Shingle me timbers! Seagate brags of 1m SMR drives - where are they? Seagate has said it's shipped a million shingled disk drives to date. Shingled Magnetic Recording (SMR) increases a device's capacity by squeezing more readable tracks of data onto a platter's surface, albeit by hammering rewrite speeds. When committing bytes to disk, the tiny electromagnet in the disk head that writes the …
So, if for every band I want to write I actually read+write ten, I'm getting 10x write errors + 9 band reads that can also fail.
Quite nice for cat videos, but woudl it be even possibe to rebuild an array using these disks? Is it for more than archival/ backup to disk? Because I can't see them for "normal" use.
If they are able to sell 7 TB 31 1/2 " for archival.. well, a 100 library would be 700TB.. quite impressive for a supposedly relatively small price tag (say 40K $? )
If the chance of getting write errors is almost zero, then having to do ten writes is still going to be almost zero. That's assuming there is any significant correlation between the number of write head activations and the number of write errors - I'd suspect it is more likely to be imperfections in the platter that end up resulting in write errors and that is more likely to occur with higher density regardless of how the density is achieved.
I don't know if they have implemented it but if the drive is using TRIM to keep track of which tracks actually have data then potentially it doesn't need to rewrite all the other tracks if they are empty anyway. This would make it particularly useful for set top boxes which tend to record/delete in large contiguous blocks, or for array rebuilding onto a fresh drive without slowdown on the initial write.
...although that could mean shipped into a warehouse, like Windows RT tablets.
Seems like write-once-update-never applications would be biggest benefit - like Facebook photo albums - but these would be great for streaming backups too as well.
If they were really clever, they'd keep (say) the outer 10% of the disk for non-overlapped use, and migrate hot data into that area.
"Seagate" and "reliability" together is an oxymoron, and I've a box of dozens of dead drives, mostly Seagate, to prove it.
In my post to the El Reg story "Do you think spinning rust eats flash's dust? Join the hard drive daddies club" of 14 August, I gave good reasons for it (and I suggested what Seagate should do about the problem to improve customer confidence):
http://forums.theregister.co.uk/forum/1/2013/08/14/keep_the_platters_spinning_club_formed/#c_1925303
In this announcement, not one of the issues that I raised in that August post has been addressed by Seagate, yet its hard disk densities are still creeping up and up (and seemingly the reliability still creeping down and down).
Yes, so it's now possible to pack densities of >1TB onto a singe platter—big deal! However, it's of very little practical use if at whim one's data just wafts off into the aether shortly later. Absolutely nothing in this story convinces me that Seagate has made any of these drives more reliable, and it's time we unfortunate consumers started whingeing loudly about the problem.
It's all very well having a track up Everest, albeit comparatively well worn since 1953, but it's not a path the average punter strolls along on a Sunday afternoon.
Announcements from Seagate about 'Everest'-type hard disk densities might well be headline-grabbing, but in the light of its very questionable past record with respect to the reliability of its hard disks, I wouldn't touch these new disks from Seagate with a barge pole.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
