Implementing a more efficient view tracking mechanism

Oct 13, 2025 by recoveryshake.com 102 views

Introduction

In the realm of consensus algorithms, particularly in the context of simplex and threshold-simplex, the activity_timeout parameter plays a crucial role in determining the number of views to track and whether to skip a leader suspected to be offline. However, this dual functionality often leads to a trade-off between maintaining a reasonable buffer at the tip of recent views and triggering a "fast exit" mechanism. In this article, we will delve into the importance of separating activity_timeout from leader_skip_timeout and explore the benefits of this configuration change.

Current Implementation

The current implementation of activity_timeout in both simplex and threshold-simplex is as follows:

// threshold_simplex/actors/voter/actor.rs
// ...

// activity_timeout determines how many views we track
let activity_timeout = config.activity_timeout;
let num_views = activity_timeout;

// activity_timeout also determines whether to skip a leader
if leader.is_offline() {
    if leader.skip_timeout <= activity_timeout {
        // skip the leader
    }
}

As evident from the code snippet above, the activity_timeout parameter is used to determine both the number of views to track and whether to skip a leader suspected to be offline. This dual functionality can lead to a situation where we want to maintain a reasonable buffer at the tip of recent views but trigger a "fast exit" mechanism much faster.

Separating `activity_timeout` from `leader_skip_timeout`

To address this issue, we propose separating activity_timeout from leader_skip_timeout. This would allow us to configure these two parameters independently, providing more flexibility and control over the consensus algorithm.

Benefits of Separation

Separating activity_timeout from leader_skip_timeout offers several benefits:

Improved flexibility: By configuring these two parameters independently, we can tailor the consensus algorithm to our specific use case, ensuring that we maintain a reasonable buffer at the tip of recent views while triggering a "fast exit" mechanism much faster.
Better control: With separate configuration options, we can fine-tune the consensus algorithm to meet our specific requirements, reducing the likelihood of conflicts between these two parameters.
Simplified debugging: In the event of issues with the consensus algorithm, separating activity_timeout from leader_skip_timeout makes it easier to identify and debug problems, as we can isolate the root cause of the issue.

Implementation

To implement this change, we can introduce two new configuration options: activity_timeout and leader_skip_timeout. These options would be used to determine the number of views to track and whether to skip a leader suspected to be offline, respectively.

// threshold_simplex/actors/voter/actor.rs
// ...

// activity_timeout determines how many views we track
let activity_timeout = config.activity_timeout;
let num_views = activity_timeout;

// leader_skip_timeout determines whether to skip a leader
let leader_skip_timeout = config.leader_skip_timeout;
if leader.is_offline() {
    if leader.skip_timeout <= leader_skip_timeout {
        // skip the leader
    }
}

Conclusion

In conclusion, separating activity_timeout from leader_skip_timeout offers several benefits, including improved flexibility, better control, and simplified debugging. By introducing two new configuration options, we can tailor the consensus algorithm to our specific use case, ensuring that we maintain a reasonable buffer at the tip of recent views while triggering a "fast exit" mechanism much faster. We believe that this change will have a positive impact on the consensus algorithm and recommend implementing it in the next release.

Related Issues

This change is related to the following issues:

Alto Issue #15

Future Work

In the future, we plan to explore other ways to improve the consensus algorithm, including:

Implementing a more efficient view tracking mechanism: We plan to investigate alternative view tracking mechanisms that can reduce the overhead of tracking views and improve the overall performance of the consensus algorithm.
Introducing a more robust leader election mechanism: We plan to explore alternative leader election mechanisms that can improve the resilience and fault tolerance of the consensus algorithm.

Introduction

In our previous article, we discussed the importance of separating activity_timeout from leader_skip_timeout in the context of consensus algorithms. In this article, we will address some of the frequently asked questions related to this change.

Q: What is the current implementation of `activity_timeout`?

A: The current implementation of activity_timeout is as follows:

// threshold_simplex/actors/voter/actor.rs
// ...

// activity_timeout determines how many views we track
let activity_timeout = config.activity_timeout;
let num_views = activity_timeout;

// activity_timeout also determines whether to skip a leader
if leader.is_offline() {
    if leader.skip_timeout <= activity_timeout {
        // skip the leader
    }
}

Q: Why do we need to separate `activity_timeout` from `leader_skip_timeout`?

A: We need to separate activity_timeout from leader_skip_timeout because the current implementation leads to a trade-off between maintaining a reasonable buffer at the tip of recent views and triggering a "fast exit" mechanism. By separating these two parameters, we can configure them independently, providing more flexibility and control over the consensus algorithm.

Q: What are the benefits of separating `activity_timeout` from `leader_skip_timeout`?

A: The benefits of separating activity_timeout from leader_skip_timeout include:

Improved flexibility: By configuring these two parameters independently, we can tailor the consensus algorithm to our specific use case.
Better control: With separate configuration options, we can fine-tune the consensus algorithm to meet our specific requirements.
Simplified debugging: In the event of issues with the consensus algorithm, separating activity_timeout from leader_skip_timeout makes it easier to identify and debug problems.

Q: How do we implement the separation of `activity_timeout` from `leader_skip_timeout`?

A: To implement this change, we can introduce two new configuration options: activity_timeout and leader_skip_timeout. These options would be used to determine the number of views to track and whether to skip a leader suspected to be offline, respectively.

// threshold_simplex/actors/voter/actor.rs
// ...

// activity_timeout determines how many views we track
let activity_timeout = config.activity_timeout;
let num_views = activity_timeout;

// leader_skip_timeout determines whether to skip a leader
let leader_skip_timeout = config.leader_skip_timeout;
if leader.is_offline() {
    if leader.skip_timeout <= leader_skip_timeout {
        // skip the leader
    }
}

Q: What are the related issues to this change?

A: This change is related to the following issues:

Alto Issue #15

Q: What is the future work related to this change?

A: In the future, we plan to explore other ways to improve the consensus algorithm, including:

Implementing a more efficient view tracking mechanism: We plan to investigate alternative view tracking mechanisms that can reduce the overhead of tracking views and improve the overall performance of the consensus algorithm.
Introducing a more robust leader election mechanism: We plan to explore alternative leader election mechanisms that can improve the resilience and fault tolerance of the consensus algorithm.