Cross-Site Storage Replication for Read-Write Order Consistency

The patent introduces a mediation mechanism through the primary storage site that coordinates write operations between secondary storage sites. When a write operation is received at the primary site, it is replicated to secondary sites, and the primary site mediates the consistency by tracking replication status and managing re-synchronization. This intermediary approach maintains read-write consistency without requiring complex peer-to-peer coordination between all storage sites.

Instead of having each storage site independently manage its own consistency (which would increase complexity), the patent inverts the approach by designating the primary storage site as the consistency manager. The primary site actively tracks and manages the replication state of all secondary sites, reversing the traditional model where each node is autonomous. This inversion simplifies the overall system architecture while maintaining consistency.

The patent implements partial synchronous replication by differentiating between critical data that requires immediate replication (zero RPO) and non-critical data that can tolerate asynchronous replication. The system identifies and prioritizes critical data blocks for synchronous replication while allowing other data to be replicated asynchronously, thus maintaining data consistency for essential operations without the performance penalty of synchronizing all data.

The system dynamically adjusts replication parameters based on data criticality and system load. For non-critical data or during high-load periods, the replication mode can shift from synchronous to asynchronous, changing the RPO parameter temporarily. This parameter flexibility allows the system to maintain zero RPO when necessary while improving performance when data consistency can be relaxed, optimizing the trade-off between consistency and productivity.

The patent implements preliminary actions by establishing a grace period or threshold mechanism before declaring a storage site as failed. Instead of immediately triggering failure protocols upon missing a single heartbeat, the system waits for multiple consecutive missed heartbeats or extends the monitoring window. This preliminary action filters out transient network issues that would otherwise be misinterpreted as actual failures, reducing false positives while maintaining reliable failure detection.

The system applies preliminary anti-action by implementing counter-measures against false failure detection. When network partition conditions are detected (such as asymmetric connectivity patterns), the system preemptively adjusts heartbeat evaluation criteria or suppresses automatic failure declarations. This preliminary anti-action prevents the harmful effect of false positives before they can trigger incorrect failover or data loss protocols.