Xiotech's CorteX to change the storage world Xiotech wants CorteX to be the door to a new storage world. CorteX is a REST-using API that lets an application access and manage Xiotech's storage brick, its ISE (Intelligent Stage Element), the sort of super disk drive with fail-in-place components, a five-year warranty and terrific performance from its multiple spindles. …
This reads horribly like a fan piece with a regurgitation of a manufacturers marketing speak.
I've seen far, far too many magical sounding storage solutions to believe this sort of stuff. Unsourced annecdotes like a customer replacing 10,000 enterprise drives per year is not helpful. What on earth is the context? How many drives does the customer have? If they have a few hundred thousand drives and they are mostly pre-emptively replaced under a service contract and without disruption, where's the big deal? Sealed units with a 5 year guarantee where the whole thing has to be replaced in one go - is that such a great idea? Let's see what the real effect is on a business and some evidence to back it up. As iot is, this is just an annecdote.
No - at the bottom end of this, all storage manufacturers use the same limited set of physical drives from the same limited set of manufacturers. They are going to all fail at much the same rate from any given batch from a given manufacturer (assuming the storage array manufacturer does the engineering job properly). You can stuff a bit more redundancy at the front end and not bother replacing failed drives, but sooner or later you are going to have to replace something or throw the whol lot away.
Let's see some realistic benchmarks from this, some real costing exercises on comparable configuration and then, maybe, we might have something worthwhile to discuss. Let's see what the latency is, what the performance numbers are. How about some discussion of relative strenghts and weakness.
For Steven -
The xiotech self healing stuff is somewhat unique so I'm not surprised you question the stats. They integrate(currently) only with Seagate drives. It's fairly well known that most drives that "fail" do so without suffering an actual failure. Xiotech, because of their close relationship with Seagate(they were spun out from them a few years ago) have the ability to run the Seagate RMA scans/repairs on the disks in the array, the same sorts of things the support department would run if you returned the disk.
On top of that they have a very innovative method of failing portions of the drive, if enough errors pop up they stop using the head associated with those errors, the rest of the drive is still OK to be used so the system can continue to operate. They do their RAID at the disk head level rather than the whole disk itself.
That said, for my own storage I don't see disk failures as a big deal, the array comes with 3 years of hardware support up front, cost is not big, and we've had one disk failure in the past 1.5 years (the disk was less than 4 weeks old, so I think it's just a luck of the draw there). More than 200 disks in the array.
As for TP not being needed in the controllers I don't agree myself, while TP is certainly nice to have in the apps, if your dealing with a shared storage system with many apps running on it simultaneously it will be advantageous to have a common method of thin provisioning - at the array level. If ALL you do on your array is VMFS volumes then maybe you can get away from array-based TP but I'd suspect the situations where that is the case is in the minority.
There are also next generation TP technologies involving storage reclamation as well, this more or less requires TP at the array level because the array has to be able to reclaim the space and make it available to other volumes.
Despite having TP available in vSphere the only place I ever specifically enable it is on systems that use local storage(so don't have TP natively). I detail out my VMware TP strategy here, which is somewhat unique in that I've never seen anyone else post on a similar strategy, it has worked well for me over the past few years with ESX -
http://www.techopsguys.com/2009/11/06/thin-provisioning-strategy-with-vmware/
Don't be fooled by false claims from Xiotech. Their technology isn't all that spectacular. All of the big storage array vendors do similar things. That's why they have their own disk drive firmware - it allows them to do very close monitoring, fail drives proactively and copy most of the data even from failed drives instead of going through a RAID rebuild. And guess what, the only reason they can develop their own drive firmware is a very close relationship with the drive vendors. Xiotech doesn't get any special treatment from Seagate (why should they, other companies buy a lot more disks). Telemetry/phone home features are quite common as well. And other storage vendors even get the failed disks back and analyze them in order to improve their monitoring algorithms. Can't do that with a sealed brick that is never returned...
Xiotech is a dumb storage brick and all they do has already been done by other storage vendors for years. They just don't talk about it that much as they have other features as well...
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
