News - TDK takes magnetic-disk densities to the extreme

[HEXUS]

Achieves densities of 1.5Tb per square inch.

Read more.

[shaithis]

It doesn't seem like a great leap tbh.

While 2TB per platter may be possible, I doubt they will go with that right away and we will see a gradual increase just like all previous platters.

1TB platters have been in production for a while now.

And while a lot of people seem to disagree, I still say 2.5" HDDs have an extremely limited future. Mostly used in laptops and servers. Laptops are increasingly becoming SSD-driven and with the latest price cuts, I can see server farms going SSD as well considering the price and performance difference between a 2.5" enterprise class mechanical drive and an SSD.

[Bunjiweb]

What about the data degradation that seems to be inherent to NAND based storage? I couldn't fill up a SSD with my precious data and expect all that data to be intact in a few years time if I haven't had the SSD powered up in the interim.... At least that's how I understand it.

HDDs have shown they can retain data for a long time with excellent integrity and so IMHO are still going to be around for a while to come yet...

[Unique]

so this means potentially 8tb 4 platter drives or even 10tb 5 platter drives?

a 2tb 2.5" drive would be great for me for my large laptop as i could have a fast SSD bootdrive and a large storage space. or likewise people who just want a small form factor HTPC etc

[technodean]

agreed, I have to back up all my data from my ssd to another hard drive as I simple don't trust it.

[Shad]

You shouldn't trust any storage medium. Flash, mechanical or anything else, they all can and will fail. Putting all your eggs into one basket is a recipe for disaster

[scaryjim]

... I can see server farms going SSD as well considering the price and performance difference between a 2.5" enterprise class mechanical drive and an SSD.

Enterprise-class SAS 2.5" drives are ~ £100 for 300GB. A 300GB Intel 710 (enterprise, MLC, SATA II SSD) is £800+. No enterprise is going to pay 8x as much for storage unless you can demonstrate an overriding business case for the higher performance storage - and when you're talking huge redundant arrays then I doubt the real-world performance increase is noticeable. Sure, comparing one SSD to one HDD it's night and day, but compare 8 of each in RAID10 on a storage subsystem and then tell me you can qualitatively tell the difference. Then ask yourself whether you'd rather use one 300GB SSD, or 8 300GB HDDs in RAID10. I know which I'd choose

[Kumagoro]

I can't see 2.5 inch drives disappearing or significantly reducing in production in comparison to3.5 inch drives. If anything I would expect the opposite. With the trend for smaller more power efficient machines I think its more likely that we will see the box shifters employing an ssd with a 2.5 hdd to save space I think. Or at least hybrid drives when they stop being ridiculously expensive compared to standard, I suppose out depend on how competitive toshiba will be with theirs.

[snootyjim]

What about the data degradation that seems to be inherent to NAND based storage? I couldn't fill up a SSD with my precious data and expect all that data to be intact in a few years time if I haven't had the SSD powered up in the interim.... At least that's how I understand it.

HDDs have shown they can retain data for a long time with excellent integrity and so IMHO are still going to be around for a while to come yet...

Come again?

In what possible way are hard drives reliable? They fail almost like clockwork, when analysed on a grand scale - remember the google survey from a few years back?

IIRC the typical scenario with NAND storage is that it degrades until it's no longer writeable, but it should still be readable. In the long-term I should think they'll be far more reliable than spindle-based storage.

[CampGareth]

Oh yay! Higher capacity drives! Now if only they were made in a 5.25" form factor so we could get more like 20TB per drive instead of the tiny numbers we see nowadays.

[jnutt]

You shouldn't trust any storage medium. Flash, mechanical or anything else, they all can and will fail. Putting all your eggs into one basket is a recipe for disaster

As someone who looks after other peoples systems I wholeheartedly agree.

[howdee]

In what possible way are hard drives reliable? They fail almost like clockwork, when analysed on a grand scale - remember the google survey from a few years back?

IIRC the typical scenario with NAND storage is that it degrades until it's no longer writeable, but it should still be readable. In the long-term I should think they'll be far more reliable than spindle-based storage.

We simply don't have any reliable long-term storage solution based on single drives, SSD being no exception. With this in mind, obviously there's a need for fail-safe storage solutions (parts will fail, but that doesn't necessarily mean the whole system has to) that would involve many of such drives in some sort of on-line storage arrays (like RAIDs) and, of course, backed-up properly (off-line). Since there's redundancy, single parts can fail and the array is still functional and can rebuild itself to nominal state by using hot spares. You're probably thinking by now such systems are only reserved for larger corporations and cost an arm and a leg, however this isn't true for quite some time now. I don't use HDDs on their own (JBODs) since Intel first introduced Matrix RAID (I use mostly RAID 10 arrays for their simplicity and speed gains). It might not be as fast as using hardware based RAID controllers, and there's a slight system overhead introduced with it, on the other hand it is cheap to implement even for individual needs, comes integrated on most mobos nowadays, and if any single of my drives fail I not only can replace the failed part but also continue using such "degraded" array as if nothing happened. I can afford this with hard drives, however I couldn't do this with SSDs, it would simply be too costly. Especially, if I wanted partitions of usable size.

[Funkstar]

I wonder if there will come a point when 2.5" drives are more economical for smaller capacities, such as up to 1TB?

Pretty sure set top boxes and DVRs are still based on 3.5" drives. Seems a bit of a wast of power these days.

[Unique]

my PVR has a 2.5" drive

[howdee]

I wonder if there will come a point when 2.5" drives are more economical for smaller capacities, such as up to 1TB?

Interesting point. Assuming we're talking about same number of platters and heads with either size then 2.5" should be cheaper simply for using less materials, and since they're smaller, lighter, they could transport more of them in same number of transport containers for the same price too, but they still sell with higher street prices than their corresponding 3.5" parts. It would be interesting to see numbers in which they sell worldwide as compared to 3.5" drives as I can't think of any other reason why would 2.5" drives be more expensive. Are slightly smaller DC motors and head transport mechanics really that more expensive to produce, or is there something else afoot? Hmmmmm..... "Some wee little gremlins stirring up mischief, perhaps?" (from The Borrowers by Mary Norton)

[badass]

Enterprise-class SAS 2.5" drives are ~ £100 for 300GB. A 300GB Intel 710 (enterprise, MLC, SATA II SSD) is £800+. No enterprise is going to pay 8x as much for storage unless you can demonstrate an overriding business case for the higher performance storage - and when you're talking huge redundant arrays then I doubt the real-world performance increase is noticeable. Sure, comparing one SSD to one HDD it's night and day, but compare 8 of each in RAID10 on a storage subsystem and then tell me you can qualitatively tell the difference. Then ask yourself whether you'd rather use one 300GB SSD, or 8 300GB HDDs in RAID10. I know which I'd choose

It's all about the IOPS. The RAID10's SSD's will be able to handle 10 or more times as many simultaneous connections in a database for instance.

SSD's really, really make a huge difference in certain types of workload. Certainly not all workloads and you need a very meaty back end to take advantage of them but they do have their place.