Methods, systems, and devices for data management are described. The described techniques may enable a data management system (DMS) to determine an amount of data to be pulled from a source cluster prior to performing a replication of a source file stored on the source cluster. For example, before copying any data from the source cluster to a target cluster, the DMS may read an index file associated with the source file to determine data blocks that are different from data blocks of a base file stored on the target cluster, which may indicate indices of data blocks of the source file that may be replicated from the source cluster to the target cluster as part of the replication operation. By profiling the data to be copied before copying the data to the target cluster, the DMS may optimize the replication process to reduce latency, among other examples.