the article mentions sun's cheaper offering as well as ibm's 'data balancer' but neglects to explain that the zfs filesystem in solaris does exactly what the data balancer does: use the sdd tas a clever cache by moving frequently used datasets onto it.

check the benchmarks, zfs is pretty good at this and does it automatically already. its also waaaaay cheaper than aix and let's you use commodity sdd's.

and then there is dtrace...

correctly there?

So, when a cell fails this will move the data to another cell?

Wear levelling is more about attempting to keep the erase/write cycles the same across cells, from what I understood.

But this is an interesting twist, this approach will mean the cells at the front are more often used and you will more quickly reduce the available space on the drive.