Hi Stephen, Good article on RAID. My comment: not so fast. RAID5/6 implementations based on Patterson et.al. are still being heavily purchased and deployed in the industry. One reason: the mathematical lookup of data, as described by Patterson, is not only fast, but more importantly, it's trusted. Customers are cognizant of the value of this direct mapping. Virtualizing the location of customer data has its place (e.g. enabling snaps), but mathematical lookup will continue to be a valuable role at the very bottom of the stack. Keep up the interesting posts, Steve
sfoskett
· 1 year ago
Steve,
Great point! I'm very much a pragmatist, as are most successful storage managers. They'll continue to use what works long into the future, even as new ideas come and go. But let me be clear - this is the end for RAID tied to specific whole disks. We'll continue to see RAID 5 and RAID 6 (and RAID 1) based on subdisks (as in DMX et al) but devices like the CLARiiON will give up their strict disk-centric model. In my opinion, of course!
All Comments
Good article on RAID. My comment: not so fast. RAID5/6 implementations based on Patterson et.al. are still being heavily purchased and deployed in the industry. One reason: the mathematical lookup of data, as described by Patterson, is not only fast, but more importantly, it's trusted. Customers are cognizant of the value of this direct mapping. Virtualizing the location of customer data has its place (e.g. enabling snaps), but mathematical lookup will continue to be a valuable role at the very bottom of the stack.
Keep up the interesting posts,
Steve
Great point! I'm very much a pragmatist, as are most successful storage managers. They'll continue to use what works long into the future, even as new ideas come and go. But let me be clear - this is the end for RAID tied to specific whole disks. We'll continue to see RAID 5 and RAID 6 (and RAID 1) based on subdisks (as in DMX et al) but devices like the CLARiiON will give up their strict disk-centric model. In my opinion, of course!
Stephen