One of the most popular features of the Storage Networking World conferences is the SNIA interoperability lab, where visitors can see a stack of different manufacturers' storage hardware and software working together in a variety of configurations.

At the recent SNW Europe event in Frankfurt the SNIA reps took it a stage further, not only featuring a range of iSCSI kit, but also running hands-on lab sessions for those who wanted to learn how to configure and provision iSCSI storage.

Attendees got the chance to work with a variety of hardware, from companies such as Adaptec, Cisco, HP, EMC, NetApp, Reldata, Sun and Westek. They could run performance tests too, to verify that iSCSI provides reasonable performance even over Fast Ethernet.

The results were impressive, and the conclusion has to be that yes, iSCSI really works, and yes, it is a practical solution for small SANs - assuming it's not asked to run over an already-congested LAN of course.

For small offices, where the iSCSI storage would probably take the form of a single box which provides both NAS and iSCSI services, such as a NetApp FAS or an Adaptec SnapServer, creating and provisioning an iSCSI volume is not much more complex than doing the same for a direct-attached disk drive or array.

However, the demo also showed that once you get a storage network involved, as you will in a deployment of any size, you will still need to learn a bit about SANs and networked storage in general. Anyone who tells you that iSCSI means you only need to know Ethernet in order to run a SAN is being economical with the truth, to say the least.

The fact is that storage networking is different. It's no great mystery though, and the management tools that the SAN companies are coming up with now mean that it should not be a huge leap for a competent network admin. That's for simpler SANs of course - but then again, would you expect a branch office LAN admin to take over a giant network with VPNs, VoIP and QoS without extra training?

It's also true that IP is no stranger to storage networking, with iSCSI being just one of several ways that IP can be used to build or extend a SAN. Briefly, here are the notable ways of running SANs over IP:

  • FCIP (Fibre Channel-IP) is a tunnelling scheme that links two Fibre Channel SANs over IP, making it look like one big SAN. It's simple in theory but less so in practice - the two SANs may have overlapping domain IDs, say, which will need fixing, and it can cause big problems if one SAN resets, because the changes propagate over the WAN link too.

  • iFCP (internet Fibre Channel Protocol) is a routing scheme, it is more intelligent than FCIP, allowing you to choose which nodes on one SAN you expose to the other SAN. For example, the two could share a single tape library but nothing else. The only supplier supporting it is McDATA with its Eclipse router technology, which comes from its purchase of Nishan Systems.

  • iSCSI (internet SCSI) is end-to-end IP, whereas the first two are for bridging two or more Fibre Channel SANs. Although iSCSI can also bridge into a Fibre Channel SAN to access storage resources - this is how Cisco's MDS devices work - the difference is that the servers would use iSCSI, not Fibre.
  • Two related developments are the pending ANSI standard for FC-IFR (Fibre Channel inter-fabric routing) and Cisco's similar IVR (inter-VSAN routing) specification. These could potentially supplant iFCP as a way to selectively links nodes on different SANs.

    And of course, there is also the option of not using IP at all: with the right end equipment and enough buffer credits, Fibre Channel can run over thousands of kilometres at 1Mbit/s, and for at least 1000km at 2Mbit/s.