Hyper Replica

Hyper Replica solves a specific problem: keeping dataset copies in sync across geographically distributed edge nodes where the WAN link is unreliable and latency-sensitive.

The Core Problem

Traditional replication approaches assume a stable, high-bandwidth connection. Edge environments don't have that. A replication protocol that works perfectly in a data centre will stall, buffer, and eventually corrupt when the underlying transport is a flaky 4G link with 200ms RTT spikes.

The Protocol

Hyper Replica uses a delta-based sync protocol over a persistent WebSocket connection. Each edge node maintains a vector clock. On reconnect after a partition, the nodes exchange vector clocks and compute the minimal changeset required to reach convergence — no full resync needed.

type VectorClock map[NodeID]uint64

func (local VectorClock) Delta(remote VectorClock) []Event {
    var missing []Event
    for nodeID, remoteSeq := range remote {
        if local[nodeID] < remoteSeq {
            missing = append(missing, eventLog.Since(nodeID, local[nodeID]))
        }
    }
    return missing
}

Results

In a 3-node test across regions (London, São Paulo, Singapore), 95th-percentile replication latency was 38ms under normal conditions. After a simulated 10-minute partition, full convergence was achieved in under 2 seconds.

Status

In active development. The core protocol is stable; the backplane management UI and operational tooling are ongoing work.