A leader-follower topology gives one node special responsibilities and makes other nodes follow its coordination. Systems using this design elect a replacement leader when the current leader fails, allowing operations to continue. section 1.2.1
ScyllaDB takes a different approach: every node has equal status. A request can contact any node, which calculates the owning node and forwards the request without relying on a centralized leader. section 1.2.1
The topology implies:
- Leader failure triggers leader election. section 1.2.1
- Coordination responsibilities concentrate in the designated leader. section 1.2.1
- ScyllaDB instead distributes ownership and replicates rows across nodes. section 1.2.1
