SSD technology is doomed

At 6.5nm, it says capacities will go to 4TB. This presumably is in a 2.5" enclosure. That's the maximum for 3.5" platter drives right now. It also says the latencies only doubled. Read/write latencies are in the order of milliseconds and you can bet the engineers building them will find solutions to any of the problems along the way. Even if we have to stop at 1-2TB, it doesn't matter, only the price matters. They can already fit 512GB onto a card that is small enough to go in a Macbook Air:

http://www.anandtech.com/show/4606/s...2gb-capacities

Intel is planning to have their SSDs up to 800GB this year.

Your implication is that tech won't change between now and then, and I don't buy that. Eventually we'll have different types of solid state storage than NAND, and many of the concerns with modern drives won't even exist anymore.

Cloud.

By 2022 we should be close to most areas having 500mbit/sec connections. (Edit: removed wrong calculations)

Even now at home with a measly 15mbit/sec connection I literally don't have to store any iTunes Store videos locally. I choose it and it streams faster than realtime on Mac and iOS devices.

My estimates are very low ball, this paper http://www.campustechnologies.net/pd..._paper_CTI.pdf estimates that by just 2017, the OECD ~average~ bandwidth will be almost 800mbit/sec. That is much faster than most ~hard disks~ in 2012 (when comparing to real world transfer rates).

Steve Jobs predicted this 15 years ago, whereby with network caching everything becomes faster than hitting the disk drive for data:
http://www.youtube.com/watch?v=yE3Ta_NK4-I

So yes, SSD is doomed, but not because of physical constraints.

By 2020 getting something over the network will be faster than reading it from a 7200rpm hard disk in 2012. Along with the compression you can do with CPU/GPUs in 2020, as well as new network technologies... Well, the rest is, as they say... history.

Regular hard disks are certainly doomed for mainstream portable computing by 2016:
http://www.supremelaw.org/systems/io....annotated.PNG

Even with my 15mbit/sec various tasks on my laptop with my 7200rpm drive, the Internet is waiting for me, not the other way 'round. Just got an SSD, amazing stuff. And there's a decent solid 10 years of SSD tech to make local storage incredibly fast, so along with the cloud we should see a convergence around 2022 whereby network and local storage is fully optimised on both sides, by no doubt Apple and maybe some other players, who knows.

Actually, that convergence could be around 2016, it's happening real fast now. By 2022 it could be that most servers will use highly optimised RAID-etc. SSDs (as they are increasingly used now) and most data for client devices will be streamed. Again, smart caching, insane compression (imagine by 2022, a BluRay-quality video could easily be 100MB) and network engineering will make all this possible.

That statement is so technically wrong the only thing I can say is that there is nothing redeeming in it at all. Period.



The only redeeming part of that post is that NAND flash work started in the 70's. Wow, magnetic tape storage started in the 1920's and is still viable today for the right applications.

You are just a very good Chicken Little simulator. One who obviously hasn't been paying ANY attention to the upcoming technologies in computing over the past 5-10 years. And also a Chicken Little who is completely forgetting about the end of line for computing because we have already run into (oops passed anyway) the first two un-passable physics limitation Moore's Law walls.

Tech is all about S-curves. You can count on that both in the META and in the micro. Just because a micro S-curve curve shows an approximate future limitation date doesn't mean the META S-curve made out of a succeeding family of those micro S-curves will hit it's hard limit too.

SSD latencies are currently about 50 microseconds. That's about 100x faster than the average rotating hard drive, so a worst case slowdown of 2x still makes for a solution that is 50x faster than what everyone finds acceptable today for secondary storage.

Capacities will continue growing, what the grad student and other researchers forgot to say is that the deadline is only in the face of currently known manufacturing techniques. 2024 could be a problem if nobody works on these things, but I doubt that will be the case. They also forget that circuit density is only one aspect of device capacity. Packaging can be changed to put bigger dies into only slightly larger packages. That would require some re-plumbing of the on-die I/O, but there isn't any manufacturing limits on that. Those kind of shenanigans can add a generation or two of cheap extension time for alternative research. Wow, then todays problem is pushed to 2027-2030. That would give SSDs a market leading life about the same as CDs, hardly a thing to get all worked up about today.

To boot, I think memristor arrays will be taking over by then, leap-frogging all the density, current and lifetime issues of flash memory arrays.

Cloud will have plenty of bandwidth. The problem is latency and that is what makes computing slow down. I don't think average cloud latencies will ever consistently be within an order of magnitude of local secondary storage latencies. 1-5ms latency is a wildly best case based on my data always being only a single router hop or two away on a lightly loaded server with the data already in RAM. More like 5+ms if the server has to pull my data from a drive array. put a whole bunch of customers on that server and watch average first byte delivered latencies rise even more.

We need a new idea of a file system that can expand the virtual capacity of your local storage and handle predictively archiving things and parts of files to the cloud that don't need to be stored locally to avoid the latency hits. Kind of keep the short latency required stuff locally and handle the rest/remainder as streams. We have a ways to go, both technically and trust-wise, to get to the place this can work generally. I think Apple is maybe trying parts of these techniques in their cloud streaming of media.

While I share your sentiment here (and some of your concern), I don't agree with your overarching implication (being that the only means by which anyone should be storing data in the future is local, spinning disk hard drives).

You seem to care a lot about not using SSDs, though


That's what lawsuits and governmental protection are for. Yeah, we'll see cloud services slowly grow to extortionate prices, but when there are free alternatives, they'll be artificially forced downward.

iCloud's going to be sued? Huh. Maybe Apple shouldn't have done it, then.

You need to stop making your hats out of metal. iCloud is free. It is one of the free alternatives to Carbonite and that jazz.

I'd ask what makes you think what you've said up there is anywhere near the truth, and I'd ask why you think cartel is spelled with a 'k', but it's obvious you've some paranoia issues.

Oh, never mind. Maybe you're Dutch and this is some big electoral push I don't understand.

And so what's your evidence of any of this happening ever?

iTools: Always free, never paid
.Mac: Always paid
MobileMe: Always paid
iCloud: Free

History tells us that it will remain free in perpetuity. If they want people to keep using it at all, they'll make it free forever. Period. It's illegal to one day start charging for something, holding one's data hostage, and since they'd have to give advance notice, you'd be able to remove all your files before the dreaded cut-off date. It's a non-issue.

This won't happen. Are you a shill for the spinning hard drive lobby? You seem irrationally against absolutely every future technology that looks to supplant it.

SSDs put us at the mercy of cloud operators?

And with that I disagree on the face of it.


By 'premier', do you mean primary? I agree with that. But again, it's moot since SSDs will be fine.


Google already does. Their stuff is free, but they steal your information "legally" in accordance with your terms of service. Apple doesn't. Neither have any intention to monetize their stuff. Google because they already steal your information and sell you ads based on it. Apple because you're already giving them money for hardware and they actually have a design team.

Claiming that iCloud will go paid with zero proof to back that up is nonsense.

You need to knock off the knocks toward religion. And no, it's not circular reasoning. I do not believe SSDs are a failure, nor are they doomed. Therefore there is no need to worry about expanding local storage in the future.


You do not and can not know this. It won't go paid without tens of millions of lost accounts and huge controversy that irreparably damages Apple. It can easily stay free with minimal issue.

iCloud is free if you use an Apple device/OS with it. That means the maintenance of it is considered an advertising expense for the devices. Apple is selling so many devices the cost amortizes out quite small per device/OS.

You can use it with any device, even a PC, and not have an iOS device at all. You can't upload iWork documents without having paid for iWork, but if you have a PC, you're not going to be doing that anyway.

Are you seriously going to listen to a graduate STUDENT who probably hasnt worked in the industry for long ( or none at all)?

Dont you think the people in the industry knows about the future limits?

This lady obviously wants her name out in the field for a prospective job offer from a company or is looking for a faculty position ( if not a PhD program).

Again, people move by self-interest only.

Corner cases that do almost nothing. iCloud connected to iTunes on a PC, what then if you don't have an iOS device? nada...

The point of my post wasn't that though, it was that iCloud isn't "free". It is an explicit part of the overall Apple marketing scheme and funded as such via sales of devices and software. When you buy Apple, you are paying for iCloud access whether you use it or not.