Don't give me any of that BS, they'll both go nuts if the write back cache
is not pushed to the disk. That sparc5 is a primary news server. It has
4*4Gig FWD drives attached to it. On many occassions (more than I care
to count) it has crashed and taken hours (and once three days) to make the
filesystem(s) stable. The three day down time was the result of a crash
during the fastrm cycle of expire -- several 100,000 files now half deleted.
And, I have done your power down test on a caching raid system (linux). With
30+Meg of filesystem data not physically pushed to the disks, e2fsck repaired
the damage in about 6hours after 15 iterations.
But, you are right, it is an unfair comparision as Solaris is a slow piece of
crap. It has been from the get-go... until 2.5, is was laughably slow.
--Ricky