Between Fibre Channel And Ethernet
The concept behind iSCSI arose from traditional storage interfaces such as ATA or SCSI being limited to a single machine and a maximum cable length. Flexibility of business and enterprise class storage applications hence is rather limited. So, ideally, storage should be detached from particular servers in order to increase flexibility and to avoid bottlenecks. In the end, administrators should be able to add, move, backup, restore or reconfigure storage without having to change running systems or look after multiple servers. Swapping out storage by using existing network infrastructure thus seems to be a good approach.
In addition, Fibre Channel can point-to-point-connect two devices, be set up using switches or be operated in an arbitrated loop similar to Token Ring. Apart from that, it can carry ATM cells, IP packages or SCSI commands by fitting these into its own frames (which are not compatible with Ethernet). For these reasons, Fibre Channel is dominant in the enterprise space.
Fibre Channel storage components are considered the only choice for high-end environments such as data processing or huge database applications. The infrastructure can be used as an interface between hard drives and host bus adapters, or for linking systems and/or storage units within a SAN - or both. However, as soon as an enterprise does not require data to be locally available at a maximum speed, the extra costs do not make sense: Why not opt instead for storage subsystems based on Serial ATA components, and use the existing Gigabit Ethernet infrastructure?
