h a l f b a k e r yGetting blown into traffic is never fun.
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
There currently exists in the marketplace, in several varieties, a box you can buy that is full of RAM chips, and which "looks", to the Operating System, just like a hard disk drive. Some versions of these things plug directly into a PCI slot on the computer motherboard; some are the same size and shape
as ordinary disk drives; many come in a number of configurations with respect to the total amount of RAM. ALL are hideously expensive, compared to other computer hardware. Consider a
drive-sized unit: it has a relatively ordinary steel or
aluminum box, a mini-motherboard into which RAM modules can be plugged, circuitry to communicate with the computer's disk system (usually SCSI, but also available in Fibre-Channel, IDE, FireWire, USB, and others), and a built-in backup power supply (sometimes this is optional). How much do you suppose the bare box-and-board should cost (yes, with no RAM!)? Well, they
actually cost at least ten times that amount!
I'm sure that some of that excess cost can be explained in
terms of chicken-and-egg/supply-and-demand: As long as the demand is low, there is no incentive in invest in the mass manufacturing that would bring the price down, and as long as the price is high, demand remains low. I think, however, that if a manufacturer knew that demand was about to ramp up, then the appropriate investment in mass manufacturing might occur, leading to lower prices. What I'd like to do here is explain exactly why everyone should want every PC to have a solid state drive, thereby perhaps creating the demand...(YES, I'm writing this because I want some that I can afford!)
Have you experienced an ordinary failure of an ordinary
hard disk drive? Most long-time computer users have, I
think. Simply stated, they are fundamentally mechanical
gadgets, and will inevitably wear out -- and when they do,
people tend to lose important data. The crucial thing that
a solid state disk drive can bring to your computer is a
probable doubling (or tripling, or more) of the lifespan of
your hard disk. It does require a minor paradigm shift in
how your computer is used, though.
Normally, when you install a new piece of software, it gets
put directly onto the hard disk. With a solid state disk,
you instead will install the software there. It will be up
to the people who write installation software to follow this
plan: (1) Offer the option to install to a solid state disk;
(2) obtain the designation for that disk from the user (you); (3) install to the solid state disk; (4) copy the entire installation to the hard disk as a single compressed block; (5) provide a program-activation icon (or equivalent). When you decide to run that newly installed software, you click the icon (or equivalent) in the usual way. The computer is then told to transfer and uncompress the saved block to the solid state disk, where it is mapped out as a completely normal installation. The hard disk can now be nearly shut down, preventing wear-and-tear, while you use the program normally. Only when you need to save some data should the hard disk be fired up again, while working with the program (and that can be avoided if you can configure things so that the default place to save data is some removable medium, like a CD-R or a magneto-optical disk). When you are done with
the program, you exit normally, but the program actually and simply deletes its entire installation from the solid state disk, in preparation for the next thing you might do.
EVERY program, including the operating system(!) should be designed to work that way! (The OS, of course, will auto-load, and not need an icon.) Consider what happens if there is a power failure: All previously-saved data is safe, since it is never saved to the solid state disk (which forgets everything when power is lost). Only any data that you might be in the middle of creating would be lost (a quite typical situation for anyone whose computer isn't plugged into an Uninterruptable Power Supply). Here is where the option NOT to buy a power pack for the solid state drive is a money-saving choice! If your whole computer is protected by a UPS, you simply don't need it, and if your computer is not protected, well, you are probably used to making frequent data-saves, so that little is ever lost because of a power failure! (NOTE: UPS equipment has become remarkably less expensive recently, so I recommend getting one for the whole computer.) Anyway, should you lose power, you will of course experience the typical delays while the system reboots, and some additional delay while re-loading whatever program you were in the midst of using. But that is nothing new, right?
Since the just-described scheme for using a computer changes the paradigm for accessing the hard disk from vast numbers of small saves and loads, to tiny numbers of large saves and loads, it should be quite apparent that your mechanical drive will be enabled to enjoy a long long lifespan. I consider this drastic reduction in dead-drive aggravation to be more valuable than the most-often-advertised feature of a solid state disk drive: SPEED. These things are typically a thousand times faster than
hard disks. Your software will run so smoothly you will wonder why you ever put up with the wait, as the next-chunk-of-data is loaded from a hard disk in today's paradigm.
Now, about that initial assertion that you should want to buy a bare box-and-board: RAM modules are standardized and nicely inexpensive these days, and hold nice large quantities of data (half-Gigabyte modules are readily available for less than $100, and I remember paying that much for 1/10,000 of that much RAM). There is no reason to pay the manufacturer of solid state disk drives for middleman-priced RAM. And, as long as the RAM modules are matched for speed, it doesn't really matter to the average PC owner what speed of RAM gets plugged into that bare box-and-board -- ALL RAM is about a thousand times faster than a hard disk!
Thus the point here is that you should be able to get your
initial solid state disk for a modest cost, and it can be
immediately available for use by many programs (think how many programs are smaller than half a Gig!) -- and you can easily upgrade to more RAM, in affordable stages. The "natural" maximum amount of RAM for a solid state disk drive is 4 Gigs, and hardly any ordinary consumer software needs that much, neither for installation nor operation. You will be able to run a decent number of programs simultaneously, with a maxed-out solid state disk drive. Just like you do now, only faster.
Barring the specifically-described software that makes it easy to install programs to a solid state disk, you can still make it work: You just have to pay attention during today's typical install process, and make sure you specify the solid state disk as the installation destination. Then you must make sure the computer is not rebooted until after you have copied the entire installation directory "tree' over to the hard disk. When you want to run the program, simply copy that tree from the hard
disk to the solid state disk before clicking the start-icon.
You will also have to make sure any data-saves are defaulted to some place other than the solid-state disk, as already described. Have fun!
But first, of course, we all have to tell the manufacturers
how much we want to switch to this new paradigm...so that the drives become affordable, so that we can make our hard disk drives last longer, and enjoy faster computing, as well. Feel free to copy this text and post it anywhere, or post links to it from elsewhere. Thanks!
RAM Drive
http://bugclub.org/...ndows/RamDrive.html
Semi-related [half, Oct 04 2004, last modified Oct 21 2004]
Solid State Disks
http://www.memorypl...m/main/ss-intro.htm
[Shz, Oct 04 2004, last modified Oct 21 2004]
Really Ancient Magnetic Storage
http://online.redwo...hr/vol1_1/daveb.htm
The Earth's crust as magnetic media... [Vernon, Oct 04 2004, last modified Oct 21 2004]
Magnetism's optical effect
http://www.physik.f...i/snom/#whatfaraday
The principle by which magnetic data can be optically scanned. [Vernon, Oct 04 2004, last modified Oct 21 2004]
Basic magneto-optical info
http://www.usbyte.c...ommon/MOsystems.htm
Lasers, Magnetic media, and the Curie Point [Vernon, Oct 04 2004, last modified Oct 21 2004]
Magnetic RAM
http://computer.how...works.com/mram1.htm
One type of forthcoming nonvolatile memory [Vernon, Oct 04 2004, last modified Oct 21 2004]
Ferroelectric RAM
http://www.nightfli...andom+Access+Memory
Another type of nonvolatile RAM, and a kind of descendent of ancient "core" memory mentioned by Phoenix. :) [Vernon]
Future alternative to Hard Disk
http://www.inphase-...ies.com/technology/
Multi-terabyte capacities forseen. [Vernon, Oct 04 2004, last modified Oct 21 2004]
Future alternative to Hard Disk
http://www.inphase-...ies.com/technology/
Multi-terabyte capacities forseen. [Vernon, Oct 04 2004]
Possible future of magneto-optics
http://www.thic.org...erastor.gknight.pdf
They keep thinking up new tricks [Vernon, Oct 04 2004, last modified Oct 21 2004]
Future of Hard Disks
http://www.nanotech...mrc-2002-report.htm
The whole industry is working on this trick. [Vernon, Oct 04 2004, last modified Oct 21 2004]
About "Long Term" data storage
http://www.macdirec...views/MO/index.html
An independent review of the options, with magneto-optical emerging as the winner, for most of the same reasons presented elsewhere here. [Vernon, Oct 04 2004, last modified Oct 21 2004]
Gigabyte iRam
http://www.gigabyte...aspx?ProductID=2180
This newcomer to the solid-state-disk market is designed to be inexpensive. Usually available for $200 or less. Could be cheaper, though; as the main text describes, there is no real need for that built-in backup battery. [Vernon, Oct 18 2006]
L Computers' PuRAM
http://www.l-computer.com
You have to kind of dig around for the PuRAM info, as the site is in various stages of (re)construction. Some stuff seems up-to-date, others not. [Agamemnon, Feb 21 2007]
[link]
|
| |
Just like hard drives, I've had a lot of RAM failures over the years, too. |
|
| |
It sounds very much like a ram drive. Unfortunately, I haven't seen much improvement in ramdrive over the years, and it still holds only a very small amount of data. (nothing useful anymore). Right now, the limiting factor of most hard drives is the interface, not the drive itself. For something around a gig, I just don't see something like this being worthwile. |
|
| |
Zzzz. I wish I had the time and patience to read this whole thing so I could refute (?) it point-by-point. I guess the short answer (and one you should already know) is cost. |
|
| |
BTW, there's an annotation floating around somewhere where I mention software on (flash memory) cartridges. Why install software at all? This has the additional benefit of letting you keep your data and applications together all the time, if you choose. You don't even need to worry if your friend/client has the application you want to use - just bring your software with you. All you need is a computer with a number of Nintendo-like slots. If manufacturers controlled access to the blank cartridges, pirating software would be all but impossible. |
|
| |
Worldgineer, I suppose this is a kind of consumer advice. However, the real idea here is the modified paradigm for using ordinary computers that have been souped up with SSDs, so that hard disks will last longer. Please do read it and see if you would like most of your software to run lots faster. The consumer advice thing is merely to promote demand, because it is demand that causes production that reduces price. |
|
| |
bristolz, yes, RAM does fail occasionally. However, any RAM that lasts longer than the warranty period also tends to last for many years. The existing makers of SSDs like to brag about million-hour lifespans. |
|
| |
freefall, the traditional definition of a "RAM drive" is to set aside some of the computer system's RAM specifically to act as a disk drive, and yes, they were often quite low-capacity. Today, if you want to shell out the $, you can get an SSD that can hold quite a few gigabytes. 4GB is a fairly common maximum, though. I do think you are mistaken about the interface. Every time they ramp up the speed of the interface, they also ramp up the speed of the hard drives, to fill the new potential. Meanwhile, the speed of RAM has consistently been 1000 times faster, all along. |
|
| |
phoenix, I personally dislike Flash RAM devices, because they have relatively small limits on the number of times they can be re-written to hold data -- and you never know when it is going to happen to run out. |
|
| |
Yes, memory is alot faster than hard drives, and AFAIK, the executable portion of a running program is loaded into ram when it's run. The only place that your proposal would make any difference is in the initial load time of a program. I'm perfectly willing to live with the sub-2-second wait times when loading up Excel or Word or IE6.0 or whatever else I happen to be running, in exchange for the extreme cheapness of modern hard drives. RAM, while 1000 times faster, is also roughly 200 times more expensive. (of course, if you're trying to multitask well beyond your system ram capabilities, or are dealing with a few absurdly large files, you're going to run into problems with disk access speed as the OS starts using the swap file as ram, but I'm not talking about times like that.) |
|
| |
I can see this being a benefit if you're making hundreds of database requests to multitudes of small files, or if you're considering the use of a device like this as part of a highly utilized network server. In this type of use, then yes, there will be a marked increase in performance and access times, and an associated improvement in performance. For most common desktop applications, the actual time spent accessing a hard disk is a very small contributor to your overall speed. (example: I have an old 1 gig hard drive in my computer (it's clear-top modded, just for looks), as well as a 160 gig. the 1 gig is an old ata33, the 160 is an ata133. Other than load times, I notice no difference in performance with one drive over the other.) |
|
| |
The whole computer scheme needs a serious re-think. Why should we be bothered with saving at all? Why can't that step simply be progressively automatic and behind the scenes? Drive letters, operating system files, program files, the whole nine yards should be simply out of sight and out of mind. |
|
| |
Liquid state disk drives, now I'd pay for one of those. Liquid disk, mmm. |
|
| |
freefall, let me clearly state that I'm not suggesting that SSDs replace ordinary hard disks. I'm saying that they should do the never-ending work work of small-file loads/saves, during the running of a program. Also, while the extra speed of an SSD over a normal hard disk is not going to be spectacularly obvious, when many programs are run, please remember that the extra speed is not the primary reason for promoting SSDs here. The primary reason is the greatly increased lifespan of hard disks, because of reduced wear-and-tear. I said that I consider the savings in aggravation over lost hard drives to be worth more than the speed gain. Wouldn't YOU like to see your hard disk last an extra 5-10 years or so? "Ah, by then I would have replaced it with something more capacious!", you say? But can you guarantee that what you have now will indeed survive until you get that more capacious drive? (Hard disk makers have recently reduced there warranty periods from three years to one!) I am confident that working with an SSD as described here will give you that security (unless your drive is already almost worn out, of course). |
|
| |
I hope I'm not out of line here (this is my first post), but it seems to me that if you don't use your HD for 10 minutes or so, it goes to sleep. So, if you run Excel, and the data is moved to the RAM disk, and then your drive goes to sleep and powers down, then has to wake up when your done, doesn't that make MORE wear and tear on your drive?? Also, I've thought about this before I saw it here; what if instead of working with the HD it could be used to suppliment, like a extra HD for programs you use most often to load them faster. To get around the losing everything when the comp is turned off, why not just put a little rechargable batery in there that keeps the data "alive" while it's turned off. I know it would not last indefinitely, but it should last for a few weeks, RAM chips don't draw much power to refresh I don't think. Or just use static RAM and forget the battery.
Sorry, just ranting... |
|
| |
The problem is that this is in no way a new idea. It's done now, just not on the scale [Vernon] wants. Hell, I worked on refrigerator-sized computers will 32k of magnetic core memory 20 years ago. Go to a U.S. Navy scrap yard, maybe you can find one. |
|
| |
Katt, most computers will let you set the amount of time before they turn off their hard disk; a reasonable default is 10 minutes for some applications, and longer for others -- and perhaps only 2 minutes when the system operation paradigm includes an SSD. Regarding wear and tear during spin-up and/or spin-down, no, I don't think so. They are spun as inductive electric motors are spun, so the only wear points are the bearings (no brushes), and obviously an axle that is not turning (or is turning at some speed less than max) is wearing out the bearings less than when the axle is turning at max RPM. Perhaps of more concern are the read/write heads, which these days are so close to the surface of the disk that there is not even enough room for light to pass between them. When the drive is not turning, the heads may actually touch the disk. BUT, the disk is coated with Teflon (or better), and the heads are designed so that some minimal disk speed is associated with airflow such that the heads aerodynamically lift from the surface. By design, then, it is generally not harmful for the disk to stop and restart spinning multiple times an hour. Only the delay of getting it spun back up to its normal operating speed will be annoying. Regarding your battery suggestion, well, there are some non-volatile RAM technologies coming down the pipeline that will make such batteries unnecessary in a couple of years (and powering only part of a computer's motherboard, to keep RAM refreshed today, is problematic). You seem to have a misconception about the nature of static RAM. First, it is more expensive than ordinary dynamic RAM because each SRAM memory cell consists of more parts (usually six transistors apiece, while a DRAM cell consist of just one transistor and one capacitor). An SRAM cell only "remembers" its data constantly because of the relative on/off relationship of the transistor-group that makes it up -- which has to be maintained constantly by electricity constantly flowing through it. SRAM and DRAM both forget everything when the power is off. (Just to be complete, a DRAM cell holds its data in its capacitor, which gradually leaks and so must be dynamically refreshed at intervals.) Anyway, even though the boot-up time of computers is a bit annoying, most people are so used to it that the little additional boot-time implied in this Idea (Operating System copies from hard disk to SSD before loading to main system memory) will hardly be remarked. |
|
| |
Rods Tiger, I have never really liked hard disks (they failed rather more often in the 1980s), and personally prefer to keep my data on removable magneto-optical disks (technology based on a Natural phenomenon that lets geologists discover where the Earth's magnetic poles were, a hundred million years ago -- now THAT's data retention!). But in the 90s hard disks became (to me) acceptably reliable, and so capacious that they were the only things able to hold the bloated programs coming out of Microsoft and other places. Now, as you say, they seem to becoming less reliable, and alternatives are few (just the other day, some outfit in Europe tested old CD-Rs for data retention, and found many failures after only two years!) So, this Idea offers a way to drastically reduce the usage of hard disks, thereby letting them last longer, once again. Finally, regarding your implication of WIBNI, the only thing I'm looking for is a cost-reduction. SSDs exist, but are pricey partly because they are being directed at the enterprise/server market (folks who have $ to waste, that is). By offering this Idea, so that the average person can get the benefit of a longer-lasting hard disk by adding an SSD, I hope to send the SSD manufacturers a message that a much larger market is possible for them -- if they sell for less. |
|
| |
phoenix, I did say quite up-front that SSDs do currently exist, and fairly widely, too. And I do know that RAM disks (based on system memory) were often used to speed things up (DOS acquired RAMDRIVE.SYS even before the 1MB barrier was broken; its default is to be a 64KiloByte drive). But while others may have noted that hard-disk-lifespan can be extended if SSDs are used, I don't know that anyone before me specifically stated that SSDs should be installed just to GET that extended lifespan. Previously, the focus has always been on that 1000-times-faster data access. |
|
| |
If you're that worried about data loss, just use some RAID structure with your hard drive(s). Then if something fails, it can recover any lost data. This way, the only advantage your system has is speed. |
|
| |
jivetalkinrobot, sorry, but one recent failure I encountered had to do with the computer's power supply -- it died and took FOUR IDE drives with it (including CD_ROM)! Perhaps it would have taken an IDE SSD too, but do remember that in the described paradigm, one tries to keep the data off the hard disk, and on removable media. RAID isn't necessarily as guaranteed as you think. |
|
| |
Call me childish, say I'm not contributing positively, but can we get away from the word "paradigm"? I've only ever read it in Dilbert cartoons... <looks embarrassed> Sorry... |
|
| |
[david_scothern], here is a definition:
Paradigm
From Wiktionary, the free dictionary.
Noun
1. A way of thinking about, looking at or approach to working in some context |
|
| |
The typical manner in which data flows through an ordinary computer can indeed be called a "paradigm". And I think it can be altered/improved ("shifted"), as described above. |
|
| |
I'll try to compress this idea down to a single sentence: |
|
| |
Make solid state drive devices cheaper by creating demand and marketing in modular form. |
|
| |
You also state that "most programs are less than half a gig". I submit that most well-written software accounts for pre-caching of data stored on the hard disk, such that most of the disk transactions take place transparently to the user, without slowing things down appreciably. Most of the delay that's seen is due to the initial load process. Your action of loading the entire program to the SSD before beginning operation just slows down initialization even more. |
|
| |
Sorry, I can't vote for an idea that essentially boils down to "take an existing technology and make it cheaper so I can afford it". |
|
| |
[Freefall], you have some of your simplification backwards, besides the fact that you are utterly ignoring the wear-and-tear issue. I want the manufacturers to take their existing technology and remove the fat. That will automatically make the devices cheaper, and thereby almost-as-automatically increase demand. And yes, of course I would buy some of that less-bloated and less-expensive hardware! |
|
| |
Regarding the built-in cache of most ordinary hard drives, yes, this does serve to save some wear-and-tear. However, the Operating System normally insists that data-saves not be held in cache very long, in recognition of the need to ensure the data actually ends up on the hard disk. But in this paradigm, with the system protected by a UPS, you can do ALL data-saves to the Solid State Disk, except for the very last one at system-shut-down time. |
|
| |
[aeolis14umbra], thanks, but what is a CF card? |
|
| |
Well, solid state storage has been part of my vision for a new type of computer for a long time, but is still way too expensive. Here's the gist of it. (someone already touched on 1 important feature.) |
|
| |
**Small but very fast and cheap HDD for storing personal data and configuration files
**A bank of small slots to accept very fast non-volatile memory, about the size of SD camera memory cards
**Software is purchased on memory cards.
**The OS (Perhaps Windows 2020) would simply blink on. In fact why even turn it off? It's solid state, and runs very cool so little degradation would occur.
**Plug in your new software, and leave it plugged in if you wish.
**First time a new software is plugged in, a config menu comes up. The config is saved to the HDD and /or in a separate memory card. |
|
| |
OK, this annotation is already too long, so fill in the blanks. |
|
| |
Thanks, [Tabs]. I do know enough about Flash RAM to know that it isn't good enough for this Idea. Flash memory cells can only be used a limited number of times before they die. Sure, it is a fairly high limit, suitable for casual data storage, but not suitable for high-data-turnover that will be an inherent part of a working RAM Drive. |
|
| |
Alas, this is now fairly baked. SanDisk has a 32GB solid-state HDD that isn't /quite/ out on the market yet, but close. Some more info about RAM:
> it is much more expensive to make than regular microchips, due to greater circuit density, other manufacturing steps, and less fault tolerance.
> it can be made to be non-volatile, e.g. flash drives these days have no power source when they are unplugged; however, this does not last forever. Long-term unpowered data storage often results in data loss/corruption, so using flash drives to distribute software probably isn't a good idea.
> a company called L Computers baked this idea a while ago, and called the invention PuRAM (as in Pure-RAM). (linky) |
|
| |
[Agamemnon], it clearly says in the description of the PuRAM that it uses flash memory. My main beef with flash memory cells is that they have a limited number of write cycles, before they stop working. Ordinary RAM has no such limit. |
|
| |