The linear topology is similar to the ring topology, but a loop through all the replicas is not completed. That is, the first and last replicas do not synchronize with each other.
The linear topology is simple to implement, but it is usually not as efficient as some of the other topologies. The main problem with the linear topology is the long latency time for data dispersion.
To illustrate this, consider four replicas, A, B ,C, and D, that are synchronized by using a linear scheme, with A as their controlling replica. If a change is made to data in replica D, but to none of the other replicas, it would take six synchronizations, starting at A, before all other replicas have D’s changes. These synchronizations are: A-B, B-C, C-D, A-B, B-C, A-B. If this were set up as a ring topology, A and D would be directly connected, so it would take only four synchronizations for D’s changes to be fully dispersed: A-B, B-C, C-D, D-A.
When designing a topology for your application, it is a good idea to count the number of synchronizations that must occur in different situations to fully disperse data throughout the replica set. Although the number of synchronizations is not the only measure of a topology’s effectiveness, it can tell you something about the efficiency of your replication architecture.