A big growth area in recent times has been in the use of Storage Area Networks (SAN) and Network Attached Storage (NAS). Generally speaking, a SAN uses Fibre Channel as the communication protocol for provisioning of storage. A NAS usually uses TCP/IP as the communication protocol; storage provisioning can be done by using NFS or other file sharing applications. There are a couple of variants, namely Fibre Channel over IP (FCIP) and Fibre Channel over Ethernet (FCoE). FCIP and FCoE are not widely used at present; usually their use is confined to inter site links. This post looks at some of the differences between Fibre Channel and Ethernet. Before beginning, a definition: when Ethernet storage is mentioned, it means storage that uses Ethernet at layer 2, IP at layer 3 and TCP at layer 4.
Fibre Channel is often compared favorably to Ethernet in terms of being a lossless protocol. Frames sent must be received in order and are not (as is sometimes the case with Ethernet) dropped by the switch. Network storage that uses TCP has a built in error-correction facility, as TCP guarantees in-order delivery of packets. What TCP cannot do is guarantee against re-transmissions when frames are dropped by switches. However, it should be borne in mind that TCP was designed at a time when networks were less robust than today. It was expected that networks would become congested and overload switches, causing packet drops. Ethernet networks these days are far more reliable.
SpeedProbably the key advantage that FC SANs once held was the speed. This may still be one of the aspects that separates FC SANs from Ethernet ones, particularly given that a Fibre Channel SAN basically is isolated from the Ethernet. However, one should also take a look into what the future holds. The respective standards bodies (www.FibreChannel.org and www.ieee802.org/3) give us an idea of where each is at in terms of line speed. For Ethernet, we can see that 10GE is here now, as is 40GE. 100GE is right behind. For Fibre Channel, the picture is a little slower. 16GFC is around, though 8GFC is more likely to be found. 32GFC is some way off; likely to be around 2014.
The idea of "LUN masking" exists in both Ethernet and Fibre Channel networks. However, there is an additional layer of access control in Fibre Channel at the switch level. This is known as "zoning". Zoning associates the machine addresses of a server with the storage device. It acts as an additional security mechanism. No such mechanism exists in Ethernet storage, although you could argue that a VLAN performs a similar role.
Cost considerations are another factor. A Fibre Channel SAN involves putting in Fibre Channel (or FCoE) switches, as well as storage. A NAS is probably a little bit less; even if you procure new switches, Ethernet switches are usually much cheaper than Fibre Channel switches. The other aspect of this is trained staff. Getting trained staff for Fibre Channel networks is, of course, possible, but if you have a network admin already, then they will know enough to integrate the NAS into the network and set up a VLAN. The allocation of LUNs is a relatively simple process on most Network Attached Storage devices.
Size and cost are deciding factors
Ethernet storage is a better fit for small to medium businesses, as it offers the robustness of Fibre Channel, but doesn't require extra switches and specialized cabling. The cost of deploying a NAS is therefore less. For the enterprise, the days of Ethernet storage are probably some way off; Fibre Channel is still firmly entrenched. But with increasing virtualization, Ethernet storage is likely to become more attractive, particularly for new roll outs.
The differences between Fibre Channel- and Ethernet-based storage are significant. Whilst Fibre Channel is a specific storage protocol, Ethernet is a LAN protocol. Ethernet has undergone many changes since its inception, and combined with TCP/IP can be utilized successfully for NAS. Both types can fulfill a business need, and will probably exist side by side for some time yet.
Scott Reeves has worked for Hewlett Packard on HP-UX servers and SANs, and has worked in similar areas in the past at IBM. Currently he works as an independent IT consultant, specializing in Wi-Fi networks and SANs.