A Dynamic synchronous/asynchronous replication

Synchronous replication is highly desired in todays IT organizations as any data loss may have significant monetary effect on the organization. In synchronous replication every write is acknowledged before more data is replicated. Due to the speed of light limitation on information transfer and the typical two round trip SCSI write protocol, synchronous replication is usually limited to a 50 mile distance. Beyond such a distance the performance slowdown incurred by synchronous writing becomes prohibitive. Even customers who can afford maintaining a replica within a 50 mile distance, may have significant performance issues with synchronous replication. Synchronous replication is not only limited by the speed of light, but is also limited by the WAN bandwidth between the sites and the available storage bandwidth at the replica site. For example, if the WAN bandwidth is lower than the production bandwidth, performance of volumes replicated synchronously, will be limited to the bandwidth available by the WAN. The same is true for the replica storage bandwidth. If the current throughput available for the replica storage is lower than the available production throughput, synchronously replicated applications will slow down to the performance of the replica storage which may have different hardware. For some applications synchronous replication is desired, but low latency is significantly more important. In some cases there are temporal fluctuations in the importance of data loss. For example, many enterprises run batch operations and data warehouse operations, after working hours, or during weekends. The batch operations usually consume significantly more bandwidth than standard daily workloads, but produce less critical data, i.e., data lost during a batch operation can be easily reproduced. As opposed to synchronous data replication, asynchronous replication allows the system to continue replicating data remotely even if not all previous writes have been