Partible State Replication for Industrial Controller Redundancy

Distributed control systems are part of the often invisible backbone of modern society that provides utility services like water and electricity. Their uninterrupted operation is vital, and unplanned stops due to failure can be expensive. Critical devices, like controllers, are often duplicated to minimize the service stop probability, with a secondary controller acting as a backup to the primary. A seamless takeover requires that the backup has the primary's latest state, i.e., the primary has to replicate its state to the backup. While this method ensures high availability, it can be costly due to hardware doubling. This work proposes a state replication solution that doesn't require the backup to store the primary state, separating state storage from the backup function. Our replication approach allows for more flexible controller redundancy deployments since one controller can be a backup for multiple primaries without being saturated by state replication data. Our main contribution is the partible state replication approach, realized with a distributed architecture utilizing a consensus algorithm. A partial connectivity-tolerant consensus algorithm is also an additional contribution.