A Replication has two IO Modes: Synchronous and Asynchronous.
Synchronous Replication mode means every write IO to a replication volume waits for the IO to complete. Data stored locally and the data stored remotely is to be synchronized. The Server issuing the write IO is held until the IO is completed. If the link between the local and remote data store is slow, the performance of the Server application may be slowed by the slow writes.
Synchronous Replication mode is useful for the applications that cannot tolerate missed writes of data, or for loading/storing data to other (non-Compellent) storage devices. In this mode, the External disk is treated much like a local disk.
Synchronous Replication mode and recovery: Synchronous Replication on a volume already having data implies two operations, copy existing data, and copy the update writes as they occur. Synchronous Replication mode only has recoverable data when all source volume data and updates have been completely copied to the remote system (Sync'ed State). There is no intermediate recovery state associated with remote data using Synchronous Replication. If access to the remote storage is lost (connection failure, remote storage failure, ...), all data must be re-copied to the remote system to re-establish synchronization between locations.
Asynchronous Replication mode can only be used between Compellent SYstems, and requires Data Instant Replay to be licensed on both Compellent Systems.
Asynchronous Replication means write IO from the Server does not wait for the replication IO to complete. The Server issuing the write is returned with a successful status after completion of the write to local storage; there is no performance penalty to the Server using Asychronous Replication.
Asynchronous Replication mode is useful for applications that can tolerate recovery where not all writes may have been recorded on the remote Compellent System. This replication mode is particularly useful for replicating and/or recovering large volumes.
Asynchronous Replication mode and recovery: The Compellent System uses the Data Instant Replay feature to create checkpoints in the data. Replays are used either as resynchronization points or as data recovery points. This aspect can be used to locally prepare storage for later re-synchronization (to avoid having to move massive amounts of data over a slow link) or to speed recovery moving the minimal amount of data back from the remote site from a known replay synchronization point.