The debate over which protocol, iSCSI or Fibre Channel over Ethernet, will win the standards war may not actually be a war at all, but rather a question of semantics. "With fibre channel, what we were doing was taking SCSI packets and encapsulating them," says Dimension Data's Ronnie Aliti. "Now we're taking those same SCSI packets and encapsulating them in IP. So when you go back to basics that's all we're doing. It's just a transport medium and if we look at the reason why we have fibre channel to start with, fibre channel was certainly a faster and more reliable connectivity mechanism. If we go right back to basics, fibre channel came about because we hit the limitations of SCSI."
SCSI had distance limits and also limits on the number of attached devices which were causing storage headaches before the need for increased speed became paramount. At the time SCSI limits became a serious problem, Ethernet was only offering 10Mbps with 100Mbps emerging, and switched Ethernet was the new kid on the block. "Enter fibre channel where we then had the ability to put distance between our storage and our servers and to do so in a very robust fashion," says Altit. "That's when SANs really came into their own."
When 1Gbps Ethernet emerged from the labs and into real networks there was no longer a speed barrier. "Companies such as Network Appliance jumped on that and brought out iSCSI connectivity for their switches and now we're seeing every vendor doing the same," says Altit. "EMC is now doing it in a very big way, pushing IP storage, and they have several SAN offerings that are IP-based. And certainly we're seeing that as something that all of our customers should be looking at and should be considering. When future-proofing an environment, we think it's very important to ensure that your network has that capability over the three year average life span of that type of equipment. That is, that it has the ability to bring IP-based storage into the network.
Of course, adding any new application to your network requires you to allow for the usual caveats such as knowing your existing traffic mix and how much bandwidth is actually available, but the future certainly looks like being IP-based Ethernet for just about everything. "We're encouraging our customers to look to spend less money on their switching environment at a fibre channel level and spend that saved money on some more intelligence in storage software," Altit says. "That way we can be sure they're managing their environment in a far simpler and more cost effective manner."
"Smaller organisations don't necessarily understand fibre channel but they certainly understand Ethernet switching," Altit adds. "When implementing a fibre channel network, often people need to be trained and they need to understand how fibre channel works. Whereas if you were to go straight to an IP-based storage medium, there is no training because it's just simply connecting another box into the LAN. They're doing it already to get their workstations talking to their servers. Now all we're doing is making a workstation into a stack of disks."
Indeed, when you mention fibre channel, you start to talk big scary interface connection dollars, scary fibre optic cable dollars and usually a scary consulting fee to find somebody who knows how to connect it and make it all work. "That's certainly the way people look at it, but in reality fibre channel is not really that complicated," says Altit. "But it's often the cost that frightens people. If we can address cost, if we can address simplicity, manageability all at once while still providing the service levels that a business requires, then I think it's a win all around."