FCIP write acceleration works by spoofing the SCSI XFER_RDY command at the FCIP gateway. The local gateway “reads” all SCSI commands sent by the initiator and returns an XFER_RDY as soon as it sees a SCSI WRITE command. The result is that the initiator can begin sending data almost immediately. If the remote gateway receives the data from the initiator before it receives the XFER_READY from the target, the remote gateway must buffer the data until it does receive the XFER_READY.
This has the most impact with applications that can only allow a single outstanding I/O (such as some tape backup devices), because those applications would normally have to wait for two round trips between each write, thus wasting significant link bandwidth.