Western Digital subsidiary HGST has announced the Ultrastar Archive Ha10, a 10 terabyte helium-filled shingled magnetic recording (SMR) hard disk drive. It rotates at 7,200 RPM and has a 256MB cache.
HGST has also released libzbc, "a simple library providing functions for manipulating disks supporting the Zoned Block Command (ZBC) and Zoned-device ATA command set (ZAC)."
The new drive is intended for enterprise bulk storage that is infrequently accessed. SMR tracks are partially overlapped which can hurt drive performance. The Ha10 has lower sequential write speeds than the He8. Seagate has already released 8 TB SMR drives.
What's next? 12 TB? 16 TB? HAMR?
(Score: 3, Insightful) by RedBear on Thursday June 11 2015, @04:25PM
I could respond in kind and point out that boiling down everyone else's needs, past, present and future, to match your opinion of your own needs makes you a jerk, and makes everything you say worthless. But that would be a jerky thing to say.
We have nearly a quarter terabyte of personal photos already after just a few years, and we don't even take many photos. If we had been into making a lot of high-def family event videos like some people do we could have already needed that 10TB just for personal, irreplaceable data. No hoarding required. And that of course is just focusing on home users.
As I point out in the FreeNAS thread, full mirrors drastically increase the overall cost of storage, making it difficult to see that as a financially viable choice for most home users. And as a very knowledgeable user on the FreeNAS forums points out, the failure of a mirrored device that is part of a larger array creates a stress point in the array, and depending on how the array is being used the rebuild can take a lot longer than just a straight copy operation would theoretically take. Unless you have two or more mirrors of each physical device that creates a strong risk that the mirror you're rebuilding from might also fail before the rebuild completes. If those mirrors are part of a virtual device that is a member of a larger array, you would then lose the entire array. In other words, full mirrors may be part of the solution in the future to making data live forever, but it is far from being the full solution.
Part of using a multi-disk parity scheme like ZFS's RAID-Z2/Z3 is gaining the ability to reconstruct corrupted blocks of data from multiple other verifiable blocks that contain an uncorrupted version of that data and can be compared against each other. This aspect of data protection is not present in a simple RAID1 mirror, so there are many situations where mirrors cannot be the final word. Using a heterogenous disk environment (different brands and manufacturing batches) should definitely be a requirement in any professional setting. And yes, RAID of any level is definitely not the same as a backup.
¯\_ʕ◔.◔ʔ_/¯ LOL. I dunno. I'm just a bear.
... Peace out. Got bear stuff to do. 彡ʕ⌐■.■ʔ
(Score: 2) by frojack on Friday June 12 2015, @06:28PM
Old school thinking still prevails I see.
The drastic increase in cost using raid1 is proving to be a myth as costs fall dramatically year over year.
The theory of increased costs with mirroring was wrong 20 years ago, it was more wrong 10 years ago, its even more wrong today, and fixing to become utterly and irrevocably wrong in the future. If a theory has been proven wrong for its entire life, don't you think its time to get a new theory?
Rebuild time with a damaged large raids is cost prohibitive. The cost of resources, computing, human, and hardware, to fix a corrupted raid exceeds by several orders of magnitude the cost of simply buying twice the storage.
Raid1 is the least expensive, even if you have to use 3 drives in a mirror.
No, you are mistaken. I've always had this sig.