vac:dst:vac:2025q2-libp2p-evaluation
Description
Test libp2p on each new version or requested feature and look for regressions, learn scaling properties and run scaling studies, understand the limits of Waku and its behaviour. Deliver reports and actionable insights. Do this monthly, reliably, with documentation of findings.
The scope of this commitment depends on the P2P team work and improvements, and it is subjected to change.
Background
We want to learn specific, actionable information about libp2p’s behaviour and how it is evolving over time with each new release and with each thing we are specifically asked to check and test.
We will use a combination of real world testing, theoretical analysis and simulation to determine and measure the success, side effects and other factors of libp2p and its evolution.
Narrative
We will support the Conduit of Expertise narrative directly by analysing and evaluating new libp2p releases and their features, both with regards to features they have today and with regards to how that compares to past behaviour.
Additionally, these efforts will contribute to the Premier Research destination narrative by improving and strengthening our relationship with the libp2p team and thus increasing the reach and influence of the IFT, and improving the chances that we successfully grow our ecosystem’s products and collaborations and especially those we want to work with externally.
Additional info
Task list
Regression testing (recurring)
- fully qualified name:
vac:dst:vac:2025q2-libp2p-evaluation:regression-testing - owner: Alberto
- status: recurring
- start-date: 2025-04-01
- end-date: 2025-06-30
Description
Run different scenarios and collect evidence and data of libp2p’s behaviour.
Test for known regressions that have occurred in the past and ensure they don’t happen again.
Deliverables
- Analysis done
- Report published with all relevant details
- RFC’s GitHub issue updated with links to the analysis and results.
Mix protocol analysis
- fully qualified name:
vac:dst:vac:2025q2-libp2p-evaluation:mix-analysis - owner: Alberto
- status: 100%
- start-date: 2025-05-12
- end-date: 2025-05-16
Description
Make use of mix protocol in DST experiments. Make use of 500~ hundreds of nodes, where some (10~) of them are using mix protocol. Study it’s behavior, as in message reliability is consistent, how much latency mix is adding in the network, calculate how much time a message takes to traverse te mixnet, and compare same scenario with and without using mix.
Deliverables
- Analysis done:
Mix-gossipsub investigation
- fully qualified name:
vac:dst:vac:2025q2-libp2p-evaluation:mix-gossipsub - owner: Alberto
- status: 100%
- start-date: 2025-06-02
- end-date: 2025-06-13
Description
Investigate mix behavior with gosspsipsub. Previous results shown that gossipsub instance within a node might not be getting triggered when a message takes the exit route in the mix protocol. Detect if this is an error from the analysis, or provide accurate information as in the gossipsub instance is handling the message as it should.
Deliverables
- Analysis done:
- PRs:
IDontWant statistical analysis
- fully qualified name:
vac:dst:vac:2025q2-libp2p-evaluation:idontwant-statistical-analysis - owner: Pearson
- status: 90%
- start-date: 2025-06-09
- end-date: 2025-06-20
Description
The aim of this task is to model the impact of IDontWant
control messages in the context of Waku scalability research,
as detailed in the following link: Waku Scalability Research.
The first step is to integrate the influence of these control
messages into the model provided in the reference,
simplifying where necessary. Any simplifications should be clearly
explained and justified to ensure a proper understanding of the trade-offs involved.
The focus then shifts to determining the overhead imposed by IDontWant control
messages on the network and subtracting these costs from the total bandwidth
usage to quantify their net benefits in terms of traffic reduction.
In this phase, we can assume that all messages are small and later analyzing
scenarios assuming all messages are large. Latency effects also need to be addressed,
particularly the case where multiple control messages arrive at varying times.
To start, the model should simulate situations where three messages are received at once,
while two additional messages arrive later and are discarded by gossipsub
due to their lateness. A refined approach must consider how to reduce these
losses, potentially by introducing a probability distribution to predict
late arrivals and better handle them in the network.
Deliverables
- PRs:
- Documents: