System Architecture

Index0 core systems are designed for high-throughput, low-latency consensus in partition-tolerant networks.

The Event Loop

At the heart of the system is a single-threaded event loop based on the reactor pattern. This ensures deterministic execution of state transitions, which is critical for verifying correctness.

Note: We strictly avoid multi-threaded logic in the core consensus module to prevent race conditions. Parallelism is handled at the network I/O layer.

// Rust pseudo-code for the Event Loop
impl Node {
    fn run(&mut self) {
        loop {
            match self.rx.recv() {
                Ok(Event::Message(msg)) => self.handle_message(msg),
                Ok(Event::Tick) => self.handle_tick(),
                Err(_) => break,
            }
        }
    }
}

Consensus Mechanism

We utilize a variant of Raft optimized for geographically distributed clusters. Standard Raft can suffer from leader election storms during high latency periods. Our implementation enables PreVote checks to stabilize leadership.

State Transitions

Nodes exist in one of three states:

Follower: Passive, responds to requests.
Candidate: Active, soliciting votes.
Leader: Dominant, appending log entries.

Data Persistence

Logs are persisted to an append-only file structure using fsync on every commit to ensure durability (ACID compliance).