The Force Research | How to “save money” in Filecoin mining

As the storage scale of Filecoin grows, the total daily reward also increases, but the gas fee remains high, which indirectly affects the net income of miners. So how should we deal with it at this stage?

1. Gas remains high, is "throttling" a definite option?

In Filecoin mining, "open source" is difficult for miners to master, and is almost strictly restricted by "halving every 6 years", "baseline network", and "the scale of growth of computing power of the entire network"; but "throttling" is controllable by miners. In addition to maximizing the use of hardware, the most concerned issue at present is how to minimize the gas fee: when to increase and maintain computing power? How to set parameters to minimize the gas fee? How to configure the hardware to obtain the optimal solution? The above are all problems that miners urgently need to explore and solve in "throttling".

Total transaction fees from 2020/10/15 to 2021/2/4, source: IPFS Force Zone, 2020-02

In the face of Gas throttling, our first consideration is ProveCommit and PreCommit. According to Force Zone data statistics, from the launch of the main network to now, the total Gas consumption of the entire network has exceeded 10 million FIL, with ProveCommit accounting for a cumulative proportion of about 68.06% and PreCommit accounting for a cumulative proportion of about 24.78%. So for the above, how to analyze the current situation and solve it?

2. Factors affecting Gas Fees

According to incomplete data statistics from the Force Zone, the main reasons for exceeding the average Gas Fee are: the impact of node size, excess burning of ProveCommit and PreCommit, ProveCommit and PreCommit failure rate, etc., among which scale has the greatest impact, accounting for about 64%.

The weight of each factor affecting the cost of Gas consumption exceeding the average level, source: IPFS Force Zone, 2020-02

1. Size and Gas Fee: 64%

Single message handling fees for clusters of different sizes, source: IPFS Force Zone, 2020-02

Based on statistics of single message handling fees of 860 miners from 2021/1/25 to 2021/1/31, they are divided into four scales: below 1PiB, 1-10PiB, 10-20PiB and above 20PiB. Among them, for clusters below 1PiB, the average cost of PreCommit successful handling fees is about 0.04044FIL/message, and the average cost of ProveCommit successful handling fees is about 0.17223FIL/message; for clusters above 1PiB, the average cost of PreCommit successful handling fees exceeds 0.06326FIL/message (56.43% higher than the former), and the average cost of ProveCommit successful handling fees exceeds 0.17831FIL/message (3.53% higher than the former). The main analysis is as follows:

PreCommit has the greatest impact on scale, ProveCommit is less affected by scale, and 1PiB is the watershed;

The handling fee for a single message in clusters of different sizes, source: IPFS Force Zone, 2020-02.

Further subdividing the cluster summary, the single message fee of 15-20PiB ProveCommit and PreCommit showed a downward trend, and then slowly rose further. The main reason, the author analyzed, is that in the process of the increase in the scale of the mining pool from 1-15PiB, miners have begun to gradually overcome the regular setting of this volume, but PreCommit faces the problem of high handling fees for the growth of computing power with a larger volume (above 20PiB).

2. Excess Burning and Gas Fee: 28%

Statistics show that from 2020/10/15 to 2021/2/4, the excess burning fee was about 430,000 FIL, of which ProveCommit accounted for 74.20% and PreCommit accounted for 25.80%.

2020/10/15-2021/2/4 ProveCommit and PreCommit overburn, source: IPFS Force Zone, 2020-02The

above is a line chart of ProveCommit and PreCommit daily overburn. ProveCommit consumes an average of about 2,800 FIL per day, and PreCommit consumes an average of 975 FIL per day. Observing the above figure, it is found that the trends of the two figures are inconsistent. So what are the correlations between ProveCommit and PreCommit overburn? The author believes that the following are:

PreCommit overburn is positively correlated with the number of failed items/Base Fee fluctuations. Because PreCommit needs to complete the task within 30 minutes, and the Base Fee changes too quickly, it will cause too many failures, indirectly causing some miners to prefer to overpay rather than bear the risk of failed order burning;

ProveCommit overburn is positively correlated with computing power growth/total handling fees. Because EIP 1559 requires that once the Filecoin usage rate exceeds 50%, the basic fee will automatically increase, and it will show exponential growth. Therefore, once the computing power grows too fast, the total handling fee will inevitably rise. In order to ensure the smooth completion of the task, some miners will inevitably pay excess handling fees to complete the last proof of computing power growth.

3. Failure rate: 8%

As can be seen from the figure below, the failure rate of ProveCommit is generally higher than that of PreCommit, both within 5%; from the launch of the main network to February 4, 2021, the consumption involved in the replication proof exceeded 30,000 FIL, of which ProveCommit failure consumption accounted for 97.85% and PreCommit accounted for 2.15%. The main analysis is as follows:

2020/10/15-2021/2/4 Failure rate of replication proof, source: IPFS Force Zone, 2020-02

The reason why the failure rate of PreCommit is lower than that of ProveCommit is that the pre-penalty and difficulty of PreCommit are not high. There is a penalty in the PreCommit stage. PreCommit requires the delivery of pre-charged pledges. If the miner fails to complete ProveCommit in time within 30 minutes, the pre-charged pledge will be deducted. This part of the fee will be higher than the payment of Gas fee; at the same time, PreCommit is not difficult to implement compared to ProveCommit, but the time requirement is more stringent.

The failure rate of ProveCommit is high in the medium term and gradually decreases in the later period. In the medium term, it is because the exponential growth of Gas fees caused by the EIP 1559 mechanism and the frequent upgrades of versions directly lead to failed attempts by miners to adapt to the new rules and avoid excessive burning. However, after adapting to the network, the failure rate gradually decreases by about 40%.

So what is the idea of ​​the Force Zone for the above problems?

3. Sincerity leads to "saving money"

1. Impact of node scale: Replace small cluster computing power growth

After calculating the gas fees of nodes of different sizes and being familiar with the gas fee rules of large-scale nodes, it is more cost-effective to choose small nodes to increase computing power, which greatly reduces the customer's expenditure cost. According to the data from 2021/1/25-2021/1/31, small clusters (within 10PiB) will have lower expenditure costs than large clusters (above 10PiB), and PreCommit is about 24% lower (varies with the network and rules).

2. Excess burning: Reduce or avoid

various ways such as setting recommended values, automatic dynamic adjustment of GasUsed/GasLimit to alleviate basic burning, excess burning, and special processing of data structure compression to reduce Gas consumption, so as to reduce unnecessary burning of Gas as much as possible. For example

, ForcePool has achieved excess burning with SubmitWindowedPoSt of 0. 3. Avoidance of information accumulation

Avoid message accumulation through reasonable push of messages, and avoid more uncertainty problems in the future. In this regard, ForcePool will increase its sensitivity to the Base Fee on the one hand, and on the other hand, it will dynamically adjust the relevant parameters of the sectors to get the information on the chain as soon as possible to reduce the backlog and the number of error messages.

4. New hybrid practice: 64G/32G sector dynamic packaging

For the excessive fluctuation of the Base Fee, ForcePool will adopt the output configuration of 64G sectors and 32G sectors to solve or alleviate the Gas problem, so as to dynamically adjust the Gas and packaging speed.

In general, we have been exploring the road of "throttling".

4. Exploration never stops, and "throttling" is endless

In the future, the Force Zone will continue to pursue the road of "throttling", mainly from two directions: multi-dimensional and precise research.

Multi-dimensional is the possibility of exploring more "throttling" directions. Seeking to reduce costs in unknown and blank areas may include: maximizing the use of resources, reducing the angle of Gas fees, Venus to ensure the stability of operation, etc., and finding which definite "throttling" direction in uncertainty.

Precise research is to further explore the already determined road, from 90 points to 95 points, or even 100 points. With a clear direction, we will further explore more optimized strategies in terms of scale impact, excess burn and failure rate.