NodeSync Architecture

This page describes the internal design and architecture of the NodeSync distributed key-value store.

System Model

Client
  |
  v
Follower Node  --->  Leader Node
  |                    |
  +----- Replication --+

NodeSync follows a leader-based replication model. All write requests are coordinated by the elected leader.

Leader Election

• Nodes exchange periodic heartbeats
• Failures are detected via heartbeat timeouts
• The alive node with the highest ID becomes leader

Write Path

1. Client sends SET request
2. Follower forwards request to leader
3. Leader applies write locally
4. Leader replicates write to peers

Consistency

• Eventual: asynchronous replication
• Strong: quorum-based acknowledgments

Failure Handling

• Failed nodes are marked DOWN
• Leader failure triggers re-election
• System remains available if quorum exists

← Back to NodeSync