Geographic Consensus Layer: A Hybrid Consistency Approach for Distributed NoSQL Systems

Abstract

The Geographic Consensus Layer (GCL) presents a novel hybrid consistency architecture for distributed NoSQL database systems, specifically addressing the challenges of achieving strong consistency in geo-distributed Apache Cassandra deployments. By introducing a decoupled control plane implementing Multi-Paxos across regions, GCL provides linearizable consistency guarantees exclusively for critical operations while preserving high performance for non-critical workloads. This article effectively isolates the unavoidable latency penalties of cross-region consensus to only those operations that genuinely require strong consistency. Experimental evaluation demonstrates that GCL maintains near-baseline throughput while ensuring linearizability where needed. Architecture includes metadata-cavalry unanimous, operation batching, and pipeline processing, such as adaptation to reduce overheads. The GCL represents a practical solution for the fundamental trade-bands imposed by the CAP theorem, enabling the organizations to deploy globally distributed globally without renouncing continuity for wider operations or comprehensive systems.