Latency and Profits: MEV On Solana
Understanding Solana’s MEV lansdcape
Introduction
When the average user submits their transaction on a blockchain through a decentralized application’s front end, they are no longer responsible for ensuring their transaction is actually processed, executed and delivered. They wait for their transaction status to read “successful” and the desired amount of asset X delivered to address Y, often unaware of the competitive process taking place behind the scenes.
Entities run nodes on blockchains called validators , which are responsible for verifying transactions, processing, and eventually executing them onchain. Validators are compensated for their computational efforts in the form of block rewards and transaction fees, but the process is inherently competitive given that running nodes can be resource-intensive and costly, incentivizing actors to seek out additional revenue streams to generate profits.
MEV
Maximal Extractable Value (originally Miner Extractable Value on Proof-of-Work chains) refers to the measure of profit an entity can make by reordering transactions in a block beyond standard validator rewards. MEV is like an invisible tax which comes at the expense of regular users, yet it drives the economic value of blockchains and cannot be entirely eradicated.
Some real world analogies from the Jito blog : Disney Fastpass, UberEats Priority Delivery, HOV lanes
MEV is chain-agnostic. Different blockchain networks approach MEV in different ways but it exists virtually on all of them. Ethereum has unofficially adopted a de-facto marketplace between block builders and proposers, Osmosis uses threshold encryption to preserve transaction details from validators, and DYDX has implemented social slashing as a way to hold validators accountable and honest.
Today we are jumping into the MEV landscape on Solana, and it helps to have some context on Solana.
Getting To Know Solana
What is Solana?
- Solana is a Proof-of-Stake Layer 1 blockchain optimized for high throughput and performance. Its architecture is distinctly separate from from EVM chains, and features like timestamping, parallel execution, and continuous block production enable it to run high TPS numbers and enable Solana’s vision of being a consumer-first chain. After all, transaction fees cost less than $0.01 by default.
- Solana is a monolithic chain, operating as a single global state machine. This means all blockchain components (consensus, data availability, execution, and settlement) operate on a single layer as opposed to modular scaling designs which designate separate chains for separate tasks; where L2s and sharding are considered the endgame for blockchain scalability. The core monolith chain thesis is that building on sharded systems could become exponentially more difficult for application developers over time, in addition to the continuous fragmentation of liquidity across these networks.
- Solana’s composability makes it easier for app developers who need to access state across different contracts, though it does mean that bottlenecks in reading state are much more impactful to the network’s real time performance. A number of network outages over the past few years have exposed the greatest concern with a monolithic chain — an outage caused by reason X halts activity for the entire chain.
- Like post-merge Ethereum, Solana is secured by Proof of Stake consensus. Solana does not have an inherent stake requirement like 32 ETH on Ethereum — but hardware requirements can make up for those costs alone. The average home internet connection cannot support running a Solana validator — most operators run validators in datacenters which are specialized to handle mass data storage and processing. It becomes advantageous to colocate in datacenters with lots of network stake, as those validators process more transactions and earn more block rewards by design. 50% of all transaction fees collected by the slot leader are burned. This is intended to incentivize leaders to process as many transactions as possible during their allocated slots, a crucial part of what makes Solana so fast.
- Solana currently has 1,990 validators, and 2,919 full nodes. For comparison, Ethereum has 5,730 full nodes running. Solana validators’ superminority is 30 — this means the top 30% of validators control >33% of network stake and could theoretically collude to halt the network. Not real likely however.
- Solana code is written in Rust, not Solidity. This enables fast, low-level code that can take advantage of hardware optimizations — in addition, Rust is more adopted in the wider developer community than Solidity and can be easier to audit. Solana operates within the LLVM compiler infrastructure, which separates human readable code from low-level code and gives developers the best of both worlds. However, exciting developments like Solang — a Solidity compiler enabling developers to write Solana programs- are pushing the boundaries to make building on Solana more accessible and welcoming for everyone.
Solana Transactions
There are several important distinctions to make regarding Solana transactions:
- Solana does not have a public mempool
- Blocks are produced every 400ms
- Consensus votes are counted as transactions
- Block builders on Solana (known as leaders) are known ahead of time following a stake-weighted leader schedule.
The lifecycle of a Solana transaction
- User indicates txn X via dApp front end
- dApp created txn and sends to user for signature
- User signs > back to dApp
- dApp forwards signed tx to RPC server.
- RPC provider scans leader schedule and sends to the current slot leader, and the next two. This is a precautionary step in the event that the current leader is offline / connection lags
- Leader sends signed tx to consensus
- Consensus votes to verify tx
- Transaction is routed back to RPC > dApp front end > user
- Transactions are considered finalized on Solana after 31 blocks have been built on top of the one in question
Solana MEV
Solana’s MEV landscape is still young and immature compared to Ethereum’s. The network sees less volume and liquidity overall, and validators earn less block rewards as a result. Spamming transactions has been the most common strategy for priority transaction inclusion, which has also lead to a number of network outages in the past. A series of network upgrades rolled out earlier this year enabled mechanisms for limiting spam and ultimately punishing actors who deploy bots for this purpose — Solana’s last outage was in February and the network has been consistently running TPS upwards of 4000 since then. Furthermore, this introduces new vectors that could shift Solana’s MEV environment over time. An uptick of activity and growth on the network is bound to create a more competitive environment for MEV extraction — therefore future approaches to Solana MEV infrastructure should be focused around providing fair and beneficial distribution of MEV profits.
Searchers compete to execute profitable transactions which extract the maximal value for them. On Ethereum, blocks are produced in 12 second slots. This latency means searchers mostly compete for transaction inclusion with fees, battling to outbid one another ultimately.
On Solana, block production occurs every 400ms, and given Solana’s FIFO transaction processing mechanism, searchers must compete over latency rather than fees — being able to read state first means being the one to successfully execute a profitable trade.
Searchers can run nodes to read the latest network state — but being first means either directly controlling a meaningful portion of stake or working with high-stake validators, both of which can be costly strategies.
In order to actually extract MEV, searchers deploy bots to continuously monitor the chain for profitable opportunities and execute trades on their behalf. These bots work very similarly to shopping bots, like the ones used to buy specific items as quickly as possible on a user’s behalf.
Similar to Ethereum and other chains, the most common forms of MEV extraction are arbitrage and liquidations.
Arbitrage refers to profit made by capitalizing on price discrepancies for the same trading pair across different trading venues. These discrepancies typically don’t last very long, making it important to act in a timely manner, which low-latency networks like Solana are optimized for.
Within the context of Solana DeFi — this is like buying SOL/USDC for $20.31 on Raydium and selling for $20.24 on Orca. The $0.07 in profit might not seem like much — but using bots to execute many more transactions like this one every day can accumulate significant profits over time. Most MEV profits come from arbitrage transactions — though not all.
Liquidations are another venue for MEV extraction made possible by the innovations of DeFi — namely borrow/lending protocols. Borrowers can borrow up to a certain amount of assets against deposited collateral within a given threshold. If the price of the collateral asset drops and the position reaches its liquidation threshold, the position is then liquidated. Liquidation bots close out the unhealthy position by paying off the debt to receive the position’s collateral at a discount. This is another form of MEV extraction, though not as popular as arbitrage given that liquidations occur when prices suddenly drop in markets and borrowers have the ability to monitor and pay off their position in real time.
NFT mints are also worth mentioning — Solana NFTs had their fair share of glory and good times and saw a number of viral NFT mints which gave the network a fair share of trouble. When an NFT mint is very highly anticipated, operators can deploy bots to spam the Mint function erroneously until they have scooped up as many as possible. This method is different from arbitrage and liquidations though — there is no inherent profit guarantee as the NFT still needs to be sold in the secondary market.
Strategies for Prioritized Transaction Inclusion
Spamming
Given the first-in-first-out transaction processing mechanism on Solana, paired with a lack of any priority fee mechanisms until earlier this year, spamming transactions has been the most common way to fight for transaction inclusion. This method wasn’t so bad for searchers — who had to pay less than a penny per transaction, equivalent to virtually nothing for resourceful entities. Unfortunately for slot leaders — this meant processing a lot of failed transactions — which they don’t receive block rewards for. Spam would overload validators’ processing capabilities and cause discrepancies in consensus — forcing the network to come to a halt temporarily.
Priority Fees
Priority fees are part of the new upgrades rolled out to address and mitigate spam, by creating new incentives for transaction priority. Priority fees help reduce spam by creating a new way to get priority inclusion in a block — a costlier one. This would mean that spamming becomes much costlier and therefore disadvantageous to searchers over time. However, priority fees need to be consistently high enough to actually incur significant costs and damage to harmful spammers.
fee data from Solana Compass shows that priority fees are being continously used in the last 100+ epochs. The main question is now how high are these fees on average — and do they pose actual threats to searchers’ costs and operations. Low ceilings on priority fees would likely mean searchers aren’t affected too significantly. according to Jito’s MEV explorer, validator tips are all fairly low except for the rare exceptions of MEV bots like 2Fast1RAxFtcXipCucpJhFQACiksuySKa8E9S9ckwVYY , who have paid up to 21.78 SOL for priority blockspace at times.
Jito-Solana : Flashbots for Solana
The Jito-Solana client is the pioneer of client diversity on Solana. For most of Solana’s history, validators were obliged to run the Solana Labs client as there simply weren’t any other options available. This poses a centralization risk in the event that the client goes down, so does the entire network.
This changed when the Jito Labs team launched Jito-Solana : an open source fork of the Solana Labs client, with an extra 1000 lines added to optimize for MEV capture and distribution. Similar to the Flashbots mev- Geth client on Ethereum, Jito-Solana runs an out-of-protocol blockspace auction in which searchers submit bundles of transactions they would like to be included to generate profit, along with bids for these bundles, to validators running the client. The highest bidders win and their bundles submitted onchain, creating an additional revenue stream for validators and their stakers. 100% of tips paid are paid out to the validators and their stakers, creating a more fair and competitive environment for MEV extraction on Solana.
Jito’s mission aligns with the idea behind the MEV Supply Chain Utopia posed by the Flashbots research team in an older blog post, which represents a theoretical environment where all the MEV generated by users is returned to users, either in the form of direct cashbacks, gas fee subsidies, or network security subsidies. While true decentralized block building may never become more than a good idea — it represents the shared sentiment in the MEV community that MEV distribution needs to become more decentralized as a blockchain grows.
However, Jito faces a bit of a chicken-egg problem — it needs adoption to gain adoption. Without a high volume of searchers submitting bundles and bids, validators running the client will not be able to generate higher profits and therefore will be less incentivized to run. The good news for Jito is that adoption is in fact growing. Currently, the Jito-Solana client makes up 31% of the validator client market share, according to this dashboard from P2P.org . Validator MEV rewards as well as the number of validators adopting the Jito client (now 161) is also currently trending upwards, indicating that more protocol actors are thinking about how to utilize MEV capture/distribution on Solana.
Data/Numbers 📊
A snapshot of the past year in Solana MEV , taken from Jito’s epic explorer 📸
The past year saw over 1.3b MEV transactions, at a 3.31% success rate, averaging $0.23 per block in MEV profit and over $14.8m in total MEV profit.
MEV transactions by signer also tell us that the top 10% of mev extractors make up 50% of the total mev extracted! this tells us it is fairly centralized and there is room for expanding the gap, though it will not happen without the necessary software and tooling.
Current State and Moving Forward
The general consensus within the Solana community is that MEV is not something which can be actually eliminated — instead the focus should be on how MEV extraction can benefit regular users and applications in the ecosystem, while providing the necessary incentivizes for running a Solana validator.
Centralization Vectors
MEV becomes more centralized with network effects. The main centralization vectors for MEV on Solana are stake and geographic location.
- Stake: If a high stake validator starts to employ strategies to capture MEV at a significant advantage to other validators, they can amass more network influence over time, which poses centralization risks to Solan’s validator economy. Validators who earn more than others will be able to offer competitive returns to stakers, who will be attracted to stake with validators offering higher APY. More stakers with a validator = more network influence within the hands of that one validator.
- Geographic Colocation: Latency-oriented MEV dynamics mean that colocating with high-stake validators is advantageous for searchers on Solana. The closer a searcher is located to a data center with high-stake validators, the faster they read state and the better they can extract MEV.
The Network Upgrades Tackling Spam
Spam has plagued Solana and causes numerous outages. The good news is that Solana has rolled out a series of network upgrades to tackle its spam issue, and with the last outage taking place in February, at a high level it seems the upgrades are working as intended so far.
Given Solana’s FIFO transaction processing mechanism, spamming has been the best chance at earlier block inclusion. Validators need to process transactions even if they have failed, but they don’t earn block rewards for this computational effort, which means money is only coming out of their pockets in the process. According to Jito’s dashboard , MEV transaction success rates for the past year sit under 3% ‼️
This dilutes validator operations even further beyond the hard upper bound 50% limit imposed on earnings through the fee burn mechanism. Poorly incentivized validators make for poor decentralization efforts overall, making validator incentives a core issue to address for the network.
The good news is that Solana (quietly) rolled out a series of network upgrades in the past year aimed at mitigating spam and its externalities, and consequently reshaping MEV dynamics over time.
Fee markets — Since summer ’22, priority fees have enabled users to express their desire for priority inclusion in a fair and competitive manner. More importantly, however, the introduction of local fee markets play an important role in pricing general blockspace — they prevent “hotspots” of activity, like a hype NFT mint or token IDO, from congesting the entire network and raising gas fees in the same way Ethereum gas wars do.
QUIC Protocol — Solana has moved on from using UDP packets to transfer data between RPC servers and slot leaders and has instead adopted Google’s QUIC protocol. In short, QUIC enables validators to rate limit addresses flooding the network with spam. This means it disables one anonymous machine to spam the same transaction a bunch of times, like before.
Stake weighted Quality-of-Service
This feature intends to replace accepting transactions on a first-come-first-serve basis by instead distributing bandwidth based on stake-weight (how much SOL you stake vs other nodes). Considering stake weight is the economic primitive of PoS chains like Solana, it seems only natural to extend the standard to transaction processing quality.
The New Meta: MEV Infra
To wrap things up on an exciting note, here are a handful of new tools/developments in Solana MEV infrastructure.
JitoSOL 👨⚖️
We talked about the benefits of the new Jito-Solana client for validators. Well, validators need stakers to support their operations and partake in profits from securing the network, and stake pools enable this. Jito now has stake pools which delegate user’ stake to validators running the MEV-enabled Jito client, and earn MEV rewards from blockspace auctions on top of the standard block rewards. The jitoSOL liquid staking token (LST) allows users to earn a base yield while being able to deploy their assets for further use across Solana DeFi.
Currently, jitoSOL is offering ~7.15% APY according to their site. NFA.
Marinade Finance mTransaction 🥩
Marinade Finance is one of the true OG protocols on Solana, and actually holds the highest TVL by a landslide ($111.5m — the next closest is Lido with $52.9m…). So when they roll out new innovations they’re well worth paying attention to.
mTransaction is a transaction throughput marketplace for Solana. It provides a venue for prioritized transactions by fast-tracking user transactions to the slot leaders. Built as a ‘sidecar’ to the underlying validator client, mTX can safely integrate with any of the existing (and future) clients without requiring validators to make any changes to their code.
Priority fees are collected from users and paid to participating validators, incentivizing validators with an extra profit venue. That said, mTX has some strict requirements — validators must have at least 200,000 SOL staked (in order to have at least one guaranteed QUIC connection to the leader), and be located in Europe or US-East. Meanwhile, users must fill out a form if interested to reduce overhead.
Nonetheless, it will be interesting to see if mTX grows and develops as the demand for blockspace grows with the rise of network activity. A rising tide lifts all ships as they say…
Chorus One
The research team at Chorus One has been involved in in-depth analyses of MEV on various chains, including Solana. They released a whitepaper for the Solana-mev client earlier in February, proposing a new methodology for fair and efficient MEV extraction and distribution. The core concept of the client is that it enables validators to check every batch of processed transactions for MEV opportunities and append transactions to capture the MEV profit where possible. Similar to Jito-Solana, this client would be a fork of the Solana Labs client with a few optimizations — it would not introduce any network hops or changes to the protocol itself.
This said, I haven’t been able to find much information on the client following up the release of the whitepaper…wat mean?
Firedancer 🔥💃
Though not directly MEV infrastructure, Firedancer is worth mentioning in the context of upcoming Solana infra developments.
Firedancer is a high performance client for Solana being built and developed by the Jump team for the purpose of increasing network throughput, resilience, and efficiency, without increasing hardware requirements. It is one of the most anticipated developments on Solana — even Anatoly has alluded to rebranding to Solana 2.0 following the launch of the high performance client. For starters, client diversity means there is a significantly smaller risk of the network facing a Zero Day , a critical step forward in Solana’s resiliency.
The theoretical performance metrics are also eye-opening. 1 million TPS??
So what does this mean for MEV?
It’s hard to point specifically to how Firedancer impacts MEV dynamics on Solana, but it is fair to assume that Firedancer will onboard a lot more network participants, including regular users and validators alike. A boost in network activity means a boost in transactions which means a boost in MEV opportunities. Assuming Jito continues to see more adoption, we can be optimistic that MEV on Solana becomes more competitive, and therefore more beneficial to the greater whole of the network rather than a small number of sophisticated actors.
The new rise in Solana MEV infrastructure indicates the transformation that has begun to take place and is expected to develop as the Solana network grows as a whole. In addition to the network upgrades mentioned earlier, these tools may just become some of the most important components of Solana’s economy down the road.
More Discussion
Distributed block production has been proposed for Solana previously, which would create a similar mechanism to Proposer-Builder-Separation on Ethereum. DBP on Solana would “split out user transactions into separate user blocks that are transmitted concurrently with blocks of consensus votes. This naturally follows into asynchronous execution of the user blocks.” Anatoly had previously discussed embracing the challenge of having multiple block producers operate concurrently , and separating the block production process from the transaction execution process. These were points raised at Breakpoint 2022, and though there haven’t been too many subsequent discussions around this topic, I would certainly keep an eye out as a number of Solana thought leaders seemed to be quite enthusiastic about the idea.
Umbra Research also left a handful of great open questions in their “MEV on Solana” research paper:
- Can we monitor the degree of entity decentralization and geographic decentralization of Solana validators?
- How can we keep Solana sufficiently decentralized, so that it remains credibly permissionless and censorship-resistant?
- How can application developers reduce value leakage to the L1? Are there incentive-compatible fee mechanisms to return MEV back to protocols?
- In what ways can MEV extraction destabilize the network? How can we counteract those forces?
- Can we appropriately price blockspace with in-protocol mechanisms?
As mentioned earlier, MEV is chain-agnostic. Different networks take different approaches to addressing its externalities, but the principle truth is that there will always be a demand for extracting value capture in transaction order flow on the permissionless, distributed data ledgers we call blockchains.
Stay tuned as we continue to explore MEV across different chains to enrich our perspective on this phenomena in the world of digital assets.
Sources 📚
Jito Client Adoption : https://reports.p2p.org/superset/dashboard/jito_client_adoption/?ref=p2p.org&native_filters_key=ufLPU4yoIJGpXNm1x2NyvvaMtAgJAgbUlY4_pmGttnrQmOx0HAcWiXT6rv6adyN4
