10TB helium drive

jtr1962

Storage? I am Storage!
Joined
Jan 25, 2002
Messages
3,651
Location
Flushing, New York
#41
You clearly don't own or use the newer drives. The helium requires less power to rotate and creates less noise.
MTBF is 2 or 2.5M for the newer enterprise drives. Compare that to a 2TB consumer drive. Anyway, run RAID 1 if you are paranoid about availability and backup regularly.
The only reason to buy HDD over SSD nowadays is lower cost per TB. Much of that advantage disappears when you look at enterprise drives. And then if use RAID1 the price advantage gets cut in half. Using Coug's quote for the 10TB drive and buying a pair to use in RAID1 gives you $1320US for 10TB. Decent 1TB SSDs like the 850s are going for a little over $300. So the price advantage of HDDs is only about 2.5x. I personally don't need 10TB of storage anyway. My 2TB HDD which I bought about 5 years ago is only about 60% full.

Has Backblaze tested any of the helium drives yet? What happens if the helium leaks out?

My point remains though. Whatever advantage helium gives you in terms of noise or power usage will carry over if you reduce rotational speeds. A 5400 RPM helium drive will have less noise and power usage than a 5400 RPM conventional drive. And instead of 10TB you could probably fit 12 or 13 TB on the same number of platters. If the entire point of HDDs is cheap bulk storage, then we should be aiming to maximize that.
 

LunarMist

I can't believe I'm a
Joined
Feb 1, 2003
Messages
14,808
Location
USA
#42
The only reason to buy HDD over SSD nowadays is lower cost per TB. Much of that advantage disappears when you look at enterprise drives. And then if use RAID1 the price advantage gets cut in half. Using Coug's quote for the 10TB drive and buying a pair to use in RAID1 gives you $1320US for 10TB. Decent 1TB SSDs like the 850s are going for a little over $300. So the price advantage of HDDs is only about 2.5x. I personally don't need 10TB of storage anyway. My 2TB HDD which I bought about 5 years ago is only about 60% full.

Has Backblaze tested any of the helium drives yet? What happens if the helium leaks out?

My point remains though. Whatever advantage helium gives you in terms of noise or power usage will carry over if you reduce rotational speeds. A 5400 RPM helium drive will have less noise and power usage than a 5400 RPM conventional drive. And instead of 10TB you could probably fit 12 or 13 TB on the same number of platters. If the entire point of HDDs is cheap bulk storage, then we should be aiming to maximize that.
My point was that you should run RAID 1. Of course I wouldn't.
The concept that slower RPM allows significantly higher data density is also a thing of the past.
As Coug mentioned, there are applications where one wants better performance than the sluggish 5400 RPM, but SSD is far too costly.
If you want a cheap drive get an 8TB helium 5400 RPM instead of the 7200 RPM enterprise grade.
 

jtr1962

Storage? I am Storage!
Joined
Jan 25, 2002
Messages
3,651
Location
Flushing, New York
#43
If you want a cheap drive get an 8TB helium 5400 RPM instead of the 7200 RPM enterprise grade.
But if they standardize on 7200RPM those won't be available, correct? As it is the pickings for 5400 RPM are pretty slim if you stick to decent brand HDDs like HGST, Toshiba, or Samsung. Seagate and WD are crap as far as I'm concerned and Backblaze's tests seem to bear this out.

I'll just wait it out a few more years for SSD prices to get closer to HDD prices. As it is now 1TB SSDs are almost in the affordable range but I'd probably want maybe 4TB to future-proof myself. Don't really want to spend $1500 for one now, so I'll just wait a bit.

Note that this is probably going to be for external backup in a USB drive enclosure, not internally in my machine. Given that, chances are nearly 100% I will accidentally drop it at some point given the way my hands are, which is why I can't really trust an HDD. I don't think a helium drive will survive a fall 2.5 feet off a desk onto a wooden floor. On the other hand, if my experience with flash drives is any indication, an SSD should be just fine. I've dropped flash drives 5 feet onto concrete without any problems. Also had the microSD card in my Garmin GPS survive just fine when the GPS came loose from my handlebars from hitting a pothole on my bike at 25 mph. The GPS still works, too. :)
 
Last edited:

snowhiker

Wannabe Storage Freak
Joined
Jul 5, 2007
Messages
1,439
#44
Hard drive are approximately 1,000,000x larger than they were 30 years ago. 10MB drives available 30 years ago. 10TB drives available now.

Question: Is reliability higher now or 30 years ago? Let's suppose you have 20 TB of data and replicated it 5x across 100 TB of disk space. Would 10,000,000 10MB HDD from 30 years ago be more reliable than 10 modern day 10TB drives?
 

jtr1962

Storage? I am Storage!
Joined
Jan 25, 2002
Messages
3,651
Location
Flushing, New York
#45
I don't think hard drives were particularly reliable 30 years ago. They were a relatively new product. It took a while to get it right. I want to say hard drive reliability probably peaked before hard drives became commodity items, say in the late 1990s. However, reliability didn't degrade severely until we started pushing the real physical limits of what could be done. I might say this happened when we got much past 500 GB per platter. At that point things got spotty. Some drives remained reliable, others didn't, and there was generally no way of knowing beforehand.

As for the reason why this happened, it's pretty obvious. For a really long time hard drives increased rapidly in size. It was almost expected that next year's drives would be at least 50% larger than this year's. The problems started happening when all the relatively easy ways to increase drive size started petering out. Customers still expected the same size increases each year. The only way HDD manufacturers could deliver was to rush products out of the lab well before they were really ready for prime time. If you think about it, how many other endeavors of mankind double key parameters every year or two? In general, for lots of industries a 50% increase in a decade is moving right along. One of my hobbies is thermoelectrics. Those have seen barely a 20% increase in efficiency in the last 40 years. Even LEDs, which evolved at a meteoric pace by any standards, took about a decade to increase ten-fold in efficiency (for comparison fluorescents took 50 years to double in efficiency) . In that same decade HDD sizes increased easily 100-fold. Anyway, once we maxed out the easy ways to increase HDD size, hard drives would have ended up following the same path of very small annual size increases (i.e. <10%) if not for the pressure to rush things into production. SSDs further added to this pressure.

Bottom line is maybe 5 years ago we had 2TB drives and most were fairly reliable. Now maybe we should have been at the point where we can make reliable 3TB drives. Instead, we opted for size over reliability. Even so, I've read most of the ways to increase HDD capacity further, like HAMR, may give us larger drives but these drives will be more costly, and therefore the price per TB won't drop.

To be fair I think SSDs may well suffer the same reliability issues if huge capacity increases each year are the sole driver of the market. In fact, we were actually starting to see problems caused by ever smaller cell sizes until 3D NAND enabled a way to increase capacity without hurting reliability. Eventually 3D NAND will peter out and we'll either have to get used to smaller annual size increases or less reliability.
 
Last edited:

CougTek

Serial computer killer
Joined
Jan 21, 2002
Messages
8,692
Location
Québec, Québec
#46
Given that, chances are nearly 100% I will accidentally drop it at some point given the way my hands are, which is why I can't really trust an HDD.
A 3600rpm mecanical drive has no better chance to survive that than a 7200rpm one.

The AFR of large 7200rpm drives is already quite low. With helium-filled drives, the noise problem mostly disappear. I don't see the point of opting for slower rotational speed at this point. I don't see why a manufacturer or large-scale buyer would invest in developping or asking for a slower drive (100% of the time) just to drop the AFR from 2% to something like 1.6% (yes, I pulled that from my a**, but so you did when you wrote about significantly increased reliability of slower HDD). The advantages/drawbacks ratio just isn't good.
 

LunarMist

I can't believe I'm a
Joined
Feb 1, 2003
Messages
14,808
Location
USA
#48
It's 2016. Some common sense needs to be applied. If you think that the two main HDD companies WD and Seagate are building terrible products, then how are they staying in business under the warranty?
There needs to be a balance between cost of the drives and reliability. It is OK for drives to fail occasionally. I don't mind paying an extra 2% or even 5% average in extra hard drives per year than paying a 400-500% premium for SSDs, that are still a bit small and can fail as well. I just keep a few spares drives on hand and replace the defective when there is a (rare) failure.
 

LunarMist

I can't believe I'm a
Joined
Feb 1, 2003
Messages
14,808
Location
USA
#49
But if they standardize on 7200RPM those won't be available, correct? As it is the pickings for 5400 RPM are pretty slim if you stick to decent brand HDDs like HGST, Toshiba, or Samsung. Seagate and WD are crap as far as I'm concerned and Backblaze's tests seem to bear this out.

Note that this is probably going to be for external backup in a USB drive enclosure, not internally in my machine. Given that, chances are nearly 100% I will accidentally drop it at some point given the way my hands are, which is why I can't really trust an HDD. I don't think a helium drive will survive a fall 2.5 feet off a desk onto a wooden floor. On the other hand, if my experience with flash drives is any indication, an SSD should be just fine. I've dropped flash drives 5 feet onto concrete without any problems. Also had the microSD card in my Garmin GPS survive just fine when the GPS came loose from my handlebars from hitting a pothole on my bike at 25 mph. The GPS still works, too. :)
AFAIK, helium is not yet used in the portable 2.5" drives. Frankly those drives have been the types that tend to fail for me far more often than 3.5" drives in a tower case.
I mitigate portable drive failures with at least three copies of data. Of course at <$100 for a 2TB drive a few failures is acceptable.
I thought you mostly worked from home and don't do much travel, so why are you using the slow and somewhat fragile portables?
Samsung has 1-2TB USB SSDs (T3 series) that are compact and will not die from a small drop on the hard flooring.
 

jtr1962

Storage? I am Storage!
Joined
Jan 25, 2002
Messages
3,651
Location
Flushing, New York
#50
AFAIK, helium is not yet used in the portable 2.5" drives. Frankly those drives have been the types that tend to fail for me far more often than 3.5" drives in a tower case.
I mitigate portable drive failures with at least three copies of data. Of course at <$100 for a 2TB drive a few failures is acceptable.
I have my important data on 2 SSDs, 1 HDD, and a few USB drives. I'd feel better though if I had it on something portable besides the USB drives.

I thought you mostly worked from home and don't do much travel, so why are you using the slow and somewhat fragile portables?
Samsung has 1-2TB USB SSDs (T3 series) that are compact and will not die from a small drop on the hard flooring.
Long term a portable SSD is in my plans. It's only recently that SSDs in the 1 to 2 TB class have started to get semi-affordable. Remember until I got this consulting gig in April 2014 I was in dire straights financially. All I could afford were USB drives when they had holiday specials on them.

Of course at <$100 for a 2TB drive a few failures is acceptable.

My 2TB 5400RPM Samsung has 50,364 hours on it according to SMART and still no reallocated sectors. If I recall it only cost my $109. Maybe I should have bought another back then.
 

CougTek

Serial computer killer
Joined
Jan 21, 2002
Messages
8,692
Location
Québec, Québec
#51
I'm glad I saved you from boredom on this Saturday afternoon.

I could address every link and keep debating, but it would become (or has it already?) as fruitless as a political or religious thread.

Two quick points :

* The point made in the second article is irrelevant nowadays because the technology to make platters twice as dense isn't there, no matter the rotational speed of the platter. You wouldn't get a significantly bigger drive by slowing it down and therefore, you wouldn't get the benefit from short-stroking it.

* Regarding all the other links, it assumes every other hardware used remains constant, which isn't the way the drives are designed (and stuff in general). Manufacturers will always design their product with a reliability target in mind. If you slow the rotational speed of the platters, then they'll cheapen out the bearings and other constituants of the drives to save cost and still maintain their targeted AFR. The fact is, they can and do make 7200rpm drives meeting their targeted AFR. They won't make a 3600rpm mechanical drive with components having 7200rpm-targeted tolerances in order to increase its reliability. It won't happen. Stop dreaming.
 

jtr1962

Storage? I am Storage!
Joined
Jan 25, 2002
Messages
3,651
Location
Flushing, New York
#52
Coug,

Believe whatever it is you want to believe. I probably forgot more about electronics, reliability, and failure rates than you'll ever know. Just because 7200RPM seem to work OK for what you use them for at work doesn't prove anything except they're adequate for that particular task. There are whole bunch of issues here like heat production (makes electronics less reliable), vibration (increases the BER and decreases bearing life), and so forth. Sure, I've little doubt drive manufacturers use better parts in 7200 RPM drives to mitigate these factors. However, that makes them more expensive, and at this stage in the game all HDD has going for it is cost. Even if you're hypothetically correct that drive manufacturers would cheap out on the parts for a 3600 RPM drive so it lasts no longer than a 7200 RPM, it would certainly cost less, which is really all that matters. Even Back Blaze buys 5400 RPM drives and makes them work because they generally cost less than 7200 RPM. Also note the bearings are one of the least costly line items on the BOM. They'll likely cheapen out on other things first, like make a less stiff case, or use a smaller motor. Helium isn't a panacea, either. Sure, it lets you build bigger drives, but it really isn't driving down the cost per TB because of all the extra cost items like hermetic sealing.

HDD manufacturers are probably focusing on 7200 RPM drives for one reason-the enterprise market is still willing to pay a premium for large, slightly faster HDDs. In the mass market price is everything. If HDD manufacturers were still trying to compete there they might indeed have 3600 RPM drives but at this point it makes little sense given the rate SSDs are penetrating this market. Most people don't need even 500GB. For them a <$100 250GB SSD works just fine. Really, the only ones being screwed by the standardization on 7200 RPM are home users who might need more storage than is affordable with SSDs. 5400 RPM drives represented a good cost/reliability solution for mass bulk storage in this market. That's basically my problem here-they're just making what is essentially a premium product. I guess the market dried up enough so they don't care if they lose some people.

Another thing worth a mention is even if enterprise users see some minute speed advantage with 7200 RPM drives it's coming at a heavy cost. When you have tens of thousands of drives an extra watt or two per drive means literally tens of kW more your cooling system has to get rid of. I can't help but wonder if anyone has bothered to do the math. Unfortunately, when people are in a field for years, often their thinking remains the same. Drive manufacturers pushed 7200 RPM in a big way in the early 2000s. Some people still haven't gotten out of that mindset. Even for your use, have you perhaps considered a small number of SSDs where you put your most often retrieved data and 5400 RPM for everything else? It may not yet be cost effective to go all SSD but we both know the vast majority of data is rarely retrieved, whether in a home or business setting.

The point made in the second article is irrelevant nowadays because the technology to make platters twice as dense isn't there, no matter the rotational speed of the platter.

The heads are capable of reading a certain number of bits per second. If you slow the drive down, you can pack proportionally more bits. No, paramagnetic limits won't let use pack in twice as many bits, but we could probably increase density by 15% or 20% dropping to 5400 RPM. You might also be able to use fewer ECC bits since your BER is lower at 5400 RPM, assuming the same areal density. That's another path to putting more data on a drive.

On another note, I'm actually happy drive sizes stopped increasing as fast as they used to. It seems program and data bloat squandered a lot of those gains, same as they did with faster processors. Programmers might actually have to start figuring out more efficient ways to store data at this point instead of assuming next year's drives will be 50% larger than this year's.
 
Last edited:

sechs

Storage? I am Storage!
Joined
Feb 1, 2003
Messages
4,634
Location
Left Coast
#53
I agree with standardizing on one RPM to avoid having multiple product lines, but from where I stand we should be going to 5400 RPM or less, maybe even 3600 RPM, for power, noise, and reliability reasons. Large HDDs these days solely function as bulk storage where access times really don't matter. You can get higher bit densities with lower RPMs, so STRs shouldn't be affected at all. I just don't get the point of 7200 RPM. More noise, more heat, more chance of losing data. It seems drive manufacturers are still stuck in a pre-SSD mindset.
You're in the minority. More 7200RPM drives are being sold than 5400RPM ones. So, the money savings is in dropping the slower spindle speed.

Your opinions about power, noise, and reliability may be true, but they're irrelevant. The market has spoken. End of story.
 

jtr1962

Storage? I am Storage!
Joined
Jan 25, 2002
Messages
3,651
Location
Flushing, New York
#54
You're in the minority. More 7200RPM drives are being sold than 5400RPM ones. So, the money savings is in dropping the slower spindle speed.

Your opinions about power, noise, and reliability may be true, but they're irrelevant. The market has spoken. End of story.
Right. I've read some of the opinions singing the praises of 7200 RPM and they border on a Sunday sermon. One which really cracked me up is someone who said 7200 RPM drives are as fast as SSDs. Coug was right. It's like religion. People are stupid (not referring to you Coug but to average people in general). Can't fix stupid. Props to the marketing departments at WD and Seagate. They could probably call their next drive Helium-filled 7200 RPM blue dog shit and people will be lining up around the block to buy it.
 
Last edited:

sechs

Storage? I am Storage!
Joined
Feb 1, 2003
Messages
4,634
Location
Left Coast
#55
Decent 1TB SSDs like the 850s are going for a little over $300. So the price advantage of HDDs is only about 2.5x.
LOL wot? I just picked up a 2TB 7200RPM drive for less than $60. The 1TB drives are ~$50. That's a 6x difference in cost, which is huge.

Most consumers are going to balk at that price difference.
 

sechs

Storage? I am Storage!
Joined
Feb 1, 2003
Messages
4,634
Location
Left Coast
#56
if you stick to decent brand HDDs like HGST, Toshiba, or Samsung. Seagate and WD are crap as far as I'm concerned
Come on, man
HGST=WD
Samsung=Seagate

There are no "decent" brands any more. The market is highly commoditized and caters to the lowest common denominator.
 

jtr1962

Storage? I am Storage!
Joined
Jan 25, 2002
Messages
3,651
Location
Flushing, New York
#57
LOL wot? I just picked up a 2TB 7200RPM drive for less than $60. The 1TB drives are ~$50. That's a 6x difference in cost, which is huge.

Most consumers are going to balk at that price difference.
Are those things going to be reliable for those prices? I can believe we can make reliable 7200 RPM enterprise drives at high capacities for maybe $200 and up but I can't help but believe these $50 drives cut everything to the bone. They probably use components really only suitable for 5400 RPM but jack up the speed for marketing reasons. This is really the problem I've having in a nutshell. The ultra-cheap mass market should be 5400 RPM or lower because drive manufacturers WILL cheap out on parts there. At least lower RPM ensures the drives have some chance of being reliable over the long haul.

Note the mass market seems to work mainly on price points, with capacity only being relevant if you really need that much storage. $50 to $75 seems to be what most people will pay for a drive. And if they know they can get by with 250GB or less many will opt for an SSD. Your 2TB $60 drive really isn't competing with a 2TB SSD. It's competing with whatever size SSDs can be had for about the same price. I still stand by my remark. Decent 1TB or 2TB HDDs which I would actually trust my data with are well into the $150 to $200 range.

There are no "decent" brands any more. The market is highly commoditized and caters to the lowest common denominator.

I'm going by Backblaze's data. HGST seem to have the edge for reliability.

Samsung had a equally well-deserved reputation a few years back. I suspect they left the HDD market for three reasons. One, they saw greater profitability with SSDs. Two, they saw HDDs as an eventual dead-end. Three and most important, with the increasing commoditization of the market they may have been unable to make a drive meeting their reliability criteria at price points the market was willing to pay.
 
Last edited:

sechs

Storage? I am Storage!
Joined
Feb 1, 2003
Messages
4,634
Location
Left Coast
#58
Are those things going to be reliable for those prices? I can believe we can make reliable 7200 RPM enterprise drives at high capacities for maybe $200 and up but I can't help but believe these $50 drives cut everything to the bone. They probably use components really only suitable for 5400 RPM but jack up the speed for marketing reasons. This is really the problem I've having in a nutshell. The ultra-cheap mass market should be 5400 RPM or lower because drive manufacturers WILL cheap out on parts there. At least lower RPM ensures the drives have some chance of being reliable over the long haul.
The drives are just as reliable as when they were $70 or $100. The warranty is the same length.

No one said anything about enterprise. Even if we were, the price multiple between enterprise spinning disks and SSDs isn't better.

You need to take the blinders off on slower drives. If they make a drive that spins slower, they just use cheaper parts with the same reliability ratings. There's no advantage but the price.

Note the mass market seems to work mainly on price points, with capacity only being relevant if you really need that much storage. $50 to $75 seems to be what most people will pay for a drive. And if they know they can get by with 250GB or less many will opt for an SSD.
That does not follow. If the mass market is sensitive to price, they won't pay twice as much for a quarter of the space. Most people don't understand SSDs, let alone are willing to pay the premium for the performance.

Decent 1TB or 2TB HDDs which I would actually trust my data with are well into the $150 to $200 range.
The enterprise version of 2TB drive that I bought is about $70. This has been mentioned, but you could just get three "indecent" drives and run a double mirror with much greater reliability and performance.

Your point appears to apply exactly to one person, and that's you.
 

jtr1962

Storage? I am Storage!
Joined
Jan 25, 2002
Messages
3,651
Location
Flushing, New York
#59
The drives are just as reliable as when they were $70 or $100. The warranty is the same length.
Warranty periods went down to a year on a lot of the cheaper drives. Standard for consumer drives a few years back was 3 years.

No one said anything about enterprise. Even if we were, the price multiple between enterprise spinning disks and SSDs isn't better.
Enterprise SSDs are typically faster, not necessarily more robust. It's a fair comparison using enterprise HDDs and decent consumer SSDs like Samsung. Both have a 5-year warranty.

You need to take the blinders off on slower drives. If they make a drive that spins slower, they just use cheaper parts with the same reliability ratings. There's no advantage but the price.
Less power consumption and less noise are also advantages.

That does not follow. If the mass market is sensitive to price, they won't pay twice as much for a quarter of the space. Most people don't understand SSDs, let alone are willing to pay the premium for the performance.
SSDs have been hyped much like the 7200 RPM drives before them. Also, if you ask most computer users how big their drive is you'll get a blank stare. Most people buy new drives solely on price and speed (hence the reason 7200 RPM sells). When a salesperson says you can get something way faster than 7200 RPM for the same price, don't think most people won't go for it. Only those who actually know they need a certain amount of space won't. Also, the cost of the drive is a fraction of the system cost. You get more performance gains by going from HDD to SSD than you would by upgrading any other component. Salespeople will harp on that fact indefinitely. I know because once I went in MicroCenter asking about a 4TB HDD they had on sale. The salesperson kept pushing me to get a 500GB SSD instead with stuff like "this will make your system fly", "how many people need more than 500GB?" and so forth. I hadn't intended on buying anything actually but this exchange was very enlightening. As mostly a bunch of computer people, I don't think too many here have a pulse on what the average person thinks.

The enterprise version of 2TB drive that I bought is about $70. This has been mentioned, but you could just get three "indecent" drives and run a double mirror with much greater reliability and performance.
So you need the space and power for three drives, plus you have to cope with the extra noise? Doesn't seem like a great answer unless you're running a data center. I don't even know how to set drives up in RAID, much less have any desire to. Your answer here is like me telling someone who wants to measure their car battery voltage to get a Fluke 189 like I just bought instead of a $3 multimeter at Harbor Freight. Sometimes I need to take off my EE blinders when giving people advice. I'd recommend the computer experts here do the same.

Addendum: I did the math for the extra power consumption. At let's say 6 watts per drive that's 18 watts. At the power meter it's maybe 21 watts accounting for power supply efficiency. My machine is generally on 24/7, so in a year that's ~184 kW-hrs. Over the 5 years I would probably use the drives it's 920 kW-hrs. At NYC prices of about $0.30 kW-hr delivered I'd be paying $276 for power plus and $210 for three drives. Total cost then for this solution using three drives is $276 for power plus $210 for the drives. Grand total $486. Now look at using an SSD instead. Average power consumption is close the idle power consumption of 60 mW given that I rarely access bulk storage. Let's call it 100mW for argument's sake. Over the same 5 years of 24/7 operation total power usage at the meter is ~4.4 kW-hrs at a whopping cost of $1.32. The 2TB SSD would cost me about $630 going by Newegg's recent prices. So grand total is $631.32. I'd rather pay an extra ~$150 for a solution which uses far less space, makes no noise, and is likely just as reliable, if not more so. Within less than a year I'd say TCO between these two solutions will be the same. We just need 2TB SSDs to fall to around $480.

Before you say BUT you can have the HDDs spin down to save power, well, sure you can but then if you occasionally access data your access times go WAY up waiting for those drives to spin up. Kind of negates the entire notion that "7200 RPM is faster". I actually tried letting HDDs spin down a few times. The lag was so annoying I just let them run 24/7 now. Probably better for mechanical reliability as well.

Your point appears to apply exactly to one person, and that's you.
It applies to anyone who evaluates these things logically. HDDs are great at one thing-cheap, bulk storage. As such, they should consume as little power as possible, make as little noise as possible, and be as cheap as possible. In general, slower RPM maximizes these three criteria. 7200RPM means more noise and power and heat, and either lower reliability or a more expensive drive to get the same reliability. So what's the point? A little extra speed? For bulk storage it just doesn't matter if your access times are little less. 7200 RPM gets you exactly 1.38888 ms less rotational latency and therefore access times (assuming seek times are identical) over 5400 RPM. It may or may not get you 33% better STRs. That depends upon whether or not you increase areal density if you go to 5400 RPM. My point here is even at 3600 RPM seek times and STRs would be "good enough" for 99% of bulk storage needs. Maybe this doesn't work for some data centers, but that's why you have the more pricey 7200 RPM enterprise drives.

This is also why I can't wait for SSDs to become cheap enough to put spinning disks in the dust bin of history for good. If HDDs makers are going to eliminate products which were a better fit for lots of people then they deserve to go out of business. Here's hoping Seagate and WD file chapter 8 within the next few years.
 
Last edited:

jtr1962

Storage? I am Storage!
Joined
Jan 25, 2002
Messages
3,651
Location
Flushing, New York
#61
See my TCO analysis in my last post. Also, no mention of a warranty. 3 years? 5 years? 1 year? Anything in that size with a 5-year warranty seems to cost $120 and up.
 

LunarMist

I can't believe I'm a
Joined
Feb 1, 2003
Messages
14,808
Location
USA
#62
TCO needs to include the other computer components, enclosures, ordering, building, physical space, and FBC for storage management.
Then you can better judge the importance of the hard drive costs.

You seem to be hung up on cheap and reliable, which makes no sense in the marketplace.
Stop thinking of drive reliability and think about data reliability through redundancy.
 

jtr1962

Storage? I am Storage!
Joined
Jan 25, 2002
Messages
3,651
Location
Flushing, New York
#63
You seem to be hung up on cheap and reliable, which makes no sense in the marketplace.
Stop thinking of drive reliability and think about data reliability through redundancy.
Maybe I need to clarify things a bit. HDDs are inherently unreliable simply because they're mechanical devices. You can make reliability acceptable for relatively low cost. Let's call a 5-year median lifespan "acceptable" for the sake of argument. If you make them too cheap, they're unacceptably unreliable. Maybe in theory you can make HDDs with a 10 year median lifespan but they would cost so much nobody would want them. So you hit a balance and aim for a 5-year lifespan. However, that's not the entire story. As with any mechanical device, all other things being equal it wears out faster if it moves faster. That means a faster drive requires more robust parts or it won't reach an acceptable reliability. So to make your drives acceptable reliable and as inexpensive as possible, it makes sense to reduce the speed of anything which moves. Now why do you want HDDs as inexpensive as possible? The reason is simply because their only reason for existing at this point is they cost less per TB of storage. SSDs are superior in every other metric, so you have to make HDDs as inexpensive as possible while still being acceptably reliable or the market just won't want them. It makes perfect sense in the marketplace. People don't want unreliable products at any price. They want something which works at as low a price as possible.

Now let's look at SSDs. Yes, they have failure mechanisms also, but with no moving parts they're inherently reliable. Moreover, the failure mechanism is well-known and can therefore be monitored, as well as minimized by wear-leveling. If an HDD bearing or head fails, the drive is history. If a flash cell in an SSD fails, you may possibly lose the data in that cell, but another will take its place and the drive will continue functioning. It would take a major failure of one of the main circuit boards to put the SSD out of commission. Does that happen? Sure, but modern electronics is orders of magnitude more reliable than mechanical devices. How often does RAM fail for example despite being written and read millions of times per second? If we came up with a type of non-volatile storage which didn't have a finite number of write cycles, arguably an SSD could have a median lifespan of many decades. So you have inherent reliability, along with much greater speed. These things are why the market has been willing to pay a large premium for SSDs but not for HDDs.

That brings us to redundancy. Redundancy is needed more when you're using a device which is inherently unreliable. If I only backed up my data to HDDs, I probably wouldn't feel safe unless I had it on at least 10 drives. If I back it up on SSDs and USB drives I feel pretty comfortable having it in 4 or 5 places. Even that might be overkill.

In summary, if a technology is inherently unreliable, then I expect it to be as inexpensive as possible while still meeting an acceptable reliability criteria. If its inherently reliable, I'm willing to pay more for it. Good example is the Fluke 189 I bought two weeks ago. It's a wonderful, robust multimeter, and I paid $375 for it. It also has more functionality than less expensive meters. Those two things made it worth the premium. On the other hand, I sometimes just need a very basic multimeter or two so I can measure voltage or current of several devices simultaneously. I don't care much if it's very robust because I won't use it in heavy duty applications. I also don't expect it to hold up to repeated use. And I won't use it where I need high accuracy. However, it's inexpensive enough so I can buy several if one or two fails. In fact, the only reason I buy it, despite its limitations, is precisely because it's very inexpensive. I won't pay even $20 but for $3 or $5 I'm more than happy. That pretty much sums it up with HDDs. I know their limitations. I know I can't trust storing important data on just one HDD, so I won't. However, I don't want to pay a penny more than I have to given that I'm buying an inherently inferior product. And drive manufacturers not offering the lowest possible RPM such that the drive still works is directly contradictory to that. Ideally an HDD would be 0 RPM but you need a certain speed for the heads to fly. Sure, data redundancy is important but when you do the math HDDs, even inexpensive HDDs, are really too expensive for an non-business user to implement redundancy given the alternative SSDs provide. Adding enclosures, space, etc. as you suggest would only skew my TCO example more in favor of SSDs. If I were running a data center, I'd actually take all that into consideration, not just the cost per TB of the drives themselves.
 

time

Storage? I am Storage!
Joined
Jan 18, 2002
Messages
4,795
Location
Brisbane, Oz
#64
As with any mechanical device, all other things being equal it wears out faster if it moves faster. That means a faster drive requires more robust parts or it won't reach an acceptable reliability.
You're neglecting scale when drawing your conclusions. A surface exposed to friction will indeed wear out faster if it moves faster, but if the time it takes to wear out is 50 years, it's not a relevant datum. I don't know exactly how long it takes to wear out a hydrodynamic bearing in a hard disk drive, but that doesn't seem at all unreasonable as a guess.

In other words, you are assuming that every part of the device will only last as long as the life expectancy of the device as a whole. From that, you deduce that any change in the operating parameters would affect the manufacturing cost.

There are many points of failure in a modern hard disk drive, but spindle friction is now way down the list. I was under the impression that controller and media problems accounted for most failures?
 

Howell

Storage? I am Storage!
Joined
Feb 24, 2003
Messages
4,740
Location
Chattanooga, TN
#65
On top of what time said, the cost for the size most people want is too high in the SSD form. If a spinning drive is fast enough for what it cost, people will continue to buy it.
Fyi, people still want their bulk storage to be faster rather than slower.
 

Stereodude

Not really a
Joined
Jan 22, 2002
Messages
10,179
Location
Michigan
#66
Instead you should trust flash cells that are guaranteed to slowly leak charge and eventually be unreadable instead of a mechanical component that might go bad. I'd trust a mechanical HDD that sat for 20 year to have the data recovered more than a SSD that say 20 years.
 

jtr1962

Storage? I am Storage!
Joined
Jan 25, 2002
Messages
3,651
Location
Flushing, New York
#67
Instead you should trust flash cells that are guaranteed to slowly leak charge and eventually be unreadable instead of a mechanical component that might go bad. I'd trust a mechanical HDD that sat for 20 year to have the data recovered more than a SSD that say 20 years.
Neither device is really suitable for archiving. The magnetic bits on an HDD decay also due to superparamagnetism. This wasn't an issue when drives were less dense. Indeed, a 20MB drive from the 1980s would probably still be readable in a century. I wouldn't bet on one of today's drives being readable in 20 years.
 

LunarMist

I can't believe I'm a
Joined
Feb 1, 2003
Messages
14,808
Location
USA
#68
Neither device is really suitable for archiving. The magnetic bits on an HDD decay also due to superparamagnetism. This wasn't an issue when drives were less dense. Indeed, a 20MB drive from the 1980s would probably still be readable in a century. I wouldn't bet on one of today's drives being readable in 20 years.
Maybe some of the pros can answer, but I'd be surprised if drives are normally intended used much beyond 5 years in storage servers.
 

jtr1962

Storage? I am Storage!
Joined
Jan 25, 2002
Messages
3,651
Location
Flushing, New York
#69
Maybe some of the pros can answer, but I'd be surprised if drives are normally intended used much beyond 5 years in storage servers.
Backblaze has some interesting data on that.

5 years sounds like a reasonable number here.

It would be interesting to know a similar ballpark figure for the archival life of SSDs. I know the data in 3D NAND decays more slowly because they're using larger cells. How long out can one expect that data to be viable?
 
Top