WD HDD Industry Will Be Supply Constrained Due to Thailand Flooding

[Pentium100]

More than 10TB. The drive I have has 8 platters, but there were other drives that had 16 (I don't know if there was a drive with more platters). However, since each platter is 5.25" it would hold more data than a 3.5" platter, because it has more surface area (about twice, since area is proportional to the square of the radius). So, the 5.25" full height drive would probably be 32TB (16 2TB platters). I think these drives will return someday, when SSD takes over as the "fast" storage. Then these drives will be "slow, but large and cheap" storage because they will most likely be cheaper to make than a bunch of 3.5" drives that have the same capacity.

I wouldn't want to have my system partition on such drive, but for my anime collection it would be perfect.

[CharredChar]

Again, having such large single drives is worthless. A good reason Id consider that is because of this.

http://www.zdnet.com/blog/storage/why-raid-5-stops-working-in-2009/162

Id rather a RAID with multiple drives so when one fails (and Ive had three fail on me already) you can swap out the drive and have it rebuild itself. Youll find the more relable drives have less parts, platters, read heads, etc. Of course, even increasing the density of a platter will increase failure rate, but Id rather a two platter 2TB drive than a four platter 2TB drive.

[Freedom Kira]

Well, the point was more for the storage space, not really the reliability. Obviously, the savvy would keep their critical data off of drives like those. It'd probably be good for keeping a cheap backup for non-critical data.

Not too sure that full-size disks would be slower, either. I would expect them to be faster - more platters, and larger platters means the outside data can be accessed more quickly. They'd probably use a lot more electricity though, unless they decide to decrease the rotational speed to account for the platter size.

Implementing a built-in platter-wise RAID 1/5/6/10 would be nice too, if drive manufacturers happen to realize the dangers of drive failure instead of pushing forward with disk sizes.

[Pentium100]

Again, having such large single drives is worthless. A good reason Id consider that is because of this.
Id rather a RAID with multiple drives so when one fails (and Ive had three fail on me already) you can swap out the drive and have it rebuild itself.

So, have a RAID of the big drives. A bunch of 3.5" drives will take up more physical space then one 5.25" drive of the same capacity. So, a bunch of 5.25" drives will take less space than a larger bunch of 3.5" drives.

Not too sure that full-size disks would be slower, either. I would expect them to be faster - more platters, and larger platters means the outside data can be accessed more quickly.

Seeks will be slower. Linear speeds will be higher.

Implementing a built-in platter-wise RAID 1/5/6/10 would be nice too, if drive manufacturers happen to realize the dangers of drive failure instead of pushing forward with disk sizes.

Not worth it - it is more likely that the drive fails completely instead of developing a lot of bad sectors on one platter. All the drives I have, some work perfectly, some have a few bad sectors and some do not spin up or cannot read/write even a single byte. I have never encountered a drive that has 25% (or some other large number) of its sectors bad.
Have you seen how much a hardware RAID controller costs? I wouldn't want each drive to come with its own RAID controller and not have redundant motors/heads/etc but still pay a lot for that controller.

[Freedom Kira]

Seeks will be slower. Linear speeds will be higher.

Ah, yes, I forgot about that.

Not worth it - it is more likely that the drive fails completely instead of developing a lot of bad sectors on one platter. All the drives I have, some work perfectly, some have a few bad sectors and some do not spin up or cannot read/write even a single byte. I have never encountered a drive that has 25% (or some other large number) of its sectors bad.
Have you seen how much a hardware RAID controller costs? I wouldn't want each drive to come with its own RAID controller and not have redundant motors/heads/etc but still pay a lot for that controller.

True on the drive failure. On the RAID controller though, call me optimistic, but I don't think a platter-wise controller would cost as much as a regular SATA one, mostly because it doesn't need to interface to the generic SATA/SAS/SCSI/whatever connection, nor would it have to be as fast, since the output of the whole thing would be bottlenecked by the drive's connection anyway. Would it not be similar to an SSD's controller, concept-wise?

[CharredChar]

For data across multiple platters, yes it would work just like SSDs that have multiple chip boards.

Well, the point was more for the storage space, not really the reliability. Obviously, the savvy would keep their critical data off of drives like those. It'd probably be good for keeping a cheap backup for non-critical data.

I would never and dont know anyone else either that would buy a drive that large if it didnt have a higher reliability of 3.5" drives equaling the same space. And if youre buying a bunch of 3.5" drives or even another 5.25" drive just for backup of a unreliable drive, whats the point? Id rather spend the money on the more reliable ones in the first place.
Of course, if you can keep the reliability the same or better and lower the price? Im all for that, but that is a Really hard thing to do. Id think the reason the industry stuck with the 3.5" drives is because its a good mix between reliability, memory size, and price.
Honestly, why wouldnt they keep making 5.25" just for data servers then if it didnt run a higher risk of issues? Im sure it would be cheaper for a admin to run half as many physical drives for the same space. Moreso when...

[Pentium100]

RAID controllers are expensive because they need to calculate checksums very fast or the controller itself becomes the bottleneck. Same thing would happen if you move the controller to the drive itself. It would still need to calculate the checksums. Yes, the controller on the drive could be a bit slower but then the read/write circuitry would have to be multiplied so all the platters could be written at once (since if you only wrote with one head at a time, it would be slower than a regular drive writing sequentially, also, regular drive does not need to write checksums). You also need another platter, so even if the electronics were free, a 2TB RAID drive would cost as much as a 3TB drive because of the additional platter.

All that to save the data in case of a very unlikely event - a platter or a head fails, but the rest of the drive works normally. It would be better to just use more error correction so a bad sector can be recovered.

Im all for that, but that is a Really hard thing to do. Id think the reason the industry stuck with the 3.5" drives is because its a good mix between reliability, memory size, and price.

Because 5.25" drives are slow when random access is needed. 3.5" drives are faster, that's why they are the industry standard. However, when SSDs become cheaper and take over the "fast drive" function, then, I think, 5.25" drives will come back as the "slow, but big and cheap" drives.

As for reliability - I do not know what the statistics are, but my drive is probably 20 years old and is still working perfectly (zero grown defects).

[CharredChar]

Because 5.25" drives are slow when random access is needed. 3.5" drives are faster, that's why they are the industry standard. However, when SSDs become cheaper and take over the "fast drive" function, then, I think, 5.25" drives will come back as the "slow, but big and cheap" drives.

As for reliability - I do not know what the statistics are, but my drive is probably 20 years old and is still working perfectly (zero grown defects).

Just like a DVD, the larger diameter of the disk/platter would cause slower read/write speeds at the outer edge, yes, but a 10,000rpm drive (lol Would that even be possible with such a large platter?) would limit that as being the bottleneck.
Youre already seeing slower drives being used as a "big cheap drive", almost all larger drives are 5,400RPM. They use less power, produce less heat, and I can only assume are generally more reliable than the same drive running at a higher RPM. I myself only use 10,000RPM drives or SSDs in all my PCs, even the server, for system drives. As for storage I just RAID the large 5,400RPM drives on the server.
Read that link I posted before, it would also explain why older, smaller drives have a better MTBF.

[fohfoh]

I've personally had more issues with 5400rpm HDDs than 7200rpm ones.

Never used a 10000rpm one, and my SSD is chugging along ok. But too early to really say anything about reliability.

[AnimeJanai]

I have that nasty feeling that 5900rpm and 5400rpm platters are the ones that failed to qualify at 7200rpm but were still good enough to work at the lower speed.  No proof, but when thinking about what the chinese factories do with their 7200rpm failures and I cannot envision them throwing it away just because it has too much vibration jitter to work at 7200rpm.   The chinese places have fairly unethical prosperity practices in my opinion.  But ethical practices just cannot compete with the low-priced seller and I can only assume that eventually the lowest common denominator pulls the other competitors down into the muck as well.

[CharredChar]

I have that nasty feeling that 5900rpm and 5400rpm platters are the ones that failed to qualify at 7200rpm but were still good enough to work at the lower speed.

Would you say the same with CPUs?

[AnimeJanai]

I have that nasty feeling that 5900rpm and 5400rpm platters are the ones that failed to qualify at 7200rpm but were still good enough to work at the lower speed.

Would you say the same with CPUs?

Certainly, we all know Intel and AMD re-qualify their CPUs for sale at either a lower speed or fewer cores.  But a defective hard drive platter is a different matter as compared to disconnecting the bad core on the die or running at a lower speed.  The CPU will not have an early death.  Defective drive platters may run for awhile at 5400rpm and then the drive will have an early failure if the problem with the platter is serious enough.  That's the part that makes it unethical.

[AceHigh]

I simply love this whole hard drive crisis!

Just read an article that pointed out how HD crisis just gave a boost for the SSD market. Thanks to the flood it seems we will get more affordable SSDs much sooner than we would otherwise, they estimated 1$ per 1GB in late 2012.

[kitamesume]

I simply love this whole hard drive crisis!

Just read an article that pointed out how HD crisis just gave a boost for the SSD market. Thanks to the flood it seems we will get more affordable SSDs much sooner than we would otherwise, they estimated 1$ per 1GB in late 2012.

LOL later!? can't they make it much more sooner? ohh well, 60$ for a 60GB SSD~

[Lupin]

I have that nasty feeling that 5900rpm and 5400rpm platters are the ones that failed to qualify at 7200rpm but were still good enough to work at the lower speed.  No proof, but when thinking about what the chinese factories do with their 7200rpm failures and I cannot envision them throwing it away just because it has too much vibration jitter to work at 7200rpm.   The chinese places have fairly unethical prosperity practices in my opinion.  But ethical practices just cannot compete with the low-priced seller and I can only assume that eventually the lowest common denominator pulls the other competitors down into the muck as well.

It's not the platters that dictate the drive speed. It's the electronics (both motor and drive controller) that dictate those. They cannot just slow down a 7200 drive to a 5900/5400 one because each uses different set of controllers/firmwares. Swapping boards on these won't work and will most likely destroy the motors of the drive. Also note that it's the slower drives that gets the new tech first (higher platter densities) while higher speed models get the refined tech.

Just like a DVD, the larger diameter of the disk/platter would cause slower read/write speeds at the outer edge

This is wrong. The outer edges have higher read/write speeds since the head covers much more distance (at the same rotational speed) in this region. The reason we see less 10K models is because the platters wobble more at that speed and that causes errors. Since platter densities are higher, there would less margin for error. WD remedied the problem by releasing their 10K at a smaller form factor but the gains are getting smaller and not worth the effort. 10k drives will be sucking more power but provide less and less performance gains.

[vuzedome]

Love it when simple physics and economics were ignored.

[Pentium100]

15kRPM drives use even smaller platters. That's why there are no 2TB 15kRPM drives.

[CharredChar]

Just like a DVD, the larger diameter of the disk/platter would cause slower read/write speeds at the outer edge

This is wrong. The outer edges have higher read/write speeds since the head covers much more distance (at the same rotational speed) in this region. The reason we see less 10K models is because the platters wobble more at that speed and that causes errors. Since platter densities are higher, there would less margin for error. WD remedied the problem by releasing their 10K at a smaller form factor but the gains are getting smaller and not worth the effort. 10k drives will be sucking more power but provide less and less performance gains.

Oops, it seems I was picturing it all backwards in my head.

[kitamesume]

since we're talking about HDDs here i'm gonna ask a silly question. how far are the failure rates of laptop vs desktop HDDs on the same capacity?

[CharredChar]

since we're talking about HDDs here i'm gonna ask a silly question. how far are the failure rates of laptop vs desktop HDDs on the same capacity?

Excluding specific drive issues? It should be the same.