What Is Snowflake Consensus?

The most basic member of the Snow family is Slush: the node takes a random sample and switches to the majority's preference, but this method is not resistant to malicious participants. Snowflake adds a counter to this method. In each round, the node queries a certain number of random validators; if a sufficient majority of them support the same option, the query is successful. If this option is the node's current preference, the counter increases by one; if it is different, the node changes its preference and the counter resets. When the counter reaches a predetermined threshold, the decision is finalized.

This structure provides resilience against Byzantine faults, that is, against malicious or faulty nodes: it is extremely unlikely that an attacker controlling a small minority of participants could mislead the network over many consecutive rounds. Because each node talks only to a small sample, the message load stays low and the protocol can scale with a large number of validators.

Snowflake's weak point is that the counter remembers only consecutive successes; temporary fluctuations can slow progress. Snowball solves this problem by keeping persistent confidence counters for each option. The term should not be confused with the cloud data warehouse company of the same name.