This page lists the best practices for block-level replication:
Block-level Replication source cache is created in the job results directory. To have an optimal replication performance with high I/O rate, move the job results folder to a different SSD on the source server.
To avoid replication lag, make sure that you have a dedicated network link with good bandwidth between the source server and the destination server.
For optimal replication and permanent boot performance, configure granular replication and RP store on a separate disk and a destination server. Configuring the RP store on an SSD disk will improve performance.
For optimal granular replication performance, set the minimal ACRP interval to 1 hour.
Creation of frequent RPs such as specifying RP interval in seconds with larger RP retentions (such as days or weeks) might degrade RP store performance.
Pseudo off-peak time might degrade RP store performance.
Make sure that the RP intervals, retention, and off-peak time are aligned so that the RP store is able to merge and produce more coarse RPs with time.
Overprovisioning RP store space might lead to performance degradation and failure during the merge operation.
Merging RPs requires space. Make sure that you provision enough space or RP store.
If you run a resync operation manually, then all existing RPs will become invalid.
To optimally use the memory on the destination server, granular fan_in configuration using the same destination server must be shared with the local RP store.
During an active replication operation, the destination volume explorer view is not refreshed with the current data. To see the refreshed view of the destination volume, suspend the replication operation, or you can create a replica copy without suspending the replication operation.