Clear DefinitionsSnowball Consensus
1 min read
Pronunciation
[snoh-bawl kon-sen-suhs]
Analogy
Imagine a large group of people trying to decide between two colors, red or blue. Each person initially has a preference. Periodically, everyone asks a few random friends what color they prefer. If a person hears most of their queried friends prefer red, they are more likely to also switch their preference to red. This process repeats, and eventually, due to this repeated sampling and influencing, the entire group rapidly converges on one color. This 'snowballing' effect leads to consensus.
Definition
A family of probabilistic, leaderless Byzantine Fault Tolerance (BFT) consensus protocols used by the Avalanche network. It works by having validators repeatedly query a small, random subset of other validators about their preferred state or transaction, and then incrementally adopting the preference of the majority they observe.
Key Points Intro
Key Points
Used by the Avalanche blockchain platform.
A probabilistic BFT consensus protocol.
Leaderless: no single validator is responsible for proposing blocks at any given time.
Validators repeatedly poll a small, random sample of other validators for their preferred decision.
Decisions gain confidence with each successful round of polling the same preference, eventually becoming accepted.
Designed for high scalability, quick transaction finality (typically seconds), and robustness.
Example
In Avalanche, when a transaction is broadcast, validators using the Snowball protocol query other validators. If a validator sees that a supermajority of its queried peers favor including the transaction, its confidence in that decision increases. After enough rounds of consistent majority responses, the transaction is considered finalized.
Technical Deep Dive
The Snowball protocol has parameters like k (sample size), α (quorum size, i.e., how many sampled nodes must agree to sway opinion), and β (decision threshold, i.e., how many consecutive successful polls are needed to finalize a decision). A node maintains a confidence counter for its current preference. If α out of k sampled nodes agree with its preference, its confidence increases. If they disagree, it may flip its preference. This repeated sampling process leads to rapid convergence towards a single decision across the network with high probability, even in the presence of Byzantine validators (up to a certain threshold).
Security Warning
The security of Snowball relies on the randomness of sampling and the assumption that the network is sufficiently well-connected for queries to propagate. The specific parameters (k, α, β) must be carefully chosen to balance safety, liveness, and performance.
Snowball Consensus - Related Articles
No related articles for this term.