Lyle,
Sorry, kind of a long-winded answer.
Multiple, separate logical volumes are a recommended I/O segregation best practice for any database server (such as Exchange, SQL, Oracle, DB2, etc.). Various "types" of I/O traffic (sequential versus random) benefit system performance by being segregated. An easy way to do this is separate physical volumes (arrays) for the different I/O types. BTW, each volume should be an FT array to support your business continuity plans. Imagine one array for OS + pagefile + application binaries (all random I/O), a second array for the database application's transaction/indexing logs (sequential I/O), and a third array for the databases themselves (random I/O again). This indicates the "three array" cfg. There are performance and other reasons why it is undesirable to blend the random I/O of the binaries and the databases together. The four disk (2 x RAID1 array) cfg is not well suited to hosting a heavily loaded Exchange instance or any dB app. We prescribe the four disk cfg as the Unity server as its co-homed instance of SQL2k is not under significant I/O duress. BTW, as a further improvement for I/O segregation, there is also the concept of "splitting" a storage backplane to use multiple I/O channels to chain your arrays, but not possible with your PE2600.
Every dB SW and HW vendor writes white papers on storage best practices, so there is ample study material out there.
Dell has historically been a bit less intuitive with their hot swap disk implementation. It has been a while for me, but I do recall that you needed to use their OpenManager Array Manger to take a disk offline prior to hot-removal.
VK
