The big problem with having only copper Fibre Channel connectors in the back of the Xserve RAID is that copper cabling must be used to hook up the unit to a Fibre Channel switch. As a result, the Xserve RAID unit cannot be placed very far from the switch, which could be a problem for sites that want to centralize storage on a site away from where servers are located.
The Xserve RAID has 14 drive bays controlled by two RAID controllers (each controller handles seven drives).
To create RAID sets with more than seven drives, we had to use software RAID to bind the two RAID sets together. Under this scheme, an IT manager who needed a 14-disk RAID 0 set would have to construct two seven-disk RAID 0 sets and bind them together with software RAID. Considering that this RAID system was designed to work with the relatively small, two-processor Xserve, having two separate controllers could be a more flexible option than having two clustered controllers in an active/active setting. IT managers could easily hook two servers to it (one on each controller) without having to buy a separate Fibre Channel switch to hook everything together.
The software management tools were fairly easy to use and provided a host of diagnostic information that could be used to manage multiple Xserve RAID units.
We liked the Xserve RAIDs management tools. The tools can be run on Windows and Unix machines using a Java Archive file. Because the management tools can be used on non-Apple machines, and because the Xserve RAID unit looks like any other Fibre Channel RAID unit (from a hardware perspective), the barriers to using Xserve RAID on other platforms are largely political, not technological. This means the Xserve RAID unit might be made available to other platforms, in much the same way that Apples highly successful iPod MP3 player eventually became available for Windows.
The Xserve RAID has redundant, hot-swappable power supplies and cooling modules for fault tolerance. The RAID controller modules can be replaced without rebooting, when no transactions are running. This means that IT staff can fix one side of the RAID without having to interrupt transactions on the other side with a reboot.
Senior Analyst Henry Baltazar can be reached at email@example.com.