Notebook flash SSD market: fantasy or mirage?

by Robin Harris on Sunday, 27 April, 2008

Fresh off the HD-DVD fiasco, Toshiba execs are stepping up to pursue another expensive flop: notebook SSDs. Memo to Toshiba: people won’t pay huge SSD premiums for nothing. And almost nothing is what flash SSDs provide today – and for the foreseeable future.

Please sir, may I have another!
Given the multi-billion dollar cost of semiconductor fabs, getting the notebook SSD market wrong would make Toshiba’s $250 million HD-DVD loss look cheap. The president of Toshiba semi, Shozo Saito, recently opined that flash drives will be in 25% of notebooks by beginning 2011.

He is so-o-o wrong.

Hand me the back of the envelope, please
Guessing 200M notebook sales in 2011, 50 million flash drives of, say 250 GB, for total sales of 12.5 million TB of flash. Assuming a cost reduction curve of 50% annually from today’s spot market MLC $2500/TB to ~$320/TB in 2011 . . . hmm-m . . . $4 billion in chip sales.

Give or take. Yummy!

If Toshiba projects winning 20% of the market, $800 million in sales would justify over $1 billion in flash factory capacity. And if the market doesn’t appear, a billion dollar write off.

Same power, same performance and way more costly – I’m sold!
If flash drives delivered what proponents claim there would be no problem. But they don’t and they won’t.

Power: no SSD notebook has gained more than 10 minutes battery life over disks. Since flash is already power-efficient that won’t change. Disks have multiple opportunities to improve power use – and with over a $1 billion a year in R&D behind them – they will.

Performance: tested application performance hardly changes either – even with a $3,800 flash drive. Notebook I/O doesn’t favor flash drives – and the engineering contortions needed to fix flash aren’t cheap.

The one big win for flash performance: boot and app load times. It makes the system feel a lot snappier – if you often reboot. Sleep mode makes that much less important.

Reliability/durability: flash vendors tout 2 million hour MTBFs and superior shock & vibe specs. Yet Dell reports that their SSD infant failure rates are about the same as disks. And the return rates are higher.

So where, exactly, is the flash advantage? Plus, it is only conjecture that flash drives will prove to be more reliable in actual notebook use. Only time will tell.

And what about the 4-bit MLC that Toshiba is counting on to drive costs down at 40-50% per year? This will less durable than current SLC. No hard numbers from the vendors – depends on how good their signal processing algorithms are – but it could easily be 5,000 writes – down from 10,000 today.

How do you explain that to consumers?

Data integrity: the unasked question Of all the questions about flash drives, this is the biggest. I have yet to see an SSD read error spec.

Flash has read errors – that’s why vendors implement error detection.

But flash has a problem disks don’t: flash drives move your data around a lot more often than disks do. Every time a flash drive writes a page, it has to erase the entire block that page is in.

So what happens to the data in the block? It gets read – almost always correctly – and rewritten along with the new page. The new location must be tracked by the drive.

The map that keeps track of where your data is rapidly gets very complex – and itself is regularly read and rewritten. How well protected is this critical data structure? If it isn’t bulletproof you can kiss your data good bye.

If FTL’s are like every other storage product, catastrophic failure modes are hiding in the statistical weeds. Enterprise IT is rightly suspicious of storage that “auto-magically” moves data around. Consumers have no idea. SSD vendors better have their act together or the class action suits could be as big a problem as the empty fabs.

The StorageMojo take
The further I wade into flash issues, the worse it gets. My sense is that the flash industry close to creating a multi-billion dollar fiasco. Why?

  • Over-promising on performance, reliability, battery life and data integrity. Take a systems level perspective, folks. Consumers do.
  • Over-broad positioning of flash drives as a general replacement for notebook hard drives – when pricing clearly says they aren’t.
  • Relying on system OEMs like Dell to market SSDs to consumers is a freeway to failure. They don’t have the bandwidth. The flash vendors need to market flash SSDs directly to consumers. Not sell them – market them.

The flash guys are caught in a vise: big expensive fabs that need to run all year; and seasonal demand that whipsaws their pricing all year.

Notebook flash drives can help even out demand – but only if consumers accept them for the right reasons. Otherwise Toshiba’s new fabs will build chips for a non-existent market.

Update: Flash has a place in one notebook niche: below the $40-$50 minimum cost of a disk. As we’re already seeing with the Asus Eee, replacing $50 of disk with $10 of flash makes a big price difference. But those units won’t solve the seasonality problem and may even make it worse. End update.

Comments welcome, of course.

{ 12 comments… read them below or add one }

the storage anarchist Monday, 28 April, 2008 at 5:01 am

Just to be sure I understand the target of your concerns: you’re talking exclusively about MLC-based Flash SSDs targeted at the laptop market, right?

Enterprise-class Flash Drives, such as the ZeusIOPS from STEC already address all of your performance and reliability concerns, albeit at a higher price point.

You are making the distinction between the two different classes of Flash SSDs, aren’t you?

Bill Todd Monday, 28 April, 2008 at 5:49 am

As one who personally has little interest in flash drives for notebooks right now (nor much use for it as a general-purpose disk replacement for the foreseeable future), I feel fairly well-positioned to call you on what looks like a fair amount of FUD.

1. “no SSD notebook has gained more than 10 minutes battery life over disks”

Rubbish. The anandtech.com link that you provide in the very next paragraph mentions a 5% – 15% improvement in battery life for one experimenter (and if you actually follow the relevant link the detailed numbers turn out to be 11% – 16.8%, resulting in increased run times of 16 to 43 minutes on the MacBook Air being tested): did you even read it?

It’s also important to consider the new range of LED-backlighted models with a relatively modest (i.e., very-low-power, which today implies a modest clock-rate) processor. Those are the wave of the future – because they’ll allow use of far lighter-weight batteries *plus* increased working time (a recent Li-polymer design breakthrough may make battery weight and size much less important – but only if it becomes practical, and projections are that it won’t by 2011).

2. “Disks have multiple opportunities to improve power use – and with over a $1 billion a year in R&D behind them – they will”

Says who? Just spinning the platter(s) already consumes most of the power used by a disk (especially in typical non-seek-intensive laptop applications), so the only obvious way to reduce that is to reduce air resistance: do you have *any* basis for believing that it’s not already optimized in this area?

3. “tested application performance hardly changes”

Obviously, that would depend fairly heavily on the application. For example, a read-mostly random-access-intensive application like a database too large to fit in RAM would benefit *greatly* from a flash drive, your anandtech.com link measured a 20x improvement in random 4KB read performance over a conventional hard drive, and one of the links in your Dell citation noted a 35% overall performance edge over a conventional 5400 rpm laptop drive measured by SYSmark ’07 (and that the SSD performance on average even exceeded that of a 10Krpm desktop drive).

This brings up a point related to power as well: RAM consumes enough power to become a fairly important component in that respect after display (especially if you don’t need the latest, greatest, and therefore power-hungriest in graphics performance), processor, and disk power consumption have been minimized – and flash storage potentially allows use of far *less* RAM in your notebook without compromising performance (you may need little more than enough to create a decent-sized write-back cache in front of the flash).

4. “Dell reports that their SSD infant failure rates are about the same as disks. And the return rates are higher”

No, they don’t. Your link states “Our global reliability data shows that SSD drives are equal to or better than traditional hard disk drives we’ve shipped”, and the Marc Farley blog to which it links indicates (also without quantification) that reliability (as measured by returns) is perhaps a bit *better* than that of hard drives. More importantly, these results are obtained by a skewed customer population that to a significant extent has paid a premium for flash drives *because their operating environments are hostile to conventional drives* – i.e., an apples-to-apples comparison in the same environments would likely prove that flash holds a *considerable* reliability advantage over conventional disk there.

And that article does not state that ‘return rates are higher’: one might suspect it given the wording, but in the context one can also strongly infer that *if* they are higher this has nothing to do with relative reliability/durability (the context in which you present this assertion).

5. “it is only conjecture that flash drives will prove to be more reliable in actual notebook use”

Actually, it’s considerably more than ‘conjecture’: it’s a direct consequence of dramatically decreased susceptibility to shock damage (especially while operating).

Not to mention the resulting improvement in peace-of-mind – at least I know that *I* worry about shock when using a laptop, and it’s mostly worry about whether it’ll crash my drive heads (not that the rest of the box is indestructible, of course, but it’s pretty immune to the minor shocks that are the most common).

6. “This will less durable than current SLC”

All you can reasonably project is that it will be less durable than the future SLC that’s contemporary with it – and (as you seem to admit) you don’t know how durable that may be.

But the fact that *you* don’t happen to know that vendors will be able to build adequately cost/performance-effective units doesn’t mean squat: if *they* believe that they can to the extent that they’ll sink large amounts of money into doing so, chances are that they’re right (and there’ll be plenty of independent testing done to establish whether they are before consumers have to worry about the issue).

7. “I have yet to see an SSD read error spec”

So ask, already: meanwhile, don’t babble about dangers that you can only imagine.

(I assume that you’re talking about *undetected* read errors, since they’re the only ones with potentially visible consequences.)

8. “How well protected is this critical data structure?”

That’s possibly the clearest example of pure FUD in your post. Again, if you don’t know – ask, and meanwhile, stop babbling about imaginary dangers.

(Sure, this data structure should be well-protected – but even Microsoft took steps long ago to protect similarly-important data structures like the good old FAT, so assuming that contemporary flash drive vendors are too stupid to do so doesn’t seem very reasonable – and any that are will likely be taken care of by natural selection long before they become threats to consumers who purchase name-brand products.)

9. “If FTL’s are like every other storage product”

They aren’t, though, are they? For example, they have no moving parts and in consequence have *dramatically* simpler firmware.

So this generalization becomes just more FUD.

10. “Enterprise IT is rightly suspicious of storage that “auto-magically” moves data around”

Then I guess they don’t use disks, which will happily revector bad sectors when they discover them.

If you were referring to storage that moves data around *between* devices (e.g., in response to load fluctuations), then that’s not at issue here, is it? The flash behavior that you’re talking about is much more analogous to that of log-structured storage: are you asserting that ‘enterprise IT is rightly suspicious of’, say, NetApp’s products?

To sum up:

As far as I can tell, you’re just annoyed about the amount of hype that flash is getting when contrasted with certain holes in immediate real justification for it:

a) its write performance may often not be much (or even any) better than a disk’s, which means that in many common applications (especially in environments with large amounts of read cache fronting the storage) using it won’t see much improvement over using a disk (though even there there may be considerable room for improvement if the file system is optimized to use flash storage rather than disk storage),

b) no *major* reduction in laptop battery consumption can be claimed (at least not with *currently typical* laptop configurations), and

c) other claimed areas of superiority have not been adequately quantified for your taste (though my impression is that few other people have any reservations about flash’s dramatically superior shock-resistance).

Now, I’m not a fan of hype and would also be critical in such areas, though more careful to stick to facts (rather than FUD) in presenting them.

But ‘hype’ is your middle name: with many other technologies you embrace it, regurgitate it enthusiastically at any opportunity, and vigorously resist far more reasonable attempts to tone it down than those you’ve made here regarding flash (your affection for ZFS comes to mind, but you seem to lean that way for many new, shiny things that cross your path unless their provenance is fairly obviously somewhat dubious).

What is it about flash that so irritates you?

- bill

hirni Monday, 28 April, 2008 at 6:38 am

@Bill:
so what exactly is the URE rate for flash ?
From my theoretical point – if you don’t have any ECC (which costs bits => less usable capacity) you can’t even detect it… and that worries me.
Sure – we could use filesystems which fixes this shortcoming – but then again – I’m overpaying and feel cheated.

If my camera screws up on one file some bits – fine – the jpeg-format recognizes it – and I can visually recognize it.
But with office-docs ? – very tricky …

(on hdds – I’m md5-summing my backups to be able to later know whether it’s corrupt or not – although I can’t fix it if corrupted)

OTOH – as I wrote in this blog’s comments on earlier articles, the whole “flash economy” will only take off if:

a.) it’s in absolute prices cheapter than a HDD
(absolute price means a size capable of holding the “full OS + data”)
(for Linux this is at around 4G – like the EeePC demonstrated right)

b.) it’s 100% plug+play replacement for a spinning HDD
(I must be able to test it before I switch – and I want a fallback if it sucks)

Current SSDs do not do well on point-a, while they do it right on point-b.

Flash-cards (SD/CF) are getting to the point, where a computer (not XP) can boot from them and use it as their root-FS.
So that’s where the price of SATA-SSDs has to be to make it worth to try.
As long as it’s not there – I see no real case to use flash.

Bill Todd Monday, 28 April, 2008 at 9:54 am

hirni -

URE (sometimes UER) is to the best of my knowledge the *uncorrectable* error rate, not the *undetectable* error rate that I mentioned. The former is hard enough to ascertain for conventional disks (though it’s still buried in their product manuals if you look hard enough); the latter I haven’t been able to find anywhere for close to a decade (it used to be about three orders of magnitude lower, but the gap could have narrowed as densities continued to increase). And I’ve never bothered to try to find either for flash devices – but Robin should if he’s going to try to use them as debating points.

I’m pretty sure that flash already contains error correction bits for its contents, so while the uncorrectable error rate is still significant in determining just how much additional redundancy higher layers of the system should provide for a given level of availability, it’s the undetectable error rate that’s the most pernicious (just as it is with disks). Fortunately, more correction bits can address both problems if that’s necessary, just as they have with disks for decades (density increases have tended to require more correction bits over time, if only to keep the need to retry marginal reads very low even if all the bits on the track were correct) – and since flash should only have to worry about stored bits being correct rather than the additional problem that they aren’t always *read* correctly as they speed by, other things being equal it should need fewer correction bits than a disk needs.

As for the flash economy ‘taking off’, it already has. Flash satisfies enough needs better than disks do for its future to be secure – it just doesn’t satisfy *all* needs better than (or even as well as) disks do, and may never do so (price per bit likely being one of the harder nuts for it to crack).

- bill

Andrey Kuzmin Monday, 28 April, 2008 at 1:02 pm

> The map that keeps track of where your data is rapidly gets very complex – and
> itself is regularly read and rewritten. How well protected is this critical data
> structure? If it isn’t bulletproof you can kiss your data good bye.
In modern flash file-systems, this concern is being mitigated using COW technique, similar to WAFL and the rest of the advanced file-system world.

Bill Todd Monday, 28 April, 2008 at 8:24 pm

Andrey -

I think the potential (though I suspect already well-understood and solved) problem to which Robin was referring occurs at a low enough level that such file system designs can’t affect it noticeably (unless they interact with the flash drive at a level lower – and far more device-specific – than that of a conventional disk, which is certainly not true of laptops today).

- bill

Jonas Lerch Tuesday, 29 April, 2008 at 5:34 am
MS Bob Tuesday, 29 April, 2008 at 11:52 am

Interesting article, perhaps overly negative. Some points (sticking to the portable sector):

1. “Disks will get better power over time… with $1 billion in R&D.” Maybe. It depends on the application. Laptops have had pressure to reduce power consumption since the year dot. Do you know of any improvements in that area over the last several years?

2. “App and boot loading times much reduced”. For me this would be the killer feature. In running low performance laptops (either new, or old as in my case) the most frustrating thing is app loading times. Boot I can deal with, that only has to be done once. It is every time that I open up openoffice or amarok that is the killer. If the app load times were reduced, the performance would be imperceptible from that of a modern performance CPU.

That is where the future market is, Fast enough for common tasks, light, cheap, no moving parts, no noise and 8+ hours of use without recharge. Works for portables, HTPC, kid’s computers, etc. Any bulk data gets stored to HDD, OS runs from flash.

The next question is, what are the reliability downsides and how to get around them? As far as I am concerned, the jury is still out for me, I’ll let the early adopters suck it and see. In a year or so we’ll know more about reliability and prices will have dropped. Win win.

Or there is always the possibility of trying to have your cake and eat it too. Why not have some sort of fast flash (e.g. on USB) for unchanging app directories and HDD for frequent writes/large storage needs?

e2eiod Tuesday, 29 April, 2008 at 10:54 pm

If Flash is the Storage “fly-by-wire” precursor where is the “Flying Wing”?
Maybe I missed it?
E2EIoD is my concept of End-2-End Information on Demand, which I haven’t seen yet either, because it needs the Storage “Flying Wing” which needs “fly-by-wire” control systems.
My name is Robert Pearson.

Randy Robertson Saturday, 3 May, 2008 at 2:14 pm

This post is very FUDy, do you work for Seagate or something? Flash will have a place in future storage- a much bigger place than it has currently. Traditional HDD will still be best for ultra high capacity needs. Mobile devices will move to all flash, except for perhaps giant desktop-replacement laptops.

Flash and SSD are just different elements of the storage stack. It is like comparing main memory to CPU cache. They have different uses and complement each other.

Here is the future storage/memory stack:
CPU Registers -> L1 Cache -> L2-L3 Caches -> Main Memory -> Phase Change memory -> Permanent Storage Onboard Caches -> Solid State Flash in permanent storage -> Traditional HDDs.

Some systems like laptops won’t bother with the last link, since they have power and shock resistance concerns.

e2eiod Tuesday, 6 May, 2008 at 7:06 pm

RE: “This post is very FUDy”… I do not work for any Storage device, software or process vendor.

The “Flying Wing” was a fundamental shift in the design of airframes. It was pretty much unusable by humans until “fly-by-wire” flight control systems came into being. It is not all things to all people even though our latest and greatest aircraft all use variants of it. All these variants have to use “fly-by-wire”. Humans are becoming retardants in the performance of aircraft. Just like in Storage…

In my mind this relates to Storage starting with the circa 1999 phrase “we can no longer manage Storage at the spindle level”. In the same time period people began to talk about “Storage being managed by the acre” in the case of Deutsche Telekom tape Storage.

Managing Storage by the acre, or above the “spindle” level, seems to beyond the desires of most humans. IMHO, Storage is still managed at the “hands-on”, if not the spindle, level.

Flash gives us a Unit of Technology that has a chance of providing “fly-by-wire” Storage Management. Unless it is retarded by only being used as a “rotating rust” replacement.
Not the manifest Destiny I would pick for it.

How well will your very fine “future storage/memory stack” handle Storage Management by Content? Not just Content Addressable but real Content that is addressable by many means. Because it is device and geographically dispersed. Many degrees of physical separation that must appear to be no more than six. I think that is the Storage “Flying Wing”.

The least Unit of Information is all that is important and the Managed Unit of Information is so valuable that by rule it must be stored on a Managed Unit of Technology.

james braselton Friday, 14 January, 2011 at 9:14 pm

hi there you right if apple sony dell and nintendo hard drive will be out of buissneiss in no time at all resistance is futile why do apple have macbook air macbook pro imac and mac pros with the macbook air flash drives are standard not optional other apple computers have ssd solid state drive options 128 256 or 512 gb ssd sony vaio p series has 64 128 or 256 gb ssd sony vaio x 128 gb ssd standar sony vaio w 128 256 or 512 gb ssd option no hdd option and dell has the dell adamo with a 128 gb ssd standard not optional and alien ware multi ssd options dell has other laptops with ssd options too

Leave a Comment

Previous post:

Next post: