this article, we aim to provide readers with a basic introduction to the Narwhal Mempool and analyze how the ordering of transactions within the DAG affects MEV.
Narwhal Mempool, source: Rohan Shrothrium
In the DAG algorithm, users submit transactions to Workers, and these Workers can process unrelated transactions in parallel, packing transactions into Batches and broadcasting them to other nodes. A primary node, Primary, manages multiple Worker nodes and is responsible for collecting the summaries of these Batches and the proof certificates to form blocks and broadcast them. Eventually, these blocks will form a vertex of the DAG, as shown in the steps above, and the transactions within these blocks have not yet been ordered and executed.
Subsequently, an algorithm like Sui's Bullshark sorting algorithm will be used, where a Leader is selected, the Leader's receipt of sufficient votes is verified, and then the Leader uses a deterministic function to perform global transaction ordering based on factors such as gas fee and randomness. After the ordering, this block will be broadcast to all validators, who will verify the block's content and ordering, and then all validators need to execute the transactions in the block.
Under the DAG + Bullshark algorithm combination, a Leader is selected each round to perform global ordering, and MEV leakage is often related to transaction ordering, as seeing the order of transactions allows the insertion of transactions to extract MEV. Within a block, if the Leader node's transactions are placed in the first half, we call it the top of the block, which can allow it to process transactions first; if the transactions are placed in the second half, we call it the bottom of the block, which can give a complete view of the block to build more complex MEV.
Sui
In Sui, the transaction ordering rule is based on the most common Gas Fees. Therefore, in this environment, it is generally some arbitrage operations, such as CEX-DEX arbitrage, when it finds arbitrage opportunities in the transactions, it only needs to initiate a copy transaction with a higher Gas Fee, so these are deterministic in the ordering.
Aptos
Aptos adopts other ordering strategies, so the Leader also needs to re-order all transactions according to the strategy before determining all transactions, and then have a complete view of the block. At this point, their transactions can only be placed at the end of the block. This also makes MEV on Aptos more complex, as these MEV are often not about rushing transactions, but complex MEV under a complete view.
From a user experience perspective, users on the Sui chain often face higher Gas Fees, as MEV strategies are all based on high Gas, which will lead to Gas competition. Aptos, on the other hand, is praised for its better user experience, as it does not order by Gas Fees, so the complexity and cost of the Leader node's MEV attacks are often higher. But the user experience is better.
In models like Sui and Ethereum that order transactions by Gas Fees, they all face a vicious cycle of higher Gas Fees during active chain transactions (such as Meme trading activity, large price fluctuations, NFT Mints, etc.), which can make some users unable to participate in the market at that time. Aptos' strategy of ordering first and then displaying allows user MEV to be minimized during the transaction process, and increases the cost and complexity of MEV strategies.
Aptos transaction fees, source: The block
SUI GAS FEES, source: Sui Explorer
Based on historical data, the on-chain average Gas Fees of Aptos and Sui are at the same order of magnitude, 0.0 0x. However, the chart also shows that Sui's Gas Fee is more volatile, while Aptos is relatively smoother. The reason for this user experience is inseparable from the MEV brought by its sorting algorithm.
References
https://www.umbraresearch.xyz/writings/mev-on-solana
Disclaimer:
The above content is for reference only and should not be considered as any advice. Before making any investment, please seek professional advice.
About Gate Ventures
Gate Ventures is the venture capital arm of Gate.io, focusing on investing in decentralized infrastructure, ecosystems, and applications that will reshape the world in the Web 3.0 era. Gate Ventures collaborates with global industry leaders to empower teams and startups with innovative mindsets and capabilities, redefining the interaction patterns of society and finance.
Website: https://ventures.gate.io/Twitter: https://x.com/gate_venturesMedium: https://medium.com/gate_ventures
