A discussion on how Filecoin deals with the impact of network computing power.

Source: Filecoin Network

TLDR

• Due to the built-in anti-fragile mechanism in the Filecoin protocol, Filecoin is able to withstand major drops in computing power.

• This mechanism provides strong economic incentives for new storage providers (SP) to join and for existing storage providers to continue staying in the network even after power outages.

• Specifically, after a significant loss of computing power, incoming computing power will receive extra rewards due to a rapid increase in reward concentration.

Introduction

Filecoin is a distributed data storage network that stores humanity's most important information. SPs provide storage capacity to the network within a specified contract period and receive block rewards as compensation. To ensure the fulfillment of contracts, FIL tokens are used as collateral. The amount of collateral and the block rewards obtained will vary depending on the type of incoming computing power (committed capacity vs real data).

Elasticity, what does it mean? The Merriam-Webster Dictionary defines elasticity as the ability to recover or adjust easily from unfortunate events or changes. Elasticity is relevant to many aspects of encryption economies like Filecoin, including 51% attacks, Sybil attacks, and Finney attacks. This article will focus on a specific aspect of elasticity: the resilience of the Filecoin network to computational shocks. In Filecoin, computational power refers to the storage capacity and data volume of the network. For various reasons, when storage providers (SP) leave the network, it can result in a power outage.

First, let's accurately address the elasticity-related questions we want to answer quantitatively: How will the Filecoin network cope with power outages if an SP experiences the following situations?

• The SP gracefully exits and disconnects from the network after completing the storage contract.

• The SP abruptly terminates the storage contract before the end date, causing a shock to the network and sudden power outage.

Modeling

How to model network resilience in a structured way and ensure that the predictions adhere to Filecoin's cryptographic economics? Our approach is to use an agent-based model (ABM) of the Filecoin economy, where the ABM consists of agents that interact with the environment. In this ABM, the environment implements the cryptographic economic mechanisms of Filecoin (such as interactions related to locking, ownership, minting, and token supply), while the agents represent the models of service providers (SPs). For more information about the Filecoin ABM, please refer to the article.

To explore resilience, we build agents to model three types of SPs in the Filecoin network:

• SPs that only load Committed Capacity (CC) sectors

• SPs that only load Deal sectors (Fil+)

• SPs that load both CC and Deal sectors

Each SP can exhibit the following behaviors:

• Average Dollar Cost behavior - In this mode, the agent accesses and continuously updates pre-configured computing power until the end date, after which the agent leaves the network (due to power expiration or termination).

• Adaptive ROI behavior - In this mode, if the predicted return on investment (FoFR) is greater than a configured threshold, the agent uses the predicted FoFR (FIL-on-FIL Returns) to determine the load of computing power.

With these agent and environment specifications, we can configure multiple counterfactual scenarios that represent different network states that could result in computing power shocks. The network states are defined by the distribution of SP types running in the network, and each network state could lead to different outcomes. Because each state has unique raw bytes and quality-adjusted computing power, it results in different dynamics for minting, pledging, and FoFR. The initial network state we tested is:

• Uniform distribution of CC, FIL+, and mixed SP

• Only CC and FIL+ storage providers have a uniform distribution

• The ratio of FIL+ and CC storage providers is 70/30 (FIL+ biased)

• The ratio of FIL+ and CC storage providers is 30/70 (CC biased)

From these points, we simulated counterfactual scenarios where the network's computing power decreases by 30% or 70% compared to the current value. We considered two cases of computing power decrease: a) gradual exhaustion of computing power, b) abrupt termination of computing power.

In summary, we simulated the 30% or 70% decline in network computing power with progressive sector termination or immediate termination events. We simulated each configuration and recorded network key performance indicators (KPIs).

Results

There are some subtle differences between gradual power outage and sudden power outage scenarios, but in both cases of computing power loss, we observed that the network computing power starts to recover after the termination event due to an increase in reward concentration, resulting in increased FoFR.

Working Principle: After termination, the network computing power decreases rapidly, but it has a counter effect on the standardized rewards for each QA sector, leading to an increase in rewards. Combined with the decrease in staking, reward concentration occurs, causing a sharp increase in FoFR. This combination of network conditions provides higher FIL-on-FIL returns for participants remaining in the network. Our simulation reflects this phenomenon—here, agents (rational actors) observe high FIL-on-FIL returns and take advantage of the situation to acquire more computing power, thereby restoring the network's computing power.

For more details, additional scenarios, and discussions, please refer to this report.

The figure above shows the KPI of the Filecoin network under various scenarios of SP leaving the network. The vertical dashed line represents the start of the simulation, and the dotted vertical line represents the date when computing power starts to flow out of the network. In addition, a baseline scenario with constant loading simulated using DCAAgent is also provided for comparison (represented by the black dotted line).

The figure above shows the KPI of the Filecoin network under various scenarios of SP leaving the network. The vertical dashed line represents the start of the simulation, and the dotted vertical line represents the date when computing power starts to flow out of the network. In addition, a baseline scenario with constant loading simulated using DCAAgent is also provided for comparison (represented by the black dotted line).

Conclusion

In conclusion, we explored the resilience of the Filecoin network in response to network computing power shocks. The conclusion is that the network has resilience due to the concentration of rewards, which increases the income of remaining storage providers, and it also exhibits anti-fragility due to the baseline recovery mechanism.

If you would like to learn more about CryptoEconLab and our work, please visit our website or follow us on Twitter.

This article is for reference only. CryptoEconLab does not provide legal, tax, financial, or investment advice, and no party should rely on or expect any related advice.