|
With the launch of the second phase of Filecoin testnet, the official also announced a 4 million FIL testnet reward plan. The Filecoin mainnet has entered the countdown stage, and the market enthusiasm is unprecedentedly high. Whether it is the mining circle, technology circle, market circle, investment circle or media community circle, they have all poured in. Various bigwigs have been calling for Filecoin recently. Almost all online and offline forums and summits have IPFS & Filecoin special sessions every day. It is hard for Filecoin not to be popular. Image from Filecoin official website Miners are also actively making final preparations for the launch of Filecoin’s mainnet. The prices of some mining hardware have soared recently, and it is said that AMD motherboards have been snapped up by miners. In order to get the 4 million test rewards next month and the last wave of marketing before the mainnet launch, various manufacturers are showing their talents. Miners with low computing power have higher unit profits? Recently, a picture has been widely circulated in various WeChat groups. The statistician used the miner's ( total coin production ÷ effective computing power) to calculate the miner 's single ton coin production. The data shows that miners with lower computing power rankings have higher unit income, while miners with the top three computing power rankings cannot even enter the top 10 in terms of single ton coin production. Many investors are getting more and more confused and don't know how to make a decision.In fact, this issue was addressed in the previous version of the test network when a self-media published an article mentioning the statement "single T output is the first". Later, they realized that this algorithm was wrong and issued a statement to correct and apologize. Why is this statement wrong? We can think of effective computing power as the goods produced by miners, and block rewards as the sales revenue of miners. Company A's total effective computing power is 3.35PiB, with an average daily effective computing power of about 500TiB, a total weekly output of about 450,000 coins, and an average daily mining output of about 64,000 coins; while Company B's total effective computing power is 413.81TiB, with an average daily effective computing power of 60TB, a total weekly output of about 77,000 coins, and a daily mining output of 11,000 coins. If both companies have 1,000 mining machines participating in the profit distribution (the number of devices participating in the profit distribution is based on the number of devices actually sold by the company), then the average daily output of a mining machine in Company A is 64 coins, while the average daily output of a mining machine in Company B is 11 coins. The difference between the single-machine output of Company A and Company B is about 6 times. Do big miners use a lot of equipment to brush data? Recently, some people have speculated that the top few big miners in the test network used thousands or even tens of thousands of devices to generate data, while the miners ranked lower only used a dozen or even one or two devices to participate in the test. Some people have also calculated how many mining machines each company used to participate in the test. See the figure below. This chart is based on the total number of new sectors added by each node per day divided by 4, and the number of mining machines used by each node is calculated. Because according to the official test data of the Protocol Lab, the average time for a device to package a sector is 5-6 hours, which means 4 sectors per day. For example, the number of sectors packaged per day for the "t01009" node ranked first is 8,400, which they calculated to be 2,100 devices. The number of sectors packaged per day for the "t011101" node is 1,000, so the number of devices used is 250. However, this data can only be said to be a guess by novices or non-professional miners. For example, last month, a domestic miner challenged the protocol laboratory and completed the data packaging in less than 3 hours. Recently, a team also said that it can process 12 sectors in parallel at the same time, and the average time for a single machine to package a sector can even be compressed to less than one hour. Therefore , many amateur players cannot understand the cluster solutions of professional players . Just like playing games, amateur players are still studying strategies on how to defeat the boss , while professional players have already developed plug-ins . Spacetime Cloud has long reconstructed the official cluster solution and built its own cluster system. At present, our cluster solution can encapsulate more than 10 sectors in parallel at the same time , so the number of devices used in our current computing cluster is less than 200. The steps of Filecoin mining can be simply divided into two steps: worker and miner. The worker is responsible for calculation, which uses the SDR algorithm to perform mathematical calculations on the original data, and then encapsulates the calculated data into the sectors of the hard disk, and submits it to the chain to generate a copy proof. The miner obtains computing power. This process consumes a lot of CPU, memory and GPU resources. Then the Filecoin network allocates the right to package blocks (that is, the right to vote for block packaging) according to the computing power held by the miner. The greater the computing power and the higher the winning rate, the miner needs to repeatedly submit time-space proof when participating in block packaging. The node that completes the time-space proof can get the reward for block packaging. What everyone understands as a mining machine is actually just a storage server. However, the equipment deployed in a professional Filecoin mining pool is far more than just a storage server. The Filecoin mining pool is different from the Bitcoin mining pool. It uses a single-machine stacking cluster solution, but a distributed computing + storage integrated cluster solution. Therefore, the hardware configuration of the Filecoin mining pool generally separates computing and storage, builds computing clusters and storage clusters separately, and then adds a zk-SNARK (zero-knowledge proof) cluster. The process of miners obtaining computing power mainly depends on the performance of the CPU, GPU and memory, that is, the performance of the mining pool's computing cluster. Then, the income is distributed in proportion to the storage capacity purchased by the customer to the total number of storage servers in the mining pool (committed capacity). Therefore, the effective computing power obtained by a mining machine company every day does not depend on how many mining machines (storage servers) it uses, but on the hardware configuration of its computing cluster, cluster solutions, and algorithm optimization capabilities. After the mining pool mines, it distributes the profits in proportion to the number of mining machines it sells (not the effective computing power). If we use effective computing power to calculate the output per terabyte, in the previous version of the test network, the "t01020" node we built stopped data packaging after packaging 170 terabytes of computing power in the first two days, and the computing power stopped growing, and only participated in block packaging. At the highest point, the cumulative output of 1020 nodes reached 320,000 coins, which is the node with the most single-terabyte output in history. There are also some nodes whose computing power has been fined to zero, but they have participated in block packaging and have income before, so the income of this node can be said to be close to infinity. As for those who say that only a few mining machines are currently used for testing, it is a bit misleading. Many companies advertised how many devices they deployed before the second test began , and after the test results came out, they said that they only used a few or a dozen devices for testing. Could it be that the hundreds or thousands of devices deployed in their computer rooms were just for photo ops? Moreover, the mainnet is getting closer and closer. The significance of the testnet is to test the security, stability and stress test of the Filecoin network on the one hand, and to test the hardware configuration, cluster solution and algorithm optimization capabilities of each mining pool on the other hand. If you are still running with a few devices now, how can you build a cluster of hundreds or thousands of devices after the mainnet is launched? Have you ever seen a construction team of migrant workers who can only build a two- or three-story building at ordinary times, and now they can build a 100-story skyscraper? They underestimated the difficulty of Filecoin mining. Moreover, due to the characteristics of Filecoin block explosion, the block explosion rate of the top three miners in terms of computing power is basically 5 times or even more than 10 times that of miners outside the top three. Therefore, the current test is to ensure the data packaging efficiency of the mining pool and strive to enter the top three in computing power to ensure sufficient yield. Of course, if the small miners behind really only use a few or dozens of devices to participate in mining in the future, the input-output ratio may not be too low. However, although many miners currently have relatively small computing power, they actually sell a lot of mining machines, so this is a very dangerous thing. Therefore, the correct algorithm for calculating the unit income of miners is to distribute the income according to the total amount of coins produced by the mining pool divided by the total number of storage servers (committed capacity) used by the mining pool. At present, many mining machine manufacturers with low computing power rankings can be said to have only used a few devices for testing, but in the future, no matter how many, dozens or even thousands of devices they use for mining, the final income distribution will be distributed to their investors in proportion to the number of mining machines they have sold. How much coin is expected to be produced by 1TB of cloud computing power? Many people are most concerned about the expected unit coin production in the future. I am asked this question dozens of times every day, but there is really no way to calculate it. Because the only thing we know is that after the launch of the Filecoin mainnet, the FIL released by mining in the first year is 152 million, an average of about 400,000 per day. However, the daily coin production of miners is related to the number of effective miners participating in mining, which is a variable. After the mainnet is launched, there will be an increase or decrease in miners every day. It is difficult for us to predict the data of the entire network, so let's take our own data of Spacetime Cloud as an example. Our expected goal is to have 1,000 devices participating in the first mining before the main network is launched (about 200PB, or 200,000TB). If our effective computing power accounts for 20% of the entire network in the early stage, and the corresponding coin production is also about 20%, then our mining pool will produce about 80,000 coins per day on average, and the average production ratio per TB is 0.4 coins. However, if the price of Filecoin performs well after the launch of the mainnet, and a large amount of capital flows in, the number of devices in our mining pool may surge to 5,000 in 3-6 months. Assuming that the computing power of the entire network increases simultaneously, our share of the computing power of the entire network is still 20%, and the daily production ratio is still 80,000, then the average production per TB is 0.08. Since the total amount of Filecoin is limited and released in a linear decreasing manner, the cake is so big (and getting smaller and smaller), the more people participate, the less coins each person gets. However, the more people participate, the stronger the consensus, and the more support and room for the price to rise. Moreover, according to the current computing power and block rules, the top five nodes in computing power are likely to mine more than 80% of the tokens in the entire network. The nodes with lower computing power may only participate in the distribution of the remaining 20% of the tokens, so it is important to choose the top mining pools. As for how many coins can be distributed per unit storage capacity, no one can calculate a specific or even approximate value now. Regarding Filecoin mining, due to its technical complexity and the fact that the mainnet has not been online yet, there is a lot of information asymmetry, and many investors have a big misunderstanding about Filecoin mining. Recently, I will write columns regularly to answer questions about these misunderstandings. Welcome everyone to follow. For more information about IPFS and Filecoin, as well as friends who want to participate in Filecoin mining, you can contact Kuangge (WeChat ID: kuangge2018) . |
![[Advanced Classroom] How many coins can a mining machine mine?](/upload/images/67e6abf575ab4.webp)
