EIP-1559 Proposal in the Filecoin Network

In the Filecoin network, the concept of Gas is used to measure the computational resources required to execute a given on-chain transaction. The cost of these resources needs to be paid, resulting in many decisions that must be made around who should bear these costs and how transactions that consume limited resources need to be prioritized. This is a summary of Juan Benet's talk.

For example, one approach might be to conduct a first-price auction. In such a system, participants who wish to execute transactions on the blockchain submit bids to block miners. Miners then collect these bids as fees when they include the relevant transactions in a mined block. Given the size of a block or an upper limit on the amount of gas that a block can be traded, miners prioritize transactions to optimize the fees they collect.

While conceptually simple, such approaches have been shown to have a number of flaws. First-price auctions can be complex to operate and can cause bidders to overpay significantly. They can also lead to undesirable behavior during periods of network congestion.

The system detailed above also rewards miners with the entirety of their gas fees, ignoring the fact that non-mining participants in the network also necessarily expend resources to process each transaction.

Ethereum Improvement Proposal (EIP) 1559 is a standard developed in response to many of these issues. Its main innovation is the introduction of a base fee (per unit of Gas) associated with each block. This fee can increase or decrease due to network congestion, pushing network traffic back to target levels.

Under EIP-1559, transaction creators specify a fee cap (the highest fee per unit of gas they are willing to pay to include this transaction). If the fee cap is less than the block's base fee, it cannot be included in the block. Transaction creators also specify a tip (also in units per unit of gas); this tip is collected by block miners.

Instead of a single-price auction, the minimum fee paid by a transaction creator is a) the fee cap and b) the base fee plus the tip, multiplied by the fee used in the transaction. All revenue from the base fee is burned, creating deflationary pressure so that the entire network is compensated for the resources spent executing transactions.

Miners, in turn, receive at least a) a tip, and b) the fee cap minus the base fee, multiplied by the gas used for the transaction.

Under this scheme, transaction creators can ensure that their high-priority transactions still make it into the blockchain, while typically paying much less in fees.

EIP-1559 has many compelling properties that align well with Filecoin’s design considerations:

1. Efficiency. EIP-1559 outlines a more efficient model for setting and calculating Gas.

2. User experience. Estimation and fee setting are much easier than participating in a price auction.

3. Critical message throughput. WindowPost messages in Filecoin are time-critical and must be processed in a way that is robust to congestion. EIP-1559 provides higher reliability and throughput for such messages.

4. Transaction rewards the network. The entire network bears the cost of processing transactions and should be compensated accordingly.

For all of these reasons, Filecoin incorporated EIP-1559 into its core protocol with minimal modifications. Filecoin uses a trick set; its transactions enter the chain before they are executed, and until this happens, the actual Gss usage cannot be determined. Therefore, Filecoin requires users to provide an estimated transaction limit. In order to properly align incentives, overestimations above a certain margin are burned.

EIP-1559 has been implemented and is live on the Filecoin mainnet. Currently, about 100k to 150k FIL is consumed per day due to network transactions. The inclusion of the standard achieves at least two key wins:

1. Fast track for high value transactions

At times, Storage messages can threaten the price of more time-sensitive WindowPost messages. However, as expected, integrating EIP-1559 provides Filecoin participants with a straightforward mechanism to address this congestion without always paying a high price.

2. Network capacity management

So far, EIP-1559’s base fee mechanism has done a good job of keeping the network capacity at 100% of its target.

Over time, transaction creators have also gotten better at estimating the gas required for their transactions, reducing overestimation of consumption in the long term.

There is still some room for improvement: the base fee varies widely and its rate of change is spiky. This is due to the massive congestion as new storage is launched. In the past this has also caused important WindowPost messages to be priced out as well.

The Filecoin community could investigate a number of approaches to address these pricing issues:

1. Gas-controlled aircraft

One possible way to address network congestion is to create dedicated gas channels for control plane transactions, reserving a portion of each block for specific messages that are critical to the blockchain’s functioning.

2. Message type fee structure

Another way to solve this problem might be to reduce the cost of WindowPost, or increase the cost of other messages (especially Pre/ProveCommit messages).

3. Basic cost change rate

The peakedness of the base fee over time indicates that the rate selection is not ideal and could be smoothed out.

Other methods that have nothing to do with the gas model can be introduced to help alleviate congestion:

1. Proof of scaling

Ideas for expanding Filecoin’s proof mechanism include batch verification of SNARKs, as well as slash-based verification to reduce the amount of verification that needs to be done up front.

2. Scaling consensus

Given current congestion rates, Filecoin may need to seek shards quickly.

Looking ahead, there is a lot of potential data analysis work on the results achieved so far by the Filecoin network, and opportunities to collaborate with other researchers and the blockchain community to build on the ideas of EIP-1559.

Some possible areas of research include introducing queuing theory into the Gas model (where there may be inefficiencies in the current model), guaranteeing quality of service for certain messages and transactions, and exploring ways to introduce encrypted transactions into the blockchain so that they can be executed at some point in the future.