I'm more curious as to why a hosting service isn't using a hypercore type system which pretty much negates a *single* drive failure taking down a site. At the very least running on a RAID configuration that would take such failure into account and allow services to continue while a replacement...