Understanding the Ethereum Network Through Periodic Oscillations in Transaction Fees
Editor's Note: This article comes fromEthereum enthusiasts (ID: ethfans)Editor's Note: This article comes from
Ethereum enthusiasts (ID: ethfans)
, Author: Nic Carter, translation & proofreading: Elisa & Min Min, reproduced by Odaily with authorization.
In 1788, James Watt invented the centrifugal governor, a small device that made the industrial use of the steam engine possible. It applies the rotational force generated by the steam engine to the two heavy spherical cone pendulums. As the cone pendulums are rotated under force, the centrifugal force makes them rise up and pushes the lever connected to the valve. When the pendulum rotates faster, the valve closes. That's it, the governor is driven by the steam engine and mechanically adjusts the flow of steam, thereby adjusting the speed of the steam engine If the speed is large, the pendulum cone will turn off the energy input of the steam engine; if the speed is low, the input will be increased, and if the speed is high, the input will be turned off, such negative feedback will stabilize the rotation of the pendulum cone at a certain speed).
first level title
Interesting Fee Cycle
After the 2017 craze subsided, I noticed that Bitcoin’s transaction fees and transaction volume movements seemed to follow a cyclical pattern. As blocks are filled, transaction fees skyrocket, transaction volume drops, and then blocks are filled again. If I'm not mistaken, this cycle repeated 6 times in 2017.
image description
- Graph visible in Coin Metrics -
Note that I have smoothed the average fees and volume with a 7-day moving average. The Bitcoin "transaction fee crisis" of late 2017 is often thought of as an isolated event, however in fact, Bitcoin transaction fees experienced at least 4 spikes, and 6 if you count the periods with lower peaks. It’s just that most people calculate transaction fees in USD, not Bitcoin, so they only notice the spike in transaction fees that has been going on for a while when Bitcoin’s USD price spikes.
On average, transaction fees peak about two weeks after transaction volume peaks. The entire cycle takes two months to complete, but from the perspective of the whole year, this cycle is gradually shortening. As the utilization rate of block space increases, once there is a small peak in the transaction volume, the transaction fee will be pushed up to the point where traders are speechless. Of course, the transaction fee is only a sign, and the deep-seated problem lies in the block capacity.
By exploring the relationship between block fullness and average transaction fees, we can see how transaction congestion occurs.
Don't forget, SegWit (Segregated Witness) was officially launched on July 23, 2017. The block size exceeded the 1MB limit in the second half of 2017 due to SegWit increasing the block size (equivalent to allowing each block to hold up to 4MB of data). However, not all clients are using SegWit, so the average block size has not increased much.
Here's a flowchart that describes how transaction fees and block space usage oscillate, like a sine wave. We can also think of it as a negative feedback loop, because when transaction fees reach a certain threshold, transaction volume is suppressed.
first level title
Ethereum's "transaction fee crisis" in 2020
This year, when Ethereum transaction fees began to rise and surpass Bitcoin, I began to wonder whether Ethereum would repeat the transaction fee and transaction volume cycle of Bitcoin that year, and whether this cycle would bring the same or even more serious impact on Ethereum. impact, because a lot of liquidity is on-chain (relatively speaking, the trading volume of centralized exchanges is basically "off-chain"). My guess is that we will see the same cycle, but the oscillations may be less violent, because the block space of Ethereum is not fixed and can expand with the increase of usage. But it turned out that the transaction fee pump was more damaging than I expected.
image description
- Source: Coin Metrics -
Like Bitcoin in 2017, Ethereum saw a spike in block utilization in 2020 as transaction volume grew, gradually pushing up transaction fees. In mid-June, the launch of Compound’s governance tokens exacerbated this situation, putting upward pressure on transaction fees. In mid-August, various tokens launched, pushing transaction fees to new highs. A couple of events here are worth noting, especially when SushiSwap and UniSwap tokens went live and transaction fees were simply eye-watering. On September 2, the average fee per transaction on Ethereum exceeded $14. On that day, a total of $16.7 million worth of Ether was used to pay transaction fees, far exceeding the newly issued Ether (worth a total of $5.98 million) (that is, the block rewards received by miners). Due to the increase in transaction fees, some users decided to delay transactions, causing transaction volume to start to drop. From mid-August, we entered a period of peak transaction fees, Ethereum's daily transaction volume began to decline steadily, and transaction fees peaked on the evening of September 2, but also began to decline.
Two days later, on September 4th, I predicted on the On The Brink podcast that high transaction fees would not only affect the usage of the Ethereum blockchain, but also the liquidity of decentralized exchanges. Here's my excerpt from the podcast (starting at 36:30):
Nic: We saw this fee and volume oscillation in Bitcoin in 2017, and I expect the same to happen in Ethereum. Therefore, as the block usage rate rises, the transaction fee increases. After reaching a certain threshold, users will not be able to bear it, and then stop trading for a period of time because the transaction cost is too high. In this way, transaction volume and transaction fees are adjusted again, transaction fees become cheaper, people start trading again, and so on.
In my opinion, the sell-off seen in the market may also be caused by this. Retail investors are being forced out of the transaction fee race due to rising transaction fees. However, for smart traders, ill-informed retail investors are their favorite counterparties. In my opinion, this leads to lower liquidity.
While Ether price dynamics are hard to pin down, and I don't think Ether's price volatility is the only cause of the oscillations, Ether's USD price started to fall after peaking on Sept. 1, while transaction fees fell throughout the month. Stay high the rest of the time.
So what's going on? There are so many things. First, I think that Ethereum’s transaction fees and transaction volumes entered a major oscillation period for the first time. Transaction fees peaked three weeks after the transaction volume peaked, and then both began to decline at the same time. Even more interestingly, the average value of Ethereum transactions and the volume of several stablecoins spiked as transaction fees rose. There is a reason for this, users are more concerned about the ratio of transaction fees to transaction volume, and as transaction fees increase, they stop small transactions, and large transactions start to dominate.
image description
- Source: Coin Metrics -
This sensitivity to transaction fees is not unique to Ethereum transactions. As ethereum transaction fees increase, the transaction volume of stablecoins like usdt also rises.
image description
This shows that traders have a transaction fee threshold in mind, that is, the highest proportion of transaction fees they can accept in the transaction volume. As transaction fees increase, traders are often reluctant to make small transactions unless necessary.
All of the above can also be explained by introducing a third variable, such as the growth of liquidity mining, which both causes congestion on the blockchain and attracts large traders. This became even more clear to me after studying the composition of Ethereum transactions. My intuition tells me that small traders are discouraged by high transaction fees, but I didn't understand until I sorted out this picture:
It may not be obvious at first glance, but the graph does show that as transaction fees increase, transaction volume decreases. More interesting is the dynamics of Ether and USDT transactions, which I divide in two at $500. We can see that the proportion of ETH transactions above $500 has grown steadily as transaction fees increase. We now remove the part of the contract call and only focus on the two types of Ethereum transactions mentioned above:
- The red line is the proportion of transactions with a transfer amount below $500; the blue line is the average transaction fee denominated in ETH -
This is ironclad evidence. In non-contract call transactions, Ethereum transactions below $500 accounted for as high as 85% in June, but after the transaction fee soared, it suddenly fell to 40%. Finally, there is evidence that my hunch was right: high transaction fees discourage small transactions.
It is foreseeable that the same phenomenon can be traced in USDT, the largest ERC20 token. It can be clearly seen that as the average transaction fee climbed in mid-August, the volume of small USDT transactions began to decrease. The volume of large USDT transactions remains strong, but generally shows a downward trend, because the transaction enthusiasm gradually cools down during the peak period of transaction fees.
first level title
Will High Transaction Fees Affect Liquidity?
We have then confirmed that there is a clear negative feedback loop between transaction fees and block interval usage in both Bitcoin and Ethereum. We already know that this cycle is about two months on Bitcoin, so what happens on Ethereum? In addition, traders obviously have a transaction fee threshold relative to transaction volume in mind, and small traders will avoid transaction peaks. This causes the average transaction value of Ether and other tokens to increase significantly during peak transaction fee periods.
But is there anything else besides these? Or is this just a repeat of Bitcoin's 2017 cycle? I bet there's something else going on here.
You should know that in 2017, Bitcoin transactions were all carried out on centralized exchanges, and the blockchain was only used for inter-exchange settlement and user deposits and withdrawals. The entire market is actually off-chain. Users can deposit fiat currency into an exchange account and then hold or sell bitcoin without actually using the blockchain. Therefore, when the transaction fee crisis occurs, although the Bitcoin economy has been affected to a certain extent, many users have funds on the exchange platform, or want to transfer funds in US dollars, and they can still continue to trade.
By contrast, in 2020, transactions on Ethereum (including those yield mining tokens) largely happen on-chain. While centralized exchanges are still critical to price formation, trading volumes on decentralized exchanges like Uniswap can sometimes exceed even the largest centralized exchanges. Exchanges with an automated market maker model are much more convenient for users because they don't require users to go through KYC reviews, escrow tokens with third parties, or go through cumbersome onboarding processes. The automated market maker model without the concept of an order book is also very easy to use. Additionally, certain asset and exposure types can only be traded on-chain, for example, small DeFi tokens or participating in liquidity mining. As a result, the on-chain liquidity industry is booming. Since either Ethereum or Ethereum-based tokens are traded on these decentralized exchanges, transaction fees are charged for each transaction. Unlike centralized exchanges, every transaction on a decentralized exchange must be settled on-chain. Therefore, on-chain transaction fees are always on the minds of all traders.
Therefore, instead of the cyclical oscillations that Bitcoin exhibited in 2017, we redraw a modified version to illustrate the new dynamics of Ethereum in 2020.
Where does this second loop come from? Why single out the protocols with greater on-chain liquidity?
The first thing we need to understand is that, as Maya Zehavi stated, high transaction fees are a regressive tax on users. The so-called regressive means that the transaction fee has nothing to do with the transaction amount, regardless of whether the transfer value is 100 US dollars or 500 US dollars in ether, the transaction fee is similar (because the transaction fee is calculated based on the gas usage of the transaction, which is different from the US dollar amount of the transaction. value is irrelevant). This is similar to a sales tax which is regressive in nature. Consumption such as food accounts for a larger share of income in the working class than in the rich, so for the latter the flat 5% sales tax actually represents a significant portion of their income.
For example, let's play a game of poker, the opening commission is a fixed fee (in US dollars), not calculated as a percentage of the winnings pool. (The commission mentioned here is the operating fee charged by the casino). Players can pay a flat fee to play a certain hand, or choose to fold without paying. There was a mix of players at the table: a few pros, a few semi-pros, and a bunch of avid gamblers who didn't know much about the game. The opening commission can be set at will. To win money, you must not only beat your opponents in the game, but also consider the cost of the opening commission. That said, it's not enough just to win: you have to have a surplus after paying transaction costs.
If the opening commission is low, everyone will continue to play happily. However, once the casino becomes greedy and wants to increase the rake, players with low stacks will be more inclined to choose to fold. If a player has to pay more than 1/10 of the chip value to start a hand, they don't play (unless the hand has double aces). The general trend is that as the absolute value of the rake increases, more and more players quit the game, and the players with the least wagers are the first to quit.
When non-professional small players start to withdraw, the profits available to other players will also be greatly reduced. After all, semi-professionals make a living beating these smaller players. Every poker player knows that no one wants to play with semi-professionals and hard-working poker players on a Tuesday morning, but it's not the same as having a table with a bunch of rambunctious amateurs on a Friday night. You can watch them lose all their chips and crawl out of the bar.
Back to Ethereum, the transaction fee is the commission in the game. The high transaction fee discourages retail investors, and these retail investors have spent a lot of time on decentralized exchanges such as Uniswap. But when the transaction fee rises to an average of $14 (if it is a DeFi transaction such as buying DAI with Ethereum on Uniswap, the transaction fee will be much higher), for small retail investors, participating in transactions or liquidity mining Mining becomes unprofitable. Small retail investors had to withdraw from the market. Because retail investors are not so well-informed, they are the favorite counterparties of professional traders (leeks). If retail investors do not enter the market, the profits of the entire market will be much less, resulting in a decrease in overall liquidity and fewer trading opportunities.
We can see that transaction volume and transaction fees have followed the same trend over the past few months. This fits with speculation that high transaction fees would reduce usage, especially from a retail perspective. The problem is that causality is hard to infer. From this graph, it is easy to deduce that higher transaction volume on decentralized exchanges is the cause of transaction fees (and vice versa), rather than high transaction fees that reduce transaction volume. Even more interesting, the graph below shows the daily new addresses participating in various DeFi protocols compared to the average transaction fee on Ethereum (taken from Richard Chen's Dune Analytical report).
In addition to the two surges before and after the launch of SushiSwap and UNI tokens, when Ethereum transaction fees rose to new levels in mid-August, it seems that DeFi no longer attracts new users as it used to.
We draw only inferences. Clearly, small users defer transactions during periods of peak transaction fees, but the impact on on-chain liquidity is difficult to measure.
first level title
The Future of Ethereum Transaction Fees
Currently, we have not seen full periodic oscillations on Ethereum, so it is difficult to estimate cycle length. However, in the spirit of making specific predictions, I will give my predictions: On Ethereum, transaction fees, transaction volume, and liquidity and transaction volume on decentralized exchanges will continue to fluctuate positively . It might look like this:
If you also believe that volumes on decentralized exchanges and the on-chain liquidity environment generally drive ether prices higher, then it is easy to make educated guesses about the potential price impact of this phenomenon. Since DeFi systems need to lock up Ether to achieve full or over-collateralization, this reserve requirement affects the price of Ether. From this perspective, high transaction fees will reduce usage, which will likely lead to a reduction in the total lock-up in the DeFi system, which in turn will put pressure on the price of Ethereum.
Nonetheless, there are some counter-cyclical features that may dampen oscillations, shielding Ethereum from high volatility in transaction fees.
dynamic block space
Since its inception, Bitcoin has only officially increased its block size limit once. Ethereum, on the other hand, can generate more block space if necessary, albeit to a limited extent. In theory, elastic block space can help stabilize transaction fees, but this will bring more verification requirements, and Ethereum's verification requirements are already very high. As Ethereum’s block usage rises, the amount of computing power miners provide traders continues to reach new highs.
However, since increasing the gas limit will increase the uncle block rate, make the network more vulnerable to DoS attacks, and increase the operating cost of full nodes, the Ethereum community is divided on whether to further increase the gas limit. While Ethereum users are more accepting of trade-offs than Bitcoin users when it comes to node operating costs, both agree that simply increasing the supply of block space is not a panacea for scalability issues. Therefore, few people are willing to support a fully adjustable block space policy, although a somewhat elastic block space supply can cope with growing demand.
Ethereum 2.0 and Sharding
Rollups
When I first wrote about the future of ethereum transaction fees, I thought that ethereum transaction fees were likely to be high for a long time, which would affect the usability of non-financial applications. Vitalik's reply to me is:
What he means is that he still believes that Ethereum can achieve its vision: to be a low-cost blockchain that can run all kinds of financial or non-financial applications on it. The solution he proposed is to adopt Rollup in the short term and Ethereum 2.0 and sharding in the long term. Vitalik said that he expects Ethereum 2.0 to increase scalability by more than 100 times, but he also believes that under the influence of "induced demand", transaction fees may eventually rise to the same level (Translator's Note: "Induced Demand" An example is that building more roads does not alleviate congestion, because more roads make cars more cost-effective, making people willing to buy more cars).
I tend to agree with him that if a good is produced more efficiently, people will find more uses for that (cheaper) good. Therefore, even taking into account the increased block space, we will see the average transaction fee reach a high level. In any case, Ethereum 2.0 is still far away, and it is difficult to draw conclusions about it. At least, we won't see fee adjustments anytime soon.
In the Ethereum community, the current orthodoxy is that Rollup is the main way to alleviate Ethereum’s current transaction fee problem. There are two main types of Rollup: ZK and Optimistic. The two mainly increase the economic density of transactions by bundling multiple payments together. In theory, they can guarantee the security of underlying transactions while greatly improving throughput. In simple terms, traders rely on relayers to aggregate large numbers of transactions and broadcast summaries of those transactions.
ZK Rollup requires the broadcast of heavily truncated transaction stubs, along with proofs of validity for this batch of transactions. Optimistic Rollup requires quasi-trusted operators to aggregate transactions, and traders need to largely assume that these operators will not do evil. Theoretically, anti-incentives are implemented by combining proof-of-fraud with financial penalties for malicious behavior. Currently, ZK Rollup only supports simple transfers, while some ORs are expected to handle all types of transactions on the current version of Ethereum. For a comprehensive overview of OR, see this comprehensive report by Daniel Goldman.
Rollup can transfer transaction data off-chain and put validity proofs on-chain (Translator's Note: The author's statement here is wrong. Rollup schemes such as Optimistic rollup put calculations off the chain and publish transaction data on the chain). Bitcoin also adopted the same idea of data simplification a few years ago, launching Lightning Network (which can reduce hundreds of thousands of transactions to a few on-chain transactions) and sidechain technology, and also advocates measures to improve efficiency, such as, Batching and SegWit Thanks to Rollup's popularity and rapid development process, Vitalik strongly supports the Rollup solution as the best short-term solution to Ethereum's scalability. While Ethereum 2.0 may not be ready yet, Rollup's vision is within reach.
Rollup may not be the panacea for Ethereum’s transaction fee problem for two main reasons. First, it is not that easy to coax all block space consumers into block stewards, especially for service providers, who could pass the fees on to end users without internal digestion. The lesson we have learned from Bitcoin is that if intermediaries can pass fees on to end users, they have very limited incentives to invest in sustainable network infrastructure.
Also, since Ethereum has repeatedly raised the gas cap to bail out blockspace consumers (at the expense of increased verification work), heavy users may be more inclined to lobby hard to further raise the gas cap rather than spend time packaging Rollup transactions. Similarly, the large block space that Ethereum 2.0 can bring may reduce the enthusiasm of heavy users for Rollup. The current main "contribution" of Ethereum 2.0 may be to cause heavy users to be lazy to improve the efficiency of block space usage.
Second, a blockchain like Ethereum has a huge user base around the world, and it cannot be called upon only by developers or advocates. Ethereum is open and does not exclude any participants, so it is welcomed by Ponzi schemes or other crooked methods. And the perpetrators of these scams (who are also usually heavy consumers of block space) do not plan for the future of the blockchain, nor do they want to optimize their on-chain footprint. Of course, it would be interesting to see Forsage use Rollup to package Ponzi scheme transactions and push the technology forward.
Introducing more complex systems such as Rollup introduces uncertainty about atomicity and composability. Since some ORs employ a trust model centered on the challenge of proof of error, the time to achieve finality becomes less optimistic. Matter Lab estimates that it may take 1-2 weeks for the OR to achieve finality. (Note: This is an estimate made in November 2019, and the existing technical level may have changed.) In order to solve this problem, there is a solution ['withdrawal period'] that allows users to withdraw the temporary funds during the challenge period as early as possible through an intermediary Frozen tokens. My understanding is that this essentially creates a tiered system where the "free version" Rollup requires a longer endgame period, while the paid accelerated version shortens that time.Chinese translation）。
In addition to finality issues, the introduction of OR may also affect composability. I'm not really a Rollup expert, but I read that one of the biggest shortcomings of OR is the composability problem (
in conclusion
Chinese translation
For the above reasons, I don't think all major smart contracts on Ethereum will switch to the Rollup system. To me, there is a stark difference between a fast-settling ether or other token transfer and a Rollup version of a token transfer, especially in terms of settlement. Also, if composability or atomicity is compromised, then the system we will build will be very different from the existing system. There must be something unique about DeFi consisting of fast-settling ether and token transfers, and I believe this will become apparent when large block space consumers start switching to the Rollup system. Therefore, even if some smart contracts are transferred to Rollup, I don't think Rollup will effectively reduce transaction fees in the short term.
in conclusion
The difference between centrifugal governors and on-chain transaction fees is that regulation fees are a side effect of a properly functioning system rather than a critical design consideration. In public chains, transaction fees exist to ensure that network resources are not wasted and to provide income to validators, not to inhibit usage of the system. But in practice, transaction fees do have such an effect, and the disincentive becomes stronger and stronger.
