El Reg gets schooled on why SSDs will NOT kill off the trusty hard drive

Some commentators think saying SSDs will kill off disk drives is wrong. Here is an exchange El Reg had with a storage media industry insider who wished to remain anonymous Commentator: I read your article about SSD replacing HDD and I think you completely underestimate the challenges of increasing NAND capacity to meet the …

[Anonymous Coward]

I just replaced the HDD in my laptop with...

...a HDD

Why? I use the thing for web browsing, messing around with photos, watching iplayer and composing documents..... and a 1TB HDD was £46 including next day delivery.

Yes, an SSD would be faster, but I already had about 220GB of stuff on my old hard disk, so it's not hard to see that a £70-£80 240GB SSD wouldn't be up to the job for long, so I'd have to fork out even more for a ~500GB SSD.

I have difficulty believing that the price per GB of a given SSD size actually reflects the relationship between development/manufacturing cost and capacity.

[LewisRage]

Re: I just replaced the HDD in my laptop with...

The solution you really need for a single disk laptop is a hybrid that gives you some of the performance that SSD offers whilst retaining the cheap capacity a spinning disk gives (in fact a 1TB 'SSHD' is on eBuyer for £47 currently, so cost isn't a differentiator).

And this is really the crux of the interview isn't it? SSD technology will be used to provide high performance, low capacity functions whilst spinning rust provides the cheap cheap capacity.

As with everything I'm sure eventually a 9-dimensional quantum nano-cube will appear that will blow them both out of the water but I suspect that we'll retain the current paradigm for a while as yet.

I mean I'm still sending offsite, long term storage out on reels of magnetised plastic and I've been being told that tape storage was dead for years now*, which is odd because there are lots of people selling large robotic tape libraries and we recently saw reported a new record achieved in tape density so it's clearly still relevant. And look at Veeam back tracking and adding in tape support.

*mainly by people trying to sell me backup solutions that don't have the technology to hook into tape drives.

[Howard Hanek]

Re: I just replaced the HDD in my laptop with...

I've set up hundreds of laptops over the last couple of years and the best compromise for speed/capacity have been hybrid drives. For the last two years for those people who require considerable amounts of data, like muscians, performers, legal, scientific a 1TB is usually sufficient. Of course the more data the larger the backup and the longer the time required to protect your data.

[dmacleo]

Re: I just replaced the HDD in my laptop with...

I have about 6 of the seagate 1tb sshd hybrids deployed across 2 small domains (all in desktops no laptops) and see no real benefit. really a small ssd for windows with a 500gb to 1tb plattter for data would work better for me.

but...it does work better for some so YMMV of course.

edit: fwiw on desktops here at home on this domain I use intel ssd (240gb) with 2-4 multi TB (mix of 2 and 3 tb) platters as secondaries.

this works real well for me.

so....ssd and platter gives me options that platter alone did not give me.

[Asylum_visitor]

Re: I just replaced the HDD in my laptop with...

This has also been my experience, hybrid drives sound like a good idea on paper but the flash caching isn't really big enough to provide any noticeable benefit. With an SSD it's hard to argue that there's a performance increase but with hybrid not so much.

[Anonymous Coward]

Re: I just replaced the HDD in my laptop with...

I have an Acer that came with a 1T HD, and an otherwise undocumented SSD interface on the motherboard. I ended up creating a hybrid.

[GoatFace]

Re: I just replaced the HDD in my laptop with...

"(in fact a 1TB 'SSHD' is on eBuyer for £47 currently, so cost isn't a differentiator)"

This little nugget stood out to me as I've been looking to buy a 500Mb SSD for most of this year and haven't seen the price drop below £100 except in rare sale events. Looking at Ebuyer today the lowest entry SSD in the 500Mb range is £129. In the 1TB range it's actually £260 - a long way away from your quoted figure! So you'll forgive me if I disagree with your abrupt dismissal of pricing not being a differentiator.

[the spectacularly refined chap]

Re: I just replaced the HDD in my laptop with...

SSHD, i.e. a hybrid drive. The pricing for those is broadly comparable to the equivalent hard drive. About twelve months ago I got a 4Tb/8Gb flash hybrid, if memory serves it was around a fiver more than the equivalent purely mechanical drive.

[Anonymous Coward]

Re: I just replaced the HDD in my laptop with...

Tape back up may still be a continued (and massively declining sector). Who are the many people still trying to selling large tape libraries? I can name them: IBM. Oracle. Spectralogic. That's too few imho. Its a declining market.

IBM making bigger arial density films matters not - especially with them (IBM) in particular, shrinking their storage HW business for something like 18 straight quarters. Tape is a relatively small portion of their dwindling storage tin revenues

Tape recovery of large systems is, on the other hand, a different fish altogether. It's not a compelling story when you are in DR and can't read back a significant proportion of tapes from the vault. Especia;;y those not retensioned regularly, spanning several data and tape formats, and maybe with the catelogues gone. It's a terrifying and career-limiting prospect for any back up Owner.

Don't get me wrong I build a career on tape, and I love the technology in principle. But it's only just hanging on in there as a viable market sector for vendors.

[Trollslayer]

Re: I just replaced the HDD in my laptop with...

SSD prices really do reflect costs especially give a 32 layer device has 32 layers of cells where a cell probably has four layers to add.

[Field Commander A9]

Re: I just replaced the HDD in my laptop with...

So like what? An external USB HDD casing never occurred to you?

Any sane tinker in your shoes would have bought an SSD together with a USB HDD casing. Put the SSD in as the boot drive and put your old HDD into the USB casing for extra data space.

[Doctor Syntax]

Re: I just replaced the HDD in my laptop with...

"Put the SSD in as the boot drive and put your old HDD into the USB casing for extra data space."

I get it: it's a use case that works for you so of course it must work for everyone else.

[Kiwi]

Re: I just replaced the HDD in my laptop with...

Any sane tinker in your shoes would have bought an SSD together with a USB HDD casing. Put the SSD in as the boot drive and put your old HDD into the USB casing for extra data space.

No, any sane thinker realises that's a pile of extra crap to cart around when it's not necessary, and for a pittance they can have 10x the storage space internally that they could have with spending a small fortune for a pitiful amount of space on SSD.

That and, given the "wonderful" speeds of USB, adding a HDD via USB so you can have the performance of SSD kinda defeats the purpose doesn't it? Just means you're a bit faster to reach the point where you have to wait for the machine to crawl.....

[katrinab]

Re: I just replaced the HDD in my laptop with...

I replaced a 1TB hdd on my laptop with a 960GB ssd, and the speed increase was like getting a brand new laptop. It wasn't by any means the fastest sdd on the market, but still well worth the extra money.

[Anonymous Coward]

Just another tier

Likely we will just end up with another tier in the cache hierarchy .... and that's before SCM / 3D-xpoint create yet another between DRAM and SSD

SRAM

DRAM

SCM/XPoint

NAND

HDD

TAPE ....

will they lose their shirts ? .. probably some will (its been happening since the 16K DRAM), but if they can move wafer starts between DRAM/SCM and NAND then maybe most will be OK.

[Fading]

SSDvHDDvTape

The death of tape hasn't happened yet so I would expect the death of spinning disc is some way off. Given HDD is still evolving there's still life in those platters (picked up a 4TB external USB drive last week - almost enough for my Steam games collection) . Enterprises are even more conscious of cost/performance ratios and HDD remains fast enough for the majority of use cases - we don't all need the performance of a Ferrari when we are shifting a transit van load around.

[Alan Brown]

Re: SSDvHDDvTape

"The death of tape hasn't happened yet so I would expect the death of spinning disc is some way off. "

There's an unspoken SPOF in the HDD industry: Platters and heads are both single-source items, so it doesn't matter that there are 2.5 OEMs left (Seagate/WD duopoly with Toshiba as a bit player) - we saw this effect in 2011 and HDD prices are _still_ higher than they were before the Thai floods, with significantly degraded warranties compared to 2011.

HDDs are evolving too slowly and SSD pricing is still falling. I believe the argument about fabs is wishful thinking as demand is elastic inasmuch as each decrease in price causes a non-linear (more like exponential) increase in demand, so more fabs won't cause a glut. Unlike RAM, NAND is a market with demand far higher than available supplies for the forseeable future and where bringing new capacity online effectively doesn't cause competition for sales.

Whilst SSD has the issue of cold storage(*), in all other respects it's far more robust than HDD whilst using less power. The only people arguing passionately for mechanical sales have vested interest in mechanical equipment or vendors.

(*) HDDs also have a problem in terms of "sometimes a cold storage drive doesn't spin up". SSDs can be used for cold storage as long as they're periodically powered up and tested/refreshed but in reality HDDs need this too (as does tape).

[DJSpuddyLizard]

Re: SSDvHDDvTape

The death of tape hasn't happened yet so I would expect the death of spinning disc is some way off.

But optical storage is dead. Archival Disc may have kept it alive as a medium, but now people can buy a 3TB HDD for $65, and it would take 30 BD-XL 3-layer disks (which cost over twice as much) to store the same data.

[John Dann]

Surely another dimension is whether the technology to build/operate FABs more cheaply is evolving too? No idea of the answer, but if it were possible to build NAND that's maybe a generation or two old, but at large scale and at eg 1/3 of the current price then the picture would change again.

[John Smith 19]

"whether the technology to build/operate FABs more cheaply is evolving too?"

AFAIK the answer is "no."

One of these SoA all bells & Whistles was $3Bn. I'd guess it's more like $4.5-5.0Bn by now and rising as the continued efforts to finally deliver the promised "Extreme UV" AKA Soft X-Ray lithography.

So no it does not look like they are getting any cheaper.

[Alan Brown]

Re: "whether the technology to build/operate FABs more cheaply is evolving too?"

" the continued efforts to finally deliver the promised "Extreme UV" AKA Soft X-Ray lithography."

Are not required for Flash.

Features smaller than 20nm are a liability in NAND, resulting in slower cells with lower endurance and significantly degraded cold-storage performance thanks to electron migration. There are strong arguments to stay at 40nm for the same reasons.

That's why 3D and chipstacking became the standard methods. The problems with parallax are well-defined and surmountable. Neither technology will significantly benefit from EUV litho.

[imanidiot]

Re: "whether the technology to build/operate FABs more cheaply is evolving too?"

Alan brown:

"" the continued efforts to finally deliver the promised "Extreme UV" AKA Soft X-Ray lithography."

Are not required for Flash.

Features smaller than 20nm are a liability in NAND, resulting in slower cells with lower endurance and significantly degraded cold-storage performance thanks to electron migration. There are strong arguments to stay at 40nm for the same reasons.

That's why 3D and chipstacking became the standard methods. The problems with parallax are well-defined and surmountable. Neither technology will significantly benefit from EUV litho.

In fact, most NAND production is done on old dry DUV tools. No need for immersion litho on NAND structures and the dry machines generally have a more robust uptime. I know of atleast one semicon fab using equipment produced in the early 2000s for NAND production.

The biggest hurdle slowing down NAND production uptick is simply available manufacturing for dry litho systems. Not many companies produce litho tools to begin with (only 3 players basically, Nikon, Canon and ASML. Of which ASML is the biggest due to better throughput performance per tool) I doubt all of them together manage more than maybe 500 tools a year in this segment.

If you really want to increase production 10 fold you need 10 times the machines already out there in the field. If there is a 1000 tools now (probably an underestimate) that means you need 10.000 systems. At current production levels it would take 20 years to meet that demand.

And that's just litho tools. The same constraints exist on all of the other equipment needed in a semicon fab. It's also not the kind of business where you can decide to increase production 10 fold in a year. I know for the tools I work on some of the critical supply chain path runs over 10 levels deep. Several of which are monopoly suppliers which won't like having to jump trough hoops just to get a small customer (for their business) some extra parts each month.

[Anonymous Coward]

Hmm, would the prices crash? The cost of entry is so high then prices wouldn't need to crash as they will still be controlled by a few big companies.

This pseudo cartel can control prices to ensure that they don't risk a loss. Any startup would struggle to build capacity while keeping prices well below market level and so wouldn't enter the market in the first place.

If prices crash and cause a loss it will only be because of a war of attrition with one big player trying to push the others out of the market but this wouldn't be sustainable and is more likely to reach a happy medium where costs are lowered gradually in line with new technologies but margins remain the similar.

I don't work for Gartner so this analysis is provided FoC.

[Eddy Ito]

But if the prices don't come down you'll still have a very large number of people sticking with the platter spinners and a whole stack of SSD/NAND piling up. One thing folks like investors don't like seeing on balance sheets is a large quantity of inventory that only grows over time.

It's a bit like the oil market has been for a few years. Some clever guys figured out an easier way to get oil out of the ground while prices were high so they did and at a very high rate to the point where it was becoming difficult to find places to store it all. Remember in 2014 started with a barrel price over $100 and it was about $50 at the end of 2014 and that's where it still sits even with a real cartel doing their best to keep prices up.

In short, I don't see the SSD players being able to sit on a large inventory like DeBeers can with diamonds because they'll be paying real money to have the manufacturing capacity and accumulating inventory doesn't pay the bills as long as there's a ready alternative in HDD. Diamonds come pretty cheap at the wages paid in underdeveloped countries and diamonds also have the advantage of being a Veblen good. DeBeers spends lots of marketing dollars to make sure you believe diamonds are rare and precious stones and that ones dug up are somehow better than the ones you grow in the garage.

[Alan Brown]

"But if the prices don't come down you'll still have a very large number of people sticking with the platter spinners and a whole stack of SSD/NAND piling up."

Prices are already coming down. The knee point for widespread adoption is about 3 times the cost of HDD and that's been surpassed at smaller sizes (you can pick up 256Gb SSD for 40quid in low end consumer drives). As soon as it happens in larger sizes, the market will buy them - the power savings in a data centre alone will pay the difference over a 5 year lifespan and our SSDs are (so far) showing failure rates about 1/5 of HDD, which translates to less labour overhead.

Due to the vast potential market, there's a non-linear relationship between price and demand. Every point decrease in price causes a multiple point increase in demand and this has shown no signs of abating for the last decade.

Once you hit a steady-state market (like RAM), then it's a different matter but you'd need to expand SSD production capability by a factor of 50+ to get close to reaching this and that's not going to happen anytime soon.

[Anonymous Coward]

> you can pick up 256Gb SSD for 40quid in low end consumer drives

Citation Needed. The cheapest I see on morecomputers.com is a Toshiba 256G at £75.46; Google Shopping finds a Sandisk 240G at £70.89

Most in this capacity are £85+

[Alan Brown]

"Remember in 2014 started with a barrel price over $100 and it was about $50 at the end of 2014"

Take a closer look at that situation. The low price was driven by conventional oil producers deliberately lowering their sale price to less than the production price of North Sea Brent Crude ($55/BBL), let alone shale ($65/bbl) or tar sands ($80-90/bbl) in order to bankrupt the tight oil producers and leave those conventional producers free to then increase prices and profit wildly.

That's classic cartel behaviour.

[Anonymous Coward]

Wrong about shale

A couple years ago some of the major shale players said most of their fields are profitable even with oil as low as $30-$40/bbl. Tar sands will always be high cost because of the amount of material that must be moved and the fact there are some steps requiring high temperatures even before it can enter a pipeline - and even then it is very heavy and much more expensive to refine.

The Saudis thought they could strangle US shale oil because originally it was being produced at $70/bbl, but they underestimated the ability for producers to become more efficient in their methods - both in identifying fields and extracting the oil and gas when found. The Chinese growth in oil consumption also slowed down considerably around the same time.

[Anonymous Coward]

Look at it this way.

Suppose Ford makes 10 million cars per year, and charges £10,000 each for them.

If it upped its capacity to 100 million cars per year, would it be able to sell them for £2,000 each?

Answer: probably not. There is a fundamental limit to the cost of production, including such things as the raw materials, beyond which scaling up doesn't help.

I think the same is true of SSD today. If you scale up your capacity by a factor of ten, that either means building ten factories where previously you had one, or one factory which is ten times larger; but most of your input costs are the same.

That's unless you believe that the SSD market is operating as a cartel, where all the players are making 300% profit margin; but surely the temptation for one of the players to steal market share by reducing margin (or for a new player to enter the market) would be too high.

Of course, there are periodic technology advances which reduce the production cost per megabyte; but such technology advances also occur in HDDs as well.

[Roland6]

Re: Look at it this way.

Suppose Ford makes 10 million cars per year, and charges £10,000 each for them.

If it upped its capacity to 100 million cars per year, would it be able to sell them for £2,000 each?

However, economies of scale do play a part, just as we see in the mobile phone market.

Suppose my facility has a capacity to make 100 million HHD's per year, which I can sell for 49 USD/GBP/whatever.

If however due to the take up of SSDs my demand falls by 20~30%, I suspect my prices would have to go up significantly; altering the SSD and HDD market pricing and thus demand...

I would not be surprised if we start seeing this effect in the low end, namely your typical sub 250GB laptop drive which is a prime candidate to have the HDD replaced by a SSD.

[Nik 2]

What's in it for me?

In order to generate the capacity required to justify the new fabs and hence lower prices, NAND needs to offer enough of a benefit that a large part of the customer base will pay the higher prices, at least in the short term.

The benefit of HDD over tape was enormous, and the opening of the home PC market allowed HDD to replace FDD as well. I don't see that NAND offers the same level of additional benefit.

Not to say that we won't get there eventually...

[big_D]

Re: What's in it for me?

Also, for films, music and photos etc. that you aren't currently editing, but looking at on occasion, the speed of an HDD is more than fast enough and it has the capacity to hold many times the data for the same price, so for slow sequential data access it is a cheap and relatively robust solution.

If I have a TB of films, music and photos, it is cheaper to archive them on an HDD than to buy a TB SSD, the performance will be more than adequate and there is an enormous saving at the moment.

We still use tape for backups today - we use a SAN, a backup SAN with regular images and that is written out to tape for disaster recovery / long term storage.

Currently, I see SSD for OS, applications and scratch / working data and HDD for data storage at home. I load my RAW photos onto my 250GB SSD, check them, edit them, then move those I want to keep to HDD and free up the space used. This gives me quick editing, but them cheap storage.

(In fact, the images are then copied to OneDrive and Carbonite and to a NAS for mobile availability and backup purposes.)

[defiler]

Re: What's in it for me?

The benefit is speed. And for the short term that's the only benefit. But it is a *huge* difference in speed.

For bulk, HDD will spin along for a while to come - that much is certain. As others have said, perhaps SSD will be "just another tier" - after all, you don't need to stream movies from SSD, for example. In that case you're probably better off with HDD and a lot of buffer RAM.

SSD is allowing huge data densities (at a cost), but I'm not convinced there's much appetite for that at the moment. Not whilst it's cheaper to just rack up the HDDs.

As always, the market will decide. I fully expect in 20 years HDDs will be an anachronism. But then we'll all be panicking about the Unix epoch rollover. :)

[Stuart Halliday]

Re: What's in it for me?

The other major benefit for users is the durability of SSDs. The huge number of home users who move their PC whilst it's switched on is massive in my experience. Eventually HDD will be used for backups mainly.

[Alan Brown]

Re: What's in it for me?

"I don't see that NAND offers the same level of additional benefit."

NAND isn't fragile, has better seek times/bandwidth and it uses significantly less power than HDD as well as having vastly lower failure rates in the warranty period.

Those advantages are more than enough to offset it being more expensive than HDD in datacentres. At 3 times the price of conventional drives it becomes a no-brainer to buy it.

At twice the price of conventional drives you'll see even budget PC-markers selling it as HDD failures are a significant part of warranty costs (a single warranty claim tends to wipe out profit form sale of 10 systems) and it's cheap insurance.

[Anonymous Coward]

Re: What's in it for me?

> At 3 times the price of conventional drives it becomes a no-brainer to buy it.

Really?

Suppose you need 10TB for a media server or backup server. The choices are £330 for a hard drive (today's price), or £1,000 (assuming SSD drops to x3 HDD price). Are you really going to pay the extra £670 if you don't need to?

> At twice the price of conventional drives you'll see even budget PC-markers selling it as HDD failures are a significant part of warranty costs

Citation needed; your implied assertion that SSDs are more reliable than HDDs is not that simple.

There is a decent Google-authored study linked from here:

http://www.zdnet.com/article/ssd-reliability-in-the-real-world-googles-experience/

"In summary, we find that the flash drives in our study

experience significantly lower replacement rates (within

their rated lifetime) than hard disk drives. On the downside,

they experience significantly higher rates of uncorrectable

errors than hard disk drives."

So for a consumer application, which is non-RAID, an SSD will appear to fail (i.e. will lose data) more frequently than a HDD, even if the entire drive doesn't die as often.

[Adam 1]

Re: What's in it for me?

A media server with 10TB of storage isn't mainstream. HDD will own that market for a few years. Whilst you are correct that SSD is more likely to just die without warning, you also assume there that a typical user will take action the first time they see an OS not detected press F1 message on boot or hear the click of death. Sorry, not buying it. My experience of typical users has been "oh yeah it did make a funny sound, blue screen, tell me there was no os last week, but I reboot it again and it seemed fine". Even a highly paid software engineer who I was working with (who definitely should have known better) had the click of death whilst I was checking something with her. I said that doesn't sound good. So she did absolutely nothing until a week later when it failed and she lost a day's work. So it's only an advantage if you act on it.

On price, the floor is much lower than a HDD. Whilst they can make a 32GB HDD, they can't do it at the price of the same capacity flash drive. At some point, the amount of storage that your Dell/hp whack into their desktops by default is going to be the same price point. The default purchase will then be a SSD, and you will flick to HDD if you need additional capacity. I don't think that is as far away as presented in the article.

[Alan Brown]

Re: What's in it for me?

"Suppose you need 10TB for a media server or backup server. The choices are £330 for a hard drive (today's price), or £1,000 (assuming SSD drops to x3 HDD price). Are you really going to pay the extra £670 if you don't need to?"

10TB is a tiny amount of storage in a datacentre. I have servers with more disk in them than that for local scratch space (and yes most of them are SSD, for speed)

The extra money on the disk side is offset by lower purchase costs for AC and lower power costs for both the drives and running that AC. When you have shelves of 40-60 drives drawing 1-2kW 24*7*365 and you can halve that, it's a significant operational saving.

[John Brown (no body)]

Re: What's in it for me?

"The benefit of HDD over tape was enormous, and the opening of the home PC market allowed HDD to replace FDD as well. I don't see that NAND offers the same level of additional benefit."

Yes, I agree, the difference isn't as a big a step change so if SSD does replace spinning rust, it will slow and gradual over many years.

[LeoP]

There is more to storage than performance

I am very sceptic of flash replacing disk altogteher as well. One of the points is, that flash performance is simply not needed for many applications. The video industry as an example has a highly sequential access pattern, enormous dataset sizes and a well-defined performance profile, that simply needs not be made faster. That's why they use disk for most of their data.

[John Robson]

Re: There is more to storage than performance

"I am very sceptic of flash replacing disk altogteher as well. One of the points is, that flash performance is simply not needed for many applications. The video industry as an example has a highly sequential access pattern, enormous dataset sizes and a well-defined performance profile, that simply needs not be made faster. That's why they use disk for most of their data."

Yep - idle power consumption is another pretty important factor, particularly in the low concurrency world of the home user.

If a flash manufacturer could get their head around that user group then we could potentially see some rather large, very lower power storage devices that don't have infinite IOPS, but are capable of pulling data in over a decent LAN connection and pushing out a couple of HD/4K streams simultaneously.

But Disk is dead - long live disk.

Tape is dead - long live tape.

SSDs are dead - long live SSD (although I suspect that SSDs are the most vulnerable of these to being replaced by something completely different)

[Naselus]

Re: There is more to storage than performance

This.

Every time we have one of these threads, it seems to be dominated by a lot of ridiculous commenters who think that the SSD in their laptop is somehow an analogy for worldwide storage demands.

It's not.

Those zottabytes of storage we generate a need for every year? Most of that is not speed-dependent. In fact, for the majority of it the only measure that matters at all is $/GB.Nothing else. I don't need to access an email archive from 12 years ago quickly. I just need to own it because of legal requirements, and so I need to store it as cheaply as I possibly can - and even if I'm doing so in the cloud, I'd rather be buying the cheapest nastiest chunk of storage on a JBOD rather than rapid access flash. So until flash can be sold at the same price per GB as HDD (meaning a fabrication method that allows it to surpass HDDs physical limits by a factor of ten or so, or a completely unforseeable paradigm shift in production methods, like learning how to grow it on trees or something), it just doesn't compete at all in the largest part of the storage market.

This is not going to change any time soon. SSD fabs cost around 10 times as much to set up as HDD fabs, and take four times as long to build and ramp up. That means you're waiting on ROI for a lot longer, since margins aren't much on either. IIRC, the oldest flash fabs in the world only started showing a genuine return (over setup and running costs) a couple of years ago. The majority of the industry still hasn't clawed back it's setup cost yet. HDD fabs, on the other hand, are in the black before the flash fab has even finished being built.

That fact alone means that HDD will not be disappearing in the next ten years or so. It might vanish from consumer devices, but will be ever-present in the data centre. This is not even a controversial argument in storage circles, it's just accepted and becomes more and more obvious the more you know about the subject.

The more interesting debate is whether flash will EVER replace HDD, or if something that replaces flash will show up first - something with flash's speed but with economics that allows it to be mass-produced cheaply enough to keep up with data demand.

[Peter2]

I've said for ages that HDD's will be around for a long time yet. Still plenty of development going, and when there isin't then the HDD manufacturers will almost certainly ditch all of their R&D and concentrate on ruthless cost cutting on producing their last generation of HDD's which will bring the price down and keep them competitive.

I'd guess that HDD's are still going to be around in 10-15 years, although they might end up being a much more niche market.

[Alan Brown]

" and when there isin't then the HDD manufacturers will almost certainly ditch all of their R&D and concentrate on ruthless cost cutting "

Guess what? That has _already_ happened - about 5 years ago.

HAMR and Shingling are the last steps. There's no more R&D for future technology coming down the pipeline. Getting HAMR from the lab to production is taking significantly longer than anticipated

[Craig 2]

I've always thought (like most things) the transition will be gradual. Ok, I may not have known the exact reasons why but even the blind squirrel finds a nut occasionally... :)

[BarryUK]

Basic economics

Basic economics would suggest that, at any point where companies believe a profit can be made from building another fab, then another fab will get built. So, while SSD keeps selling at a price where manufacturing is profitable companies will continue to expand capacity.

[returnofthemus]

.... coldish storage, and archival to HDD, which HDD still does better than tape.

LOL!

Really... I wonder in what way they mean?

PS Chris, where did you find this Anonymous Storage Chump, were they ex-EMC or from one of the last HDD manufacturer's left standing

[Anonymous South African Coward]

SSD's will take HDD's over when the fabbing process improves on yield and speed.

For now spinning rust trumps SSD's when it comes to the space:cost ratio.

[thondwe]

Flash replacing Tape too

Ignore the performance for a minute, flash needs less electricity and therefore less cooling, so it's going to make inroads into HDD and Tape purely on that basis.

Also moving data around on tapes, replacing old format tape with new format or keeping drives going are all pain points too.

So expect SSD et al manufacturing to increase. Other industries see this happen - Power (Wind/Solar replacing coal/oil/etc), Car's going electric. It's not whether more how soon...

[Charles 9]

Re: Flash replacing Tape too

The thing is, people have found that cycling backups on occasion has benefits as well. As you move the backups from one tape generation to the next every few years, you also help to maintain its integrity by verifying the backups and resetting the cold storage clock. Plus tape technology (particularly LTO) evolved to produce additional advantages: such as high data transfer rates that can surpass most rust drives and cramming more data in the same physical size than most other formats. Plus it still has its advantages regarding cold storage (rust drives apart from RDX can't be guaranteed to keep their data intact long-term).

My thought is that flash is gaining ground on rust, but rust won't go away until flash hits a price:capacity tipping point (say twice rust's price for the same amount of storage). Until then, rust's raw capacity advantage is still useful, and although so much storage in rust has issues, there are already ways around them (most notably transfer bit rot; but that's why we have error codes).

[Alan Brown]

Re: Flash replacing Tape too

"most notably transfer bit rot; but that's why we have error codes"

Those error codes aren't good enough in large capacity HDDs. We're seeing one silent error (as in not picked up by HDD internal ECC) every 45TB read on average.

Thankfully ZFS picks this up and corrects it, but not all storage systems do.

[Charles 9]

Re: Flash replacing Tape too

I'm not talking the stuff built into the drives. ZFS, as you noted, includes advanced error codes so can pick up on silent corruption, and that's one way to do it. I'm taking including things like Parity Archives for distinct finalized data sets that are being archived or (say media collections) won't be changing around. That way, even if bit rot hits during a transfer, you can usually correct for them and get back in business. You use up a little capacity to cover your butt over bit rot. This as well as pairing drives are perhaps the best way to keep consomer-level data (as in nice to keep intact but not life-and-death) protected against wear and tear.

[Doctor Syntax]

Re: Flash replacing Tape too

"tape technology... evolved to produce additional advantages: such as high data transfer rates that can surpass most rust drives"

Isn't tape also based on rust?

[Charles 9]

Re: Flash replacing Tape too

Point taken on the rust, but the rust in tape is in the cartridge, not the drive assembly, so my point still stands. Besides, we're talking linear rust instead of spinning rust along a more precise and controllable mechanism.

[Anonymous Coward]

Re: Flash replacing Tape too

Its going to have to be a lot of power and cooling or a massive drop in price to make it worthwhile to replace tapes with SSDs. Especially when tapes use no power at all and generate no heat at all when stored and have a RAW capacity right now (LTO10) of 48TB, with newer tapes demonstrated this year with 330TB of capacity.

[Peter2]

Re: Flash replacing Tape too

SSD's aren't going to be replacing tape anytime in the forseeable future. This is why:-

4TB SSD. £1300 ex VAT & delivery.

6TB LTO7 tape. £89.95 ex VAT & delivery. (50% bigger capacity, 14 times cheaper)

5 day tape rotation with 6TB LTO7's. £449.75 ex vat.

5 day disc rotation with 4TB SSD's. £6,500 ex vat.

Both figures feature RRP pricing. Yes, you can probably get either cheaper if you've got a decent relationship with a supplier and are happy to turn the thumbscrews on them a bit, but RRP is comparitive.

[Anonymous Coward]

Re: Flash replacing Tape too

> [tapes] have a RAW capacity right now (LTO10) of 48TB

For some strange definition of "right now".

"Right now" you can buy LTO-7 drives (6TB). LTO-8 is pre-release, and may start sampling around the end of the year.

A tape drive which is three generations in the future might exist in a lab somewhere, but it is not something you can pick up off the shelf of PC World.

[Matthew 17]

I've been told that SSD isn't good for cold data storage

As you need to power them up every few months or they lose their contents.

I've a beer-cooler box in my loft with a load of HDD backups of my work based on this, hopefully they'll survive a long time :)

[Peter Gathercole]

Re: I've been told that SSD isn't good for cold data storage @Matthew

Just make sure that you also store disk controllers as well.

Modern PCs have largely lost their EIDE adapters, and you can't have a PCI adapter, as again, modern PCs don't have PCI slots. And probably PCIe will change so that any PCIe EIDE disk adapters won't be usable.

I think that you'll find that if you have old SCSI disks that aren't LVD SCSI, you may already have difficulty hooking them up to a modern PC, and forget anything that's IDE, ESDI, SMD or ST506. Most likely the PCIe adapters for these controller types don't exist.

[Mario Becroft]

Re: I've been told that SSD isn't good for cold data storage @Matthew

Actually, SCSI has been extremely long-lived, SAS is simply SCSI over a different physical layer. Controllers/adapters to support plain old parallel SCSI going right back to the 80's are easily obtained. I should know something about this, I designed SCSI controller ASICs *cough* years ago.

I can't imagine that anyone sensible would ever have used [E]IDE in the enterprise.

Consumer home users; well that's a different kettle of fish entirely, and for the majority of home users who do only light tasks, choice of storage is not something they know or care much about. With the accelerating move to web-based SaaS applications and things like tablets and Google Chromebooks, combined with ubiquitous high-speed network access (even when away from home/office, i.e. HSPA, LTE,) local storage for the bulk of consumers and even business end-user devices is becoming rapidly less relevant.

[Charles 9]

Re: I've been told that SSD isn't good for cold data storage @Matthew

"With the accelerating move to web-based SaaS applications and things like tablets and Google Chromebooks, combined with ubiquitous high-speed network access (even when away from home/office, i.e. HSPA, LTE,) local storage for the bulk of consumers and even business end-user devices is becoming rapidly less relevant."

ORLY? Have you seen the data allowances these stingy providers foist upon you? Cloud seems fine and dandy until you learn you can only partake a teaspoon at a time without incurring massive overage fees. Plus what if you CAN'T access the Internet due to being in a not-spot or whatever? No, the only source you can really trust is one YOU control end to end.

[Alan Brown]

Re: I've been told that SSD isn't good for cold data storage

"As you need to power them up every few months or they lose their contents."

You need to do something similar with HDDs anyway - controllers sometimes stop working and bearings can freeze up.

[kwhitefoot]

Re: I've been told that SSD isn't good for cold data storage

Temperature variation will probably kill those drives. Lofts are bad places to keep anything temperature sensitive.

[Merrill]

What is needed is better data destruction policies

No one, including me, will ever read the roughly 100,000 files that I've collected. There is simply no reason to review the past.

Lots of storage will be freed up as people die.

[frank ly]

Re: What is needed is better data destruction policies

"Lots of storage will be freed up as people die."

This was my grandfather's HDD. It still works fine and I see no reason to junk it.

[defiler]

Re: This was my grandfather's HDD.

Not a Deskstar then?

[TheRealRoland]

Re: This was my grandfather's HDD.

Kalok, right? Both the name of the brand, and the sound they would make when it broke...

[bombastic bob]

Re: This was my grandfather's HDD.

"Kalok, right? Both the name of the brand, and the sound they would make when it broke..."

I bought a Kalok drive in the 90's. When I loaded a beta version of Win '95 it started swapping to disk a LOT, which caused the drive to eat itself. Sounded a lot like breaking glass.

(fortunately, I did regular backups, and only lost time and the cost of a new hard drive)

Prior to that, I had left a Seagate drive on top of the case, with extended length IDE+power cables going into the back of the computer (a secondary hard drive). I stored "work files" on it, mostly (whatever I was working on at the time, source code etc.) and did regular backups. But whenever the drive had been off for a while, after being on for a while, it wouldn't spin up. But the spindle stuck out through the circuit board, just a little. So I'd grab it with pliers and manually spin it until it free'd up, then power up the drive. It worked for about a year like this until I finally replaced it.

[Mario Becroft]

Re: This was my grandfather's HDD.

Yeah! In data recovery I've popped the lid on a few HDD's over the years, giving the spinde a gentle push to overcome stiction. Certain 2.5" laptop disks appeared particularly prone to this issue back in the day, perhaps due to lower power/smaller/lower torque spindle motors, or plain bad mechanical design.

I tend to think the article missed the mark somewhat on where storage is going. NAND Flash is cheap(ish) now but new technologies such as phase-change memory will probably supersede Flash *and* possibly spinning disks for some (most?) applications in the next 10-20 years.

As for tape, it's here to stay for some time. Not much else can store 150+ TB (300+ TB demonstrated recently by IBM and Sony) and stream at insanely high speed on a single small, robust, portable cartridge.

[Doctor Syntax]

Re: What is needed is better data destruction policies

"Lots of storage will be freed up as people die."

There's a very long term storage format available: ink on parchment. A lot of that must have been freed up when people died. If you're an historian that's one of the central problems of your professional life.

OTOH I think present day rates of data accumulation will also be a problem for future historians.

[MOV r0,r0]

They'll kill themselves, eventually

The longer HDDs persist the more the capital cost of their research & production is paid off - they'll just keep getting bigger and cheaper. Eventually that approach will fail (for whatever reason) resulting in insufficient profit margin to provide a return on capital invested and that is what will do them in.

The 'will they die' and the 'what will kill them' are best treated as separate questions.

[Charles 9]

Re: They'll kill themselves, eventually

There are hints that rust is approaching the point of diminishing returns. Technologies such as HAMR and shingled recording are pretty complicated beasts compared to the technologies of the past. Additional points of failure loom, so one has to wonder how much time the state of rust art has actually bought.

[Alan Brown]

Re: They'll kill themselves, eventually

"they'll just keep getting bigger and cheaper. "

Platter densities haven't changed appreciably for several years. That will change with HAMR but it's "not ready yet" and has been that way for about as long as platter density has been stagnating.

[Anonymous Cowerd]

warnings

One major issue with SSDs apart from price - SSDs give no warning of their impending failure. You just lose the data. At least an HDD usually gives some indication that it's dying, so you have time to get any unbacked-up data off it.

[Charles 9]

Re: warnings

Not always. Rust drives can suffer sudden catastrophic controller failure, too. Happened to me years back.

[Anonymous Coward]

Re: warnings

"Rust drives can suffer sudden catastrophic controller failure, too"

Sure, it can happen, as you say...

"Happened to me years back."

It's happened to me too, also many years ago - once. All the other HDD failures I've had have given some warning and allowed me to recover the data.

In my experience, SSD failures result in the immediate loss of everything stored on that SSD, with no chance of recovering anything, whereas only some HDD failures result in immediate loss.

Whilst I like the low power draw and speed of SSDs they seem like a technology that's only viable when pushed to fragility. I look forward to a better solution but can't help feeling that the current investment in NAND is holding up progress to something better.

[itzman]

Re: warnings

I dont think that is intrinsic to the technology though.

Wear levelling will use up blocks till there's no uncorrupted blocks left to use up, and could easily flag the fact.

[dmacleo]

Re: warnings

on mine intel ssd manager does give you warnings.

[Alan Brown]

Re: warnings

"SSDs give no warning of their impending failure. You just lose the data"

For the most part, "wrong". You just weren't paying attention to the error messages.

"At least an HDD usually gives some indication that it's dying"

Also for the most part "wrong". s/usually/about half the time/

In any case, that's why you have RAID (for hardware failure) and backups (for RAID failure or finger trouble)

[TheElder]

Re: warnings

SSD gives plenty of warning. You just need to keep an eye on basic error counts. Plenty of nice free software is available for that such as CrystalDiskInfo.

[John Smith 19]

Remeber folks HDD is a great archive medium, much better than any old tape....

Until it gets a head crash or the electronics goes.

Then of course you're f**ked big time.

But hey, it's real cheap.

[bombastic bob]

Re: Remeber folks HDD is a great archive medium, much better than any old tape....

well, sounds like backups and archiving ought to be a multi-plan thing then.

a) store on multiple machines/hard drives/SAN/etc.

b) periodic tape/dvd/SD card/whatever archival

c) cloudy backup for when it's appropriate

that's kinda what I do...

[Charles 9]

Re: Remeber folks HDD is a great archive medium, much better than any old tape....

Well, for me, I'm not that paranoid, but I DO take precautions for data I'd prefer not to lose. My drives are paired up and rotated periodically (mirroring as I do so) to guard against a controller failure. Bit rot I cover by using parity archives.

[Norman Nescio]

Tepid storage for home users

What I'd like is some reliable, long-term (archival) storage for home users. Something with a vast amount of capacity that I can write, lose in the back of a desk drawer for 20-30 years and reliably read afterwards, without (like a tape drive) needing a rather expensive bit of kit to read and write it.

Low density SLC 3D-NAND might do it, although cosmic rays and static discharges from nylon carpets could be an issue. I don't know if charge retention works for that duration. It could certainly be engineered to work, but current JEDEC standards, as far as I know, allow enterprise-class SSDs to have a data retention period of as little as 3-months at the END of their useful life. (There's a discussion on this on the Micron website here : https://www.micron.com/about/blogs/2015/may/addressing-data-retention-in-ssds ). What your average consumer USB stick is specified for/ capable of is anybody's guess. As ever, storing at low temperatures seems like a good idea - i.e. not left on a sunny window-[c/s]ill.

Using hard-drives for archival storage is fraught with problems - relying on a precision engineered electromechanical device over that lifetime would be brave. The usual answer is to keep multiple copies and copy over to new media well before the rated storage lifetime - which is obviously not what you want for something you want to forget at the bottom of a filing cabinet for a couple of decades. Permanent ink on good quality acid-free paper stock does remarkably well, except for the storage density requirement.

While not for home use, it is interesting to see people thinking out of the box for long-term storage. How about writing QR-codes to microfilm? See here: http://bpexchange.org.test.ibiblio.org/wp-content/uploads/2015/05/alternativedatastorage.pdf

TL;DR : Utah State Archive evaluated writing QR codes to microfilm, and said it was possible, but currently uneconomic. E.g. of the order of USD 100,000 per Terabyte.

Oh - and why microfilm? - Using the right materials and stored correctly, it is expected to last 500 years, yet can be read with very simple equipment.

[Anonymous Coward]

Re: Tepid storage for home users

I have a low-tech cold/tepid storage system at home. I use Reed-Solomon codes to create "shares" that I distribute over the LAN. The idea is that you can create n shares, where any k of them can be recombined to restore the data (n>=k, with n-k = number of tolerated failures). I've got a bunch of Raspberry Pi boards (among other ARM boards), each one attached to an external USB drive. A bunch of the drives are powered by a single ATX power supply that I can turn on/off over the network via a little circuit board attached to a Pi. Besides the ARM boards, I've got a bunch of old desktop and server machines that are in hibernate mode and can be woken up with a WakeOnLan signal, so their internal spinny drives can also take part in the RS/redundant storage network.

It all works quite well for me, though it's not really as polished or painless as I'd like sometimes, and powering up all the machines when I want to add/update files means it's not something I want to do too often. Recovering files from it isn't as bad because I only need to have k shares available.

Enrolling files has another slight problem in that the bandwidth required to shift files is n/k times the input file size. I messed around with multicast transfer (udpcast) but it didn't seem robust enough.

The major advantage here is cost. The USB drives are at or near end of life and I can make n-k large enough that I can manage the eventual failures without too much bother. So, the drives are effectively free apart from my time cost and slightly higher up-front costs for the USB enclosure. Pi boards are cheap and have other uses, and old PC hardware is also effectively free and running costs there aren't too important because the vast majority of the time they're only drawing standby power waiting for the WoL signal.

[kwhitefoot]

Re: Tepid storage for home users

Perhaps you should write it up as an Instructor or a Github repository.

[Charles 9]

Re: Tepid storage for home users

And what's the assurance THEY'LL still be around in a decade or two? Sites come and go, after all.

[Alan Brown]

Re: Tepid storage for home users

"Permanent ink on good quality acid-free paper stock does remarkably well, except for the storage density requirement."

This brings up a personal side note. BIC's "Wild Blue" ballpoint pens of the 1980s which were all the rage for a while.

I archived a bunch of notes made during classes of that period. 10-15 years later when I needed to refer to them the ink had gone entirely transparent (you could see the marks on the paper, but the ink was _gone_). Other ballpoint colours were more-or-less intact (but faded)

It was my first introduction to the idea that writing notes may not be a permanent record after all. Over the years I've run into many written records from the 80s which suffered the same fate.

[Doctor Syntax]

Re: Tepid storage for home users

"What I'd like is some reliable, long-term (archival) storage for home users. Something with a vast amount of capacity that I can write, lose in the back of a desk drawer for 20-30 years and reliably read afterwards, without (like a tape drive) needing a rather expensive bit of kit to read and write it."

The problem with this is that not only do you have to provide the equipment for 20-30 years but you also have to be able to understand it. So even if you resort to writing QR codes to microfilm you then have to hope that whatever OS you're using in 20-30 years has a library for decoding QR codes, even if the data is just plain old ASCII text that you could have streamed off a tape providing you had a tape drive.

There are no easy solutions for long term data storage except active curation: copying from the old medium and format to the latest one whilst the old is still physically and logically accessible.

[sisk]

I expect to see SSD replace HDD eventually, but it's not going to happen soon. Maybe (MAYBE) by 2030, but definitely not by 2020. There's just too much of a price difference between the two. It's going to take time for SSD technology to improve to the point that they can match the price of HDDs, and 3 years is not anywhere near enough time for that to happen.

Of course by 2050 we'll have kids who've never seen a computer with a HDD just like we've got kids today who've never seen a floppy disk.

[Anonymous Coward]

Use M-Disc. Reported life time is 1000 years, even buying one of the drives they sell on site is cheap (<$70), 4.7 GB DVD-R on up to 100 GB quad-layer BluRay capacity, and the media is inexpensive as well. Burn, toss in the back of a drawer. And they've been around long enough that if it were "snake oil", we should have heard by now.

When I spec'ed out my new game machine back in 2009, being able to take advantage of M-Disc was right up near the top of the list. It wasn't even expensive, especially compared to a tape drive which I would still love to have here given my past experience working with mainframes from the '70's but those are still too damn expensive. Just about my only worry would be being able to have a system around that could talk to the drive in the distant mists of time. However, that's a trick with all long-term cold storage.

[Norman Nescio]

Re: M-disc

@Jack of Shadows

Sounds good. What will be the availability of M-Disc drives in 1000 years? Or even 10 years? I got burnt on MO drives. I have some 230 MByte MO discs which, no doubt, have the data encoded extremely well, but obtaining a drive to read them...? (I had backups on more traditional media, so the information is not lost). It has made me wary of non-mainstream media and drives (what alerted me was the almost complete lack of vendors for new 230 Mbyte MO media).

It made me take a rule of thumb not to rely on media that require a particular mechanical operation: buses can be emulated - by soldering and jump leads, if necessary - but high-precision mechanical components for positioning head assemblies are not something that can be kludged together.

[Charles 9]

Re: M-disc

For me, it was Zip drives. It was the best option for me at the time because I was working with a 486SLC laptop with no CD-ROM capacity. Thankfully, at that point, technology was pretty fluid and by the time Zip's writing was on the wall, I had access to other techs to offload things before I lost too much.

[Doctor Syntax]

Re: M-disc

"For me, it was Zip drives."

I had to buy a drive which got used just once. My daughter had data on a disc and needed to get it off onto something more tractable.

[Alan Brown]

Re: M-disc

"I got burnt on MO drives. I have some 230 MByte MO discs which, no doubt, have the data encoded extremely well, but obtaining a drive to read them...?"

You didn't get burned on MO drives.

You got burned on assuming that you'd never need to migrate the data to newer media.

One of my cow-orkers has 2 filing cabinets full of exabyte archive tapes. She assumes that when the time comes to read them, she'll be able to do so and repeatedly declined our requests to migrate the content to LTO (the whole lot would fit on one LTO6 tape).

Our last exabyte drive died 5 years ago. Another group has one, but it's been switched off and on a shelf for years. It may (or may not) work when switched on, but given it has a UW-scsi connector and we don't have any way of interfacing to it, the point is moot.

[Lee D]

All I know is:

I will pay for a decent SIZE (I'm more than happy with basic SSD speeds for most things) SSD, even if it comes in a 3.5" form factor.

The ones I have are literally tiny aluminium boxes (not even sealed, just pop them open) with - inside them - a couple of chips on a PCB that take up about 1/3rd of a 2.5" drive space.

I would buy them by the hundreds, put them in every client, for a "free" speed upgrade. But the capacity is too small at the moment.

Servers, etc. - that's a specialist area. Literally, who cares what they use because you replace it and they are already stupendously fast and bottleneck on network, etc. in most installs rather than the drives themselves.

But I would pay through the nose for a decent capacity hard-drive-replacement SSD that was even, say, 500MB/s read/write. It would be like upgrading every machine in the place to double-speed. I know, I've done it one some of our kit with the cheapiest of cheap SATA SSDs with tiny capacities. The only problem - capacities. You also have a guaranteed replacement revenue stream five years from now if SSD lifetimes really are that limited (to be honest - I replace every machine every four years anyway, drive and all, so I really don't care).

What I hate is that ALL I want is capacity, and all people want to sell me is EVEN MORE SPEED on interfaces I don't have (NVMe etc.), in tinier and tinier formats that - collectively - could give me a 20Tb drive in a 2.5" configuration no problem, even with current tech. I really don't care about that.

NAND, SMAND, CAND, I couldn't care less. 500MB read/write. In a standard SATA format. 5-year lifetime is fine. BUT AT LEAST 1TB. At some vaguely reasonable price I would literally buy them by the dozens of hundreds. Use cheap chips, tons of them, with over-provision and error-correction so I get 1TB of usable space for 5-years. I really don't care.

But the focus on fancy interfaces, faster interconnects, more "standards", and still no improvement in price/capacity when my SATA SSD is basically a half-empty box, and the only thing that SSDs don't have is capacity is really annoying.

To be honest, make them cheap and shit and I'll just keep replacing. The speed upgrade of just maxing out a basic SATA bus is more than worth the effort. And personally, for every server drive I have, there are dozens of client drives that I could be replacing which would make a much bigger difference to the en-user.

Lose your market when it goes cheap? No. I'll start snapping them up in bulk. At the moment, about 128Gb per client is all I can run to and that's barely adequate for modern usage.

[A-nonCoward]

without the disk-drive makers becoming shirtless, history you don't know meseems

Forgetting how many disk-drive makers went under, Chris sayeth,

once upon a time disk manufacturing capacity was less than tape and disk successfully replaced tape for mainstream storage without the disk-drive makers becoming shirtless

The only ones that thrived were the ones that could spend zillions to make new technology and fabs. The lucky ones got swallowed, most staff sent to rot.

Fabs get more and more expensive, inverse proportion to size of dies, chips, whatever gets smaller but costs more to start manufacturing, in the hope that you will recoup that larger investment by selling astronomical amounts and that your prices won't be undercut. At some moment, the cost is too high to take the risk. Seen that happen. BTW, older fabs are very cheap to purchase, still good for not-the-latest chip making, thus lotsa weird clones for low-end and heavy competitive manufacturing.

BTW, Chris is quite OK if he has friends of this quality, as the gentleman being interviewed really makes sense, you don't need to know everything. Can we corroborate the numbers independently?

[I Like Heckling]

Oh no they won't...

Until Price per GB falls by hundreds of %... HDD are here to stay... and by that I mean when you can pick up a 4TB SSD for less than £100-115, then those that require proper storage capacity rather than speed for a few bits of software and to be impressed by how quick their PC boots... will still require a HHD.

What's more of a concern is the pricing cartel that we've had since the 2011 floods, a cartel that has artificially kept prices high long after production returned to normal and going against the trends of the last 20yrs that saw HHD prices fall in line with larger capacity drives being introduced. Now those larger capacity drives get introduced at ever higher prices and never drop more than a few % over the course of several years... a 4TB drive I paid £115 for in 2013, is still between £110-120 today... Yet every other drive I've ever purchased over the previous 19yrs has double capacity for roughly the same price as the previous one... going back to a 2GB drive in the later half of the 90's... £60.. 6GB.. £60... 20GB... £65.... 40GB.. £65 and so forth and so forth... right up to those floods of 2011. Pre-flood prices the last drives I had purchased was a couple of 1.5TB & a 2TB drives... for guess what... £75 each (gotta take into account a little inflation over almost 15yrs).

These days, there's a jump of around £30-35 per TB across the board... so that 4TB at around £115...... the 5TB at around £150 and the 6TB at around £185... and there those prices remain for the last few years... In spite of current production levels being high and manufacturing costs being as low as pre-flood prices. It's no more than price gouging, trying to convince people that SSD's are getting more competitive by keeping HHD prices artificially high.

So when I can buy a 4TB SSD for less than a 4TB HDD... then and only then will I believe any codswallop about HHD being a dead format.

[Anonymous Coward]

Size matters

Will the percentages shift? Absolutely. Will the technologies evolve? Without a doubt. Will HDDs die? No more than tape.

]Amazon's supply chain includes container ships, semis, vans, and, eventually, drones. But the are not going to use drones to get stock from suppliers or container ships to deliver Amazon Prime orders.

Amazon Web Services moves data from DRAM to NAND to HDD to tape (and back again). This is not going to change either.

[Anonymous Coward]

"Storage media industry insider" = shill from Seagate or WD?

SSD technology will continue to improve. HDD can only compete on cost now i.e. cost per unit storage.

There will come a time when SSD is the de facto, ubiquitous 'no brainer' choice when picking a storage drive. And people will only pick a HDD when they really want to be stingy and save some money, and even the money saved isn't worth the downgrade.

As the market responds, the manufacturing shifts away from HDDs to SSDs, resulting in HDDs becoming increasingly harder to find.

Historical example. I am old enough to have seen and used a CRT display. In some ways, the CRT display is still superior to LCD/LED displays. However, it is hard to find a CRT display now, even a refurbished one in satisfactory working condition. The manufacturing has stopped, the world has moved on.

[Charles 9]

Re: "Storage media industry insider" = shill from Seagate or WD?

"Historical example. I am old enough to have seen and used a CRT display. In some ways, the CRT display is still superior to LCD/LED displays. However, it is hard to find a CRT display now, even a refurbished one in satisfactory working condition. The manufacturing has stopped, the world has moved on."

I grew up with CRTs, too, but after seeing the transition to LCDs, I can see why CRT's disadvantages outweighed the advantages. They're bulky (takes up precious desk space), heavy (transport and labor costs, potential desk issues), and not as precise (an analog device by nature). Once LCDs overcame problems of contrast, size, and weight, they could outpoint CRTs 4 times out of 5.

[Aus Tech]

Re: However, it is hard to find a CRT display now

I have just what you are looking for, sitting in my back shed, an old 21 inch CRT monitor, still in working order, AFAIK. I haven't used it in over a year, and it is heavy. It has a VGA connector, not sure if there is any other more recent than that. Last time that I moved it, it took a good 15 minutes to go less than 50 meters. I had to stop something like 12 times before getting it into the shed. At a guess, it would probably cost more to ship it somewhere than what it is worth??? There's also an old Dell 15 inch monitor down there, that was last used with MS-DOS 6.22 and Windows 3.1. I think that donating them to a Museum would be better than trying to sell them.

[Anonymous Coward]

Working v archival storage = constant v growing demand

My prediction: SSD use will grow rapidly in the short term as it becomes the primary working storage for most computing applications, then demand will rapidly level off. However it may never displace HDD for archival storage, for which capacity demand is steadily growing.

Fast forward a few years from now, say I want to upgrade / replace my PC. My old SSD may still be working just fine, the speed and capacity are sufficient (if it gets full, move less accessed data to HDD), so why not re-use it? However if my HDD is full I have no choice but to buy another one. Admittedly the rate at which I buy HDDs has slowed because their capacity has increased by more than I need, but I wont stop buying them. But I only need 1 SSD as my working drive.

[Brian Allan 1]

Ultimately something with no moving parts will replace HDD's; not sure what?

I've lived through the eras of paper tape, punch cards, magnetic cards, magnetic tape (in many flavors), floppies (in many flavors), HDD's (in many flavors), CD's, DVD's and SSD's. All of which had their time in the marketplace. At some point something will replace HDD's and even SSD's. Technology marches on and new tech becomes old tech...

[prof_peter]

HDD/Flash and technological evolution

"Will NAND flash replace disk?" isn't really a valid question - the question is "(when) will NAND flash replace disk for application X?", and there are hundreds of applications out there with different answers.

If disks weren't increasing their capacity/price ratio as fast as flash, the answer would be simpler - given enough time, flash would be better for everything. But they're increasing at about the same speed, so the most storage-hungry applications are going to use disk far into the future, while others tip over to flash sooner. In particular, for apps with bounded requirements, once flash becomes cost-effective then it's all over for disk, because you only get the cost/GB that disk offers if you need an entire disk.

For iPods the answer was 2005, which curiously happens to be the year that 2GB of flash became as cheap as a micro disk drive. Every year after, instead of increasing the capacity of their base model as disks grew, they could keep it the same and pay less for flash.

For a wider market the transition is less abrupt - e.g. what's been happening with laptops in the last 9 years since Apple introduced their 2nd-gen Macbook Air. (about 45% of laptops are expected to ship with SSDs this year; the rest are evidently sold to 4chan trolls who need the extra storage for all the porn they download...)

For just storing lots and lots of data (like that Microsoft OneDrive account you haven't touched in a couple of years, or the Dropbox folder from a project you finished last year) hard drives are going to remain king for years and years, possibly becoming weird and specialized in the process because they don't have to remain plug-compatible with your old machine.

Finally, flash has had the speed advantage since it was introduced; if you have a small amount of data and you need it to be fast, you're stupid to put it on a disk. However as the size of that data grows it starts becoming cost-effective to play all sorts of caching and tiering games, so that you can use disk for capacity and flash for performance.

[Anonymous Coward]

Ah but there is always the big unknown factor based in consumer products that can have a substantial impact. Take for example the ipod which started with rust but then pushed nand production to it's limits.

[Thekatman]

You can't have interplanetary space travel with spinning disk.

You cannot have interplanetary space travel with spinning disk, tape libraries, CPUs that need cooling and other such current technology. Virtualization is a start, but we will need to have data stored on devices that do not need cooling, stores EB on a small device at some level other than magnetic) and can, in fact, operate with super high MBTFs at room temps. Spinning disk will be obsolete, not by SSDs, but by a follow-on data storage technology.

[neilfleminghpe]

Just a point on the speed thing....it really depends what you are comparing. 1x SSD vs 1x HDD, yes the SSD wins the speed contest hands down.

However, in my environment I have SSDs where I get ~500-550MBs sequential performance. I also have a RAID array of HDDs, that cost less than one of those SSD (significantly less), delivers 500-550MBs sequential as well, but also has data redundancy.

The power debate is also somewhat moot - all my HDD systems power down the HDDs when not in use. That makes them very power efficient. As they are storing data that is accessed rarely that is *most* of the time....but when it is accessed it requires fast sequential access...which a RAID group of HDDs gives in spades.

[Anonymous Coward]

That may fit you, but what about if your data accesses are mostly random? Or spaced out in such a way that you have to keep waiting for drives to spin up, costing you valuable time? SSDs trump even rust RAIDs in both those departments since solid state drives don't have to wait for things to spin up.

[Kiwi]

That may fit you, but what about if your data accesses are mostly random? Or spaced out in such a way that you have to keep waiting for drives to spin up, costing you valuable time? SSDs trump even rust RAIDs in both those departments since solid state drives don't have to wait for things to spin up.

They may take longer to spin up... But I can fit at least 8TB of rust in my laptop, maybe 12 if I convert my DVD bay to a HDD bay (based on 4tb being the largest I've seen in 2.5 hd's to date). I'd need to build a mansion to get that much data space with SSD.

As the previous poster said, your needs are not mine. You need a teency bit of data really really quickly, I need to have tons of stuff archived which might not be touched for years at a time. When the data has been waiting 5 years to be accessed, what's a 5-second spin-up time matter?

[Anonymous Coward]

"When the data has been waiting 5 years to be accessed, what's a 5-second spin-up time matter?"

A LOT when you need it in two or less before the rival beats you to the punch. In today's world, lag time matters. Why do you think so many financial firms vie to get so close to the New York Stock Exchange? Because they're SO cutthroat they're vying for shorter cable runs necessary to eke those those precious microseconds difference caused by the speed of light or electricity. And since these can be make-or-break propositions, money is not quite an object in this case.