In the event that there is intermittent connectivity between web applications, then adaptive mode can help by finding the best available path to transfer packages in a replication topology that includes more than two farms. With direct mode, replication will not continue until the connection is re-established.
Adaptive mode can be faster than direct mode by better distributing activity on the network. Furthermore, when implemented across multiple farms, adaptive mode can use less system resources for creating replication packages, since packages are reused instead of recreated at each step of a multi-hop replication. However, Direct mode can send fewer notifications to the target farms than adaptive mode because there are fewer paths in the replication network.
When using Differencing Compression for a connection, Adaptive mode caches packages on the source and target to anticipate the changes that need to be replicated and therefore further reduces the bandwidth used to replicate documents.
Both queued and captured events are set aside in the same space, awaiting replication. Depending on the replication schedule defined in the map family, Replicator looks for events in the waiting area and starts to process them into packages.
In the diagram above the source farm has had several events manually queued for replication by an administrator, and has also automatically captured several events for replication based on changes made on the web site level by an end user. Both types of events have been placed in the same waiting area in the replication pipeline. Both type of events will be packaged, transported and then received on the target where they are applied.