640gb PCIe Solid-State Drive Demonstrated

Lisandro writes "TG Daily reports that the company Fusion io has presented a massively fast, massively large solid-state flash hard drive on a PCIe card at the Demofall 07 conference in San Diego. Fusion is promising sustained data rates of 800Mb/sec for reading and 600Mb/sec for writing. The company plans to start releasing the cards at 80 GB and will scale to 320 and 640 GB. '[Fusion io's CTO David Flynn] set the benchmark for the worst case scenario by using small 4K blocks and then streaming eight simultaneous 1 GB reads and writes. In that test, the ioDrive clocked in at 100,000 operations per second. "That would have just thrashed a regular hard drive," said Flynn. The company plans on releasing the first cards in December 2007 and will follow up with higher capacity versions later.'"

Comments

Oblig.

[Anonymous Coward]

640gb ought to be enough for anybody.

Re:Oblig.

[GodfatherofSoul]

Maybe for your pr0n collection ;)

Re:Oblig.

[ady1]

Its not the size of the harddrive which is amazing. Its the read/write speed.
Even if you get a 32GB model, you can install windows on it and use the regular SATA2 HDD for movies/music storage. Think of the booting time.

Re:Oblig.

[Lisandro]

Uh oh. These solid state drives are not intended to be used in desktops, where things like swap files are very common - this device is flash based, and flash still has a lot of issues regarding limited write cycles (i recall the best current flash chips have a maximum of 1,000,000 write cycles per cell). You wouldn't be able to boot from them either, since this hardware is unknown to current BIOSes.

The way i see it is a very, very neat way of replacing hard disk arrays for enterprise use; the device itself it's surely very expensive, but way smaller, cooler, quieter, with less power consumption and near zero manteinance costs. A hd array with similar write/read performance would also be extremely expensive, if even feasible.

Re:Oblig.

[Anonymous Coward]

Funny...

*looks at his K6-III booting from a PCI-SATA card.*

Just FYI, the mobo is from 1997 or 1998

There's some standards for bootable PCI devices, and PCIe I suspect...

Re:Oblig.

[Lisandro]

Funny...

*looks at his K6-III booting from a PCI-SATA card.*

That'd be a PCI SATA controller card? You can boot from them alright, but the article mentions that drivers are available for several platforms - if drivers are involved, it's most likely not using a regular ATA/SATA interfase.

Re:Oblig.

[CastrTroy]

Couldn't you mostly boot from it by putting your boot information on a standard drive, just enough to load up the controller for this drive, and then loading everything else off the solid state? I mean, it wouldn't be quite as fast, but you'd still save quite a bit of time.

Re:Oblig.

[zippthorne]

Which brings to mind an interesting point..

Why even have swap files? Shouldn't caching decisions be done a bit more intelligently at the application level? I have 10 times more RAM in my current PC than all of the memory (including the HDD) on my first PC. At some point, can't we drop swap?

Re:Oblig.

[Anonymous Coward]

Uh oh. These solid state drives are not intended to be used in desktops, where things like swap files are very common - this device is flash based, and flash still has a lot of issues regarding limited write cycles (i recall the best current flash chips have a maximum of 1,000,000 write cycles per cell). You wouldn't be able to boot from them either, since this hardware is unknown to current BIOSes.

Your childish "uh oh" introduction, your completely un-cited "a lot of issues" comment, and your vague "I recall" interruption reveal the fact that you're spouting off some crap on a subject you have no direct, real-world experience with. Find another subject with which to stroke your ego kid, because you're looking like a pompous dumbass on this one.

Re:Oblig.

[Anonymous Coward]

so...

since drivers are available for several platforms for my SATA card (that I'm booting from)... It's not using a regular ATA/SATA setup and can't be booted?

Re:Oblig.

[jamie]

Once you get above $500 desktop computers, it doesn't much matter. A properly tuned system will only use swap, if at all, to drop a few MB from RAM to disk because it's just never accessed. A server that swaps during use is just not set up properly.

Re:Oblig.

[walt-sjc]

True, you have a lot more RAM now than you used to... BUT, you are also not running DOS anymore. Hell, MS Office alone take 256M just to load Word (I KID...) But seriously, modern apps are huge. Because memory is cheap, apps are not efficient. We also use our systems differently, working with massive files (images, movies, music) and tend to run lots of apps at the same time, including virtual machines. Yes, you can still run windows in 512M, but it seems that even 2G isn't enough at times.

So we still need swap.

Re:Oblig.

[ady1]

What I said apply to linux/OSX/BSD as well. :P

Re:Oblig.

[Beardo the Bearded]

Thank you. I came in here to say that. Flash can only be written to a limited number of times. Reading is basically unlimited, but writing ... ugh... wears it out a little. (Not really true, but close enough for this crowd.)

It's not a magic number, as some people assume for some reason. What they mean is this:

Statistically speaking, some of your bits are going to be incorrect in some places after a mean number of writes approximately equal to 1,000,000. At that point, you will have actual physical locations that aren't going to read back what you wrote. You could probably compensate for that like HDD do for bad sectors, but it's almost impossible to tell if your memory is bad.

Maybe the company has found a way to compensate.

Re:Oblig.

[HTH NE1]

640gb ought to be enough for anybody. What's a gram-bit?

Re:Oblig.

[Anonymous Coward]

> flash still has a lot of issues regarding limited write cycles

One thing I have wondered is if there is any RAM based solutions like this.
Where a board or a box would allow a computer to treat a bunch of RAM as a drive.
Yes you would loose your data if the system goes down, but I can work around this.
"limited write cycles" is something I cannot work around.

Re:Oblig.

[norton_I]

You can tell if flash is bad, if worst comes to worst, by reading after writing. Or reading after erasing, and looking for stuck 1's.

'worn out' flash doesn't spontaneously change state. Bits just get stuck and don't erase correctly.

I don't know how flash drives actually handle this, but it isn't magic or impossible to fix.

Also, the lifetime of modern flash is long enough that it is hardly an issue any more, even for normal desktop use. Maybe you don't want to use it for swap *IF* you swap a lot, but given the cost is in the same ballpark as RAM, you could just buy more RAM.

Re:Oblig.

[Anonymous Coward]

Because of mechanical failures, electro-mechanical devices like hard drives have an upper bound on the number of writes as well. Flash devices are now designed with a lot of "extra" space, and wear leveling keeps the same sector from being repeatedly written on.

So, we can come up with an estimate for the life of a Flash based drive. At a million writes per sector, you would need on the order of 80 million GB written to the drive (uniformly) before it will fail (uniformly). This is a rather unrealistic failure mode. Most likely, someone with a large drive will have a collection of files that remains relatively static, in addition to variable data. The static files are written once (or near there, relatively speaking). If they take up x space (in GB), and T is the total (GB), you will have to write on the order of (T - x) million GB to ruin the free space.

Flash based drives will already outlast electro-mechanical drives. They're just more expensive, for now.

Re:Oblig.

[poopdeville]

Obviously, a unit of mass times information.

Re:Oblig.

[sYkSh0n3]

Yes yes! for god's sake, think of the boot time!

Re:Oblig.

[volsung]

Yes, and they are quite expensive. 32 GB will run you about $14k: HyperDrive 4

Re:Oblig.

[Lisandro]

our childish "uh oh" introduction, your completely un-cited "a lot of issues" comment, and your vague "I recall" interruption reveal the fact that you're spouting off some crap on a subject you have no direct, real-world experience with. Find another subject with which to stroke your ego kid, because you're looking like a pompous dumbass on this one.

The specs for a 256Mb NAND flash memory chip by Samsung (which is by far the biggest NAND flash manufacturer today) quotes 100k millon write/erase cycles, and this is for an IC commonly used in USB pendrives. The figure usually tends to get worse with increased memory sizes since the memory "element" (float gate) becomes smaller. For example, Modern 16Mb chips, which are the ones i have experience with, usualy quote 1 million W/E cycles endurance.

But, it felt good stroking my ego a bit more :) Thanks!

Re:Oblig.

[Lisandro]

One thing I have wondered is if there is any RAM based solutions like this.
Where a board or a box would allow a computer to treat a bunch of RAM as a drive.

You could always stuff more memory in your PC and use a ramdisk driver. Anyway, i know such devices exists (solid-state drives using DRAM), but i never got a chance to try one...

Re:Oblig.

[Anonymous Coward]

So far this thread you done naught but prove your ignorance in this subject.

Flash drives spread out reads/writes over the drive, so that they don't die so quickly.

So, while the write to Nth block on the drive might at one time write to the the cells at the physical location of (23nm, 42nm, 35nm) on the drive, the next write might end up in the cells at (62nm, 2nm, 15nm)

Obviously the drive will use an internal routing offset, rather than a physical location, but you get the idea.

With this, you are bound more by total IO of the disk than IO to a given sector.

so a 32GB disk might end up handling 32,000,000GB worth of IO before failure.

Re:Oblig.

[Sczi]

Nah, if you can afford $30/gig for the "hard drive" you can afford to dump 8gigs or more into it and ditch swap.

Re:Oblig.

[Anonymous Coward]

You can also tell if FLASH is verging toward failure on any particular block by timing the erase cycle. As the part ages out, the erase cycle takes longer. You can also use that information to enhance wear leveling.

Re:Oblig.

[sYkSh0n3]

Why couldn't you boot from this? i seem to recall watching a video of windows xp booting off a PCI card that held 4 1gb sticks of ram and a battery for data retention when the unit was off. Booted in just under 4 seconds i believe.

Re:Oblig.

[Biljrat]

They will just use wear levelling and relocate data out of flash blocks that are underperforming or that have ECC corrections. Just because they report 640GB my guess is that they will actually have 15% to 25% additional flash to handle early failures without having to start reducing the reported capacity.

http://nwbagpipes.com/

Re:Oblig.

[Anonymous Coward]

Disregard that, I suck cocks.

Re:Oblig.

[networkBoy]

No you would boot directly from it. GP doesn't understand BIOS very well. This could hook into the boot process much the same as any RAID controller does.
-nB

Re:Oblig.

[AK Marc]

Not in Windows. It has a swap file by default. If you disable swap, it still swaps. If you have swap enabled and have something running in the background and come back hours later, it's so slow it's painful, since everything gets swapped to disk except the one thing running. "Tuning" may be possible, but well outside the realm of the average user, and even completely disabling swap in all settings, you still have swapping happening.

Re:Oblig.

[Smauler]

Why not then use the best of both worlds? Have the OS on the flash, and have everything else, including the swap file on a separate hard drive. After all, 640Gb should be enough for any OS.

Re:Oblig.

[walt-sjc]

Um, many systems can't handle 8G of ram. Windows (32 bit) can't handle more than around 3G (the exact number depends on what devices you have installed.)

64 bit windows and Linux can address lots of memory however, assuming your motherboard can (which many can't.)

But besides the limitations above, it makes sense to "swap out" the parts of memory you are not using. No sense in putting 8G of real ram in when 4G will be unused code that would have been in swap (not just the money cost, but also electrical cost.)

Re:Oblig.

[nuzak]

The wear leveling that pretty much all flash chips do now puts them on par with mechanical HDD's in terms of lifetime. Furthermore, dead cells are blocked off, and the storage space simply shrinks a little bit after a few years -- it doesn't fail catastrophically.

The limited write cycles of flash drives is pretty much a non-issue. You probably shouldn't put a swap partition on one though.

Re:Oblig.

[Sczi]

Um, many systems can't handle 8G of ram. Windows (32 bit) can't handle more than around 3G (the exact number depends on what devices you have installed.)

Again, if you can afford $30/gig, you can afford.... etc

64 bit windows and Linux can address lots of memory however, assuming your motherboard can (which many can't.)

There's a reason it's called the bleeding edge.. I've been running 64bit for years, btw.. I scoff at the half bitted plebes ;] In fact, I think half-bit will be my diminutive du jour, as in "Why did apple go back to half bit?"

But besides the limitations above, it makes sense to "swap out" the parts of memory you are not using. No sense in putting 8G of real ram in when 4G will be unused code that would have been in swap (not just the money cost, but also electrical cost.)

I know why it's there and what it does, I'm just saying that if OS's want double the ram every 3-4 years, but ram doubles in capacity/dollar every year, then eventually we should be able to get rid of swap. I mean, lord knows I do use it, but it has always kind of irked me for some reason. And of course if your goal is to ditch swap, then there is sense in it. Unless you need your swap to persist between boots, there's no sense in committing it to disk anyway.

At least with Linux and OS's that use a partition you don't have to worry about the damn thing getting fragmented. Dunno how much of an issue that is with Windows these days, but back in the day if your swap size was adjustable (and it was by default) it could get heavily fragmented.

Oh yeah, how does a solid state drive handle fragmentation? I have heard that they don't fragment, but not from reliable sources, and I just don't see how that is possible unless there is some built in mechanism to close gaps on the fly or something. If you fill the disk full then selectively delete files around the disk and then write one big file, it will fragment. Maybe fragmentation is a moot point (is it?) but that's not the same thing as not having physical fragmentation. It seems like an SSD would still need pointers of some kind.

Re:Oblig.

[Fulcrum of Evil]

Actually, I'm considering a server upgrade - AM2 opteron at 2.4G, 500G disk, 4G ram for about $1000. 8G would be another $300. Who cares about office?

Re:Oblig.

[Eccles]

Oh yeah, how does a solid state drive handle fragmentation? I have heard that they don't fragment, but not from reliable sources, and I just don't see how that is possible unless there is some built in mechanism to close gaps on the fly or something.

Hard disks have a fragmentation issue because sequential accesses are much faster than random ones with a spinning disk. Each time the next sector to get isn't right after the previous one, the head must seek to the start of the next track. Solid state "disks" have true random access, where accessing blocks in random order costs no more than sequential accesses. So while solid state will fragment, it doesn't matter for performance or reliability.

Re:Oblig.

[zippthorne]

But does the 4G of unused stuff even need to be loaded in the first place? From what I've gathered, A lot of the stuff that goes into swap is present almost verbatim in the original files on the disk. I really question the wisdom of dumping an entire seldom-used daemon into a swapfile when those the vast majority of those bits are already present on the disk. I mean, does it really matter whether you load a file from a disk or a page from a disk?

The problem is conceptual, and practical: Application programmers have been able to offload quite a bit of memory management to the OS to handle in a "dumb" but fairly reliable way. In the age of cheap, long-lasting magnetic disk, this makes a certain amount of sense. But if every write is counted, it makes sense to be a little more careful about when you really need to put there.

Re:Oblig.

[dgatwood]

Um, many systems can't handle 8G of ram.

Not many systems with a spare PCI Express slot, though.

Re:Oblig.

[sasserstyl]

Everyone knows that solid state is the future - there are so many obvious benefits to having no mechanical parts (power/reliability/cost/longevity/size etc).

But there's less to be excited about here than everyone seems to think.

At 30 USD a Gigabyte thats around 18000 USD per card. Even the 32Gb version would cost 1k USD.

And my laptop resumes from sleep in little more than a second already.

And the number of write cycles on flash memory is severely limited AFAIK - does this company provide any information on this for their product?

Re:Oblig.

[Crazy Taco]

Hell, MS Office alone take 256M just to load Word (I KID...) But seriously, modern apps are huge. Lol... you joke about Office, but what about Emacs? It's a text editor with a 141 MB footprint (I don't kid). Modern apps may be huge, but I guess emacs proves that old, cluttered apps are the hugest.

Re:Oblig.

[Cantinflas]

I don't know where you got your info, but 'modern' flash has horrible life-expectancy in terms or write-cycles.

Explanation:

Most flash vendors have moved to MLC (Multi-Level Cell) flash. It's cheaper and denser, but the bit-error rate goes up because you have more bits per cell. The typical life expectancy for MLC is somewhere in the range of 10,000 writes using single-bit error correction. This is compared to 'older' SLC flash which has a write endurance of 100,000 to 300,000 writes.

Now, most vendors making media out of flash take varying degrees of a combination of two approaches (in addition to standard wear-leveling approaches). The first approach is to assume that the majority of the users will only ever store audio or video data so the occasional uncorrectable error won't have much impact as long as it doesn't corrupt the filesystem. The second approach is to use more advanced error correcting algorithms to compensate for the higher bit-error rate.

Using more advanced algorithms, it's possible to get more than 300,000 writes out of a MLC flash-block before the errors become uncorrectable.

P.S. I may be wrong, but I believe flash can have some really odd error conditions. For example, it's possible to disturb a bit in a block just by reading it. I believe it's also possible to disturb a bit in a different block on the same matrix when writing. That's why some form of error correction is always required with NAND flash.

Re:Oblig.

[darkpixel2k]

So while solid state will fragment, it doesn't matter for performance or reliability.

Yes, but it *needs* to be defragged because I *hate* seeing red in my disk analysis...

Re:Oblig.

[ppc_digger]

I don't know about Windows, but both Linux and BSD don't keep executable code in swap space, as long as the original executable doesn't get unlinked.

Re:Oblig.

[Soul-Burn666]

My computer runs reasonably fine (including games) with 512mb ram and 1gb of swap space.
If I upgrade to 2gb ram as many people have now, I'll have more ram than my entire ram+swap and could leisurely disable the swap space.
Windows many times swaps out data wrongly, causing a lot of disk thrashing and removing the page file can eliminate this annoyance.
Even large games don't really need 2gb of ram I'm not afraid of running out of memory. And even if so, you're probably better off buying another 2gb, which is dirt cheap nowadays, and eliminate this problem completely.

Re:Oblig.

[timeOday]

I've used Linux systems for years with no swap enabled, such as the laptop I'm currently on. free total used free shared buffers cached Mem: 3631056 1649840 1981216 0 96240 1264372 -/+ buffers/cache: 289228 3341828 Swap: 0 0 0 Mostly all swap does anymore is make the system run extremely slow when a berzerk application tries to allocate gigabytes of space. Granted, swap probably also spares a few megabytes of RAM for stuff that's loaded but never accessed - but I don't think that amount is substantial.

Re:Oblig.

[F4_W_weasel]

Asta la (windows) vista....

Re:Oblig.

[Yvanhoe]

CRC checks and parity bits should help you see if something is wrong.

Re:Oblig.

[Hatta]

Uh, why the hell would I install windows on it?

Re:Oblig.

[drsmithy]

Even if you get a 32GB model, you can install windows on it and use the regular SATA2 HDD for movies/music storage. Think of the booting time.

I boot my PC maybe once a month. Usually while I'm out of the room doing something else. Further, most of the delay involved in booting has nothing to do with the devices' speed, but the time taken to initialise various bits of hardware and software.

I have never understood the fascination with using drives like this for installing the OS onto. Why ? The OS files are mostly accessed very rarely, and those that are frequently - unless you're using a 15-year od OS with incredibly bad memory management - will always be cached in RAM anyway.

Re:Oblig.

[norton_I]

I guess I was assuming that for an application like this, reliability and speed would trump density and cost, and they would just use SLC mode. I didn't think anyone considered MLC flash suitable for anything other than MP3 players and cell phones where size the overwhelming factor. If they really have the wear leveling and error correction up to where they get 300k writes, then maybe that is no longer the case.

I suspect the odd error modes are also much worse for MLC, but I don't really know that much about them. From what I have heard, SLC is pretty tame.

Re:Oblig.

[rbarreira]

For example, it's possible to disturb a bit in a block just by reading it

Much like quantum computing then?

Re:Oblig.

[McFadden]

Except anything by Adobe would still take 30 seconds.

Re:Oblig.

[aldo.gs]

You people always thinking about your booting times! Won't someone THINK OF THE CHILDREN!!?

Re:Oblig.

[jamie]

If you're right, I guess Windows can't be properly tuned. I wouldn't know :)

Re:Oblig.

[totally bogus dude]

even completely disabling swap in all settings, you still have swapping happening

I think this is because Windows counts anything which is backed by a disc copy as being "swapped", which includes executables and DLLs. This wouldn't affect the flash drive though, as Windows isn't actually writing them to a swap area. I think it also lists it as "page file usage" even if the programs are actually resident in physical memory; it's more an indicator of what it could dump from memory if it needed to, because it can just re-load it from disc when it's needed again (which makes it kind of hard to work out if a machine actually needs more memory or not).

Re:Oblig.

[totally bogus dude]

Yes, but it *needs* to be defragged because I *hate* seeing red in my disk analysis...

Then don't do a disc analysis! ;)

Alternatively, they could update the disc analysis software to know that fragmentation isn't a problem in flash drives, and not show it as red. Then everyone would be happy.

The problem with defragmenting flash drives is that it involves a lot of writing to the drive to shuffle bits around, which reduces the life (each cell is only good for a limited number of write cycles). So, you don't really want to make frivolous writes to the drive, because you're just shortening its life for no actual gain.

Re:Oblig.

[Agripa]

'worn out' flash doesn't spontaneously change state. Bits just get stuck and don't erase correctly.

This is not completely true.

Read operations can cause bit failure within pages that are in the same block. It is common enough that reserve ECC capacity should be checked when verifying writes.

Re:Oblig.

[Verte]

Because other operating systems:

1. don't need to be rebooted as frequently, and
2. don't take as long to boot anyway?

:P

Re:Oblig.

[TheLink]

"Solid state "disks" have true random access, where accessing blocks in random order costs no more than sequential accesses"

AFAIK for flash, sequential access is still faster than random access, even for NOR flash.

It's quite amazing how slow some flash is (esp NAND flash). Some are 1ms for random access, and others are even 7ms.

For comparison a 15krpm drive has a random access time of about 5-6ms. So if you have RAID10, you might get similar or faster speeds, than a flash drive 10x the total price of a RAID10 system with similar capacity.

Re:Oblig.

[cecom]

Sadly that is not true. Most people naively assume that wear leveling magically fixes everything, while in reality the situation is different.

First of all, due to the nature of NAND flash, you can only erase whole 128KB blocks and then write in 2KB chunks. You can simply not rewrite a 2KB page, unless you erase the entire 128KB block. So, it is not simply a matter of wear leveling, but of fragmentation and how to avoid it. This is much more serious. A naive algorithm means that you will be erasing 64 times more often, which means that the life drops from 10,000 rewrites/block to 160 rewrites/block. I am not aware of any ideal solution, unless battery-backed RAM is used for buffering.

Secondly, most (all?) flash controllers, do not implement "global" wear leveling - that is across the entire storage area. They separate the storage in zones and do it only inside each zone - this decreases the effectiveness.

This is not to say that there are no solutions. They may be expensive though. In any case, thinking that "wear leveling" is magic bullet is simply wrong.

Re:Oblig.

[mattyrobinson69]

Solid state drives dont have seek time, so fragmentation isn't an issue. With normal hard drives, the read head has to spin around to the part of the platter with the data you want on it (hence the latency if its scattered all over the disk).

Re:Oblig.

[dwarfking]

Odd, I have Emacs 22.1.1 on Windows XP running and task manager only shows it using 21,796kb memory. Firefox, on the other hand , is currently using 81,924k. Not certain where the 141 MB number comes from, unless you load ever single possible mode, but I've never seen it that high.

Re:Oblig.

[nuzak]

The wear leveling that pretty much all flash chips do now puts them on par with mechanical HDD's in terms of lifetime. Furthermore, dead cells are blocked off, and the storage space simply shrinks a little bit after a few years -- it doesn't fail catastrophically.

The limited write cycles of flash drives is pretty much a non-issue. You probably shouldn't put a swap partition on one though.

Hey mods, I can just keep reposting even more. Fuck all of you some more.

God, it wasn't even an interesting post. But fuck you mods anyway.

Re:Oblig.

[nuzak]

You're right about the lack of wear leveling across the entire device in current flash devices, but I wouldn't be surprised to see it in the large drives, or at least some driver support for querying the current wear levels. When you pay as much as a SAN costs for a single drive, you probably expect more than just a big chunk of storage. The current solution is either to use either a smart filesystem (like JFFS) made for flash, or strangely enough, a really dumb one like FAT that has such big blocks that it's forced to write out large chunks anyway.

It seems to me though that the best proof is anecdotal: I see people carrying around thumb drives that they've had for several years, that they write to all the time, and they're still working fine. One usually loses the device or destroys the hardware before it wears out. There's no magic bullet in wear leveling, but it certainly seems to be good enough to outlast spinning platters o' rust.

Re:Oblig.

[Sczi]

Taken to its logical extreme, though, there still would have to be something akin to a file allocation table to hold start/stop pointers. If every file on the disk were composed of 10000 fragments, wouldn't the fat would have to be huge? And wouldn't that hurt performance? But then the maximum number of fragments would be something like file size divided by cluster size, so reading the fat will still be a small fraction of the total read. Beyond that, I suppose standard defrag would still be an option, right? I guess my question is more academic/theoretical than practical. It seems that if there is a performance hit, it is probably minuscule. The reason I asked in the first place is because I had read that they simple *don't* fragment, which seemed impossible. Just wondering how similarly they behave to a regular disk..

Re:Oblig.

[mattyrobinson69]

A defrag would cause lots of writes to the disk, which you dont want on a flash drive (they wear out). They do fragment, its the job of the operating system to prevent it. I'm not sure about NTFS, but I remember FAT32 defragmenting a lot. EXT2/3 hardly fragments at all, I bet NTFS is the same, as well as reiserfs, xfs, etc.

Re:Oblig.

[Eccles]

AFAIK for flash, sequential access is still faster than random access, even for NOR flash.

Flash has to be read in blocks, so it's true that I'm oversimplifying quite a bit. I'm also not clear on whether there has to be the "seek" time between all blocks, or whether it has blocks that can be sequentially read together to avoid the "seek".

However, the access time numbers I see quoted online are dramatically different than yours. For example, AMD says their access times are 45 ns to 150 ns. Could your numbers be the total for a USB flash drive, including interface overhead? A flash "hard drive" could probably get much closer to the chip speeds, and certainly you would use faster chips, as they must be doing here.

Re:Oblig.

[TheLink]

The access times are from semi random searches for benchmarks of flash drives out there.

My guess is that's because of the flash drive controllers and the block read limitations.

But if you get a better controller or more controllers and do lots of nifty tricks you can get much better access times. Which is what might have been done in the demoed product mentioned in the article. I'd like to know what tricks they did though, maybe it only works that well in the demo scenario.

Perhaps you're right and the interface overhead is more significant (no idea though). Whatever it is, a real drive will also have similar overheads whether it's USB or ATA or SAS or SCSI. Whatever AMD says about their chips isn't so relevant.

Re:Oblig.

[Eccles]

The Sandisk 32 GB flash notebook drive was measured at 0.6 ms access time, and another article tagged one jump drive at 0.8 ms access time, both much better than a very fast hard disk. (I did see other, slower devices, but one just wouldn't use the slower tech.) And clearly we can boost flash's transfer speed by using multiple chips in parallel.

BitMicro has their own version using standard SATA. Not nearly as fast in transfer speed, but they claim 0.03 to 0.1 ms access time: http://www.linuxdevices.com/news/NS4331778531.html
So these other guys aren't alone. One wonders how SATA vs. PCIe bus affects the performance.

Uhh, Price?

[mosel-saar-ruwer]

Who, what, when, where, why?

Price would seem to be a pretty important detail...

Re:Uhh, Price?

[thegnu]

I think it's safe to assume you won't be buying one soon.

Re:Uhh, Price?

[Major Blud]

If it can achieve the performance numbers that they claim (which it probably can based on NAND technology), then price price wont' keep people from buying this. This has been what enterprise DBA's have been waiting to have for awhile; the solution to high disk queues and wait times. I can't wait to get my hands on a test unit; SQL should smoke on this sucker.

Re:Uhh, Price?

[morgan_greywolf]

FTFA:

So how much will these cards cost? Flynn told us that the company is aiming to beat $30 dollars a GB, something that should seem very cheap to large corporations, adding "You can drop ship or Fedex this card and be up and running in a few minutes... you can't do that with a storage area network." So, let's say they get to $29 a GB, a reasonable price for NAND flash-based memory devices. 640*30==$19,200. Sorry, but that doesn't seem to beat an inexpensive SAN in price. I recently priced out a 12TB iSCSI SAN for a little bit more than that, and even 1-2 TB fibre SAN from IBM should be around the same price.

Re:Uhh, Price?

[PineHall]

From the article they are "aiming to beat $30 a GB". Not cheap! Figuring $30/GB, a 640GB drive would cost $19,200. So it is way out of my price range.

Re:Uhh, Price?

[KingSkippus]

So how much will these cards cost? Flynn told us that the company is aiming to beat $30 dollars a GB

At $30/GB, that would make an 80GB drive around $2,400. Hopefully the price will eventually come down, as "He even hinted that the company is looking into some gaming applications, but didn't want to give any further details."

Re:Uhh, Price?

[t00le]

ffs rtfa:

"If you were crazy enough, you could use this in a high end game machine."

So how much will these cards cost? Flynn told us that the company is aiming to beat $30 dollars a GB, something that should seem very cheap to large corporations, adding "You can drop ship or Fedex this card and be up and running in a few minutes... you can't do that with a storage area network."

Re:Uhh, Price?

[ArcherB]

Who, what, when, where, why?

Price would seem to be a pretty important detail...

$30 a GB.

Re:Uhh, Price?

[Trigun]

I'll take 2 gigs please.

Re:Uhh, Price?

[BlueBlade]

Not to nitpick, but you're not even close to the price of flash memory with your estimate there. $29 per GB? Newegg lists several 16GB USB key drives for about $130, or $8.12 per GB. And that's retail. It's safe to assume that the actual price per GB for a mass-produced drive would be much lower than that. Even at $8, that hypothetical drive would be about $5120.

Right now, flash RAM prices drop in half every 6 months or so, meaning it won't be that long until this drive isn't outrageously expensive anymore.

Re:Uhh, Price?

[CastrTroy]

I could buy a 10 GB drive for most of my OS and software, and just keep my media on a traditional hard drive. You don't need a super fast drive for your MP3s and Videos, but it would be nice to increase boot times as well as application start up times.

Re:Uhh, Price?

[Black Copter Control]

If price is an important detail, then you should wait 5 years for this to hit the mass market.

Speaking as someone who remembers paying $2500 for a 1GB drive (and that was with a 30% educational discount...
If that kind of I/O speed is important enough for a business, I can see them not only paying the $2400 for an 80GB drive, but buying 5 of them to put in a 4 disk raid-5 configuration with hot-swap spare. (because FLASH drives have a nasty history of dying after too many writes when used in place of an OS drive).

As long as the performance increase provides a business advantage, it's worth it... Think of it this way... that 5-disk configuration would cost them about the same as 1-2 months worth of wages for the senior sysadmin who's tasked with installing and maintaining it.

Re:Uhh, Price?

[walt-sjc]

Um, we were using large RAM disks (the kind that hooked up to SCSI, had a built in UPS and disk to dump RAM contents) many years ago (8 now?) to speed up databases. That was limited by the SCSI bus, but access time and latency were near zero (which was awesome.)
Of course, large back then meant 4G, and the average hard disk was 9G. This is evolutionary, not revolutionary.

Re:Uhh, Price?

[walt-sjc]

Um dude? He is quoting the FA. It's pretty clear that this drive will release at prices closer to $30/G. Maybe next year they can get down to $8...

Re:Uhh, Price?

[hedwards]

Exactly, organizations will make a decision as to how many of these they want based upon the cost, performance and their needs. Considering the other costs of a large set up, I wouldn't be surprised if these wound up in the DB servers. Especially ones which were primarily read only under normal use.

Of course, I don't personally know if PCIe is hot swappable, which could make a difference for some people, but with the bandwidth and simultaneous connections, it shouldn't take very long at all to get a new disk up and running from scratch.

The price issue will work itself out, and It would be potentially worth it to some people out there, just to have for a small amount of swap. 2 or 3 gigs wouldn't been unreasonable for people that are looking for the highest possible performance, and I'm sure there are a few that would be willing to pay for 4 or 5 gigs for a home machine at those prices.

The performance numbers look very, very good. Beating the crud out of the ram based disks that I've seen benchmarked.

Re:Uhh, Price?

[Hal_Porter]

Nand flash is about $7 per GB

http://www.dramexchange.com/#fspot

So 640GB should be about $4480. That's the spot price on the chips too, a device with them in will be more expensive. And it would need to be a hell of a device. 160*32Gigabit (4Gbyte) devices for example. That would be a hefty PCB. And the controller that performs the way they say is not going to be cheap either.

Re:Uhh, Price?

[hackstraw]

I could buy a 10 GB drive for most of my OS and software, and just keep my media on a traditional hard drive. You don't need a super fast drive for your MP3s and Videos, but it would be nice to increase boot times as well as application start up times.

As an cluster guy, I would like to have these things as cheap, reliable, and removable. Size does not matter too much here.

As a server guy, I would like to have the same, but also with larger capacities.

Imagine a life w/o ever having a RAID array throw a disk. Imagine a life w/o having a RAID array.

Now, if only we could get rid of tapes...

Re:Uhh, Price?

[BlueBlade]

Hum yeah sorry, classic case of didn't RTFA. It's just that anyone claiming that flash costs $30 per GB is smoking some seriously good stuff (or is a govt employee paying $50k for a few hundred rolls of toilet paper or some such).

Re:Uhh, Price?

[ady1]

29$ a GB and you are modded informative? Here is a report reflecting the current Nand Flash prices. Shows that its not far off in future that these things will be available for desktops on a reasonable and competitive price.

Re:Uhh, Price?

[harryk]

I'm guessing you didn't read the article. Fusion IO hopes to beat $30/GB. True, it did not give a specific price, but the reference is available.

What I'm more interested in is which PCIe bus it plans to use? Is this 1x, 8x, 16x ??

Re:Uhh, Price?

[Ceriel Nosforit]

Sorry, but that doesn't seem to beat an inexpensive SAN in price. You're forgetting maintenance costs. Solid-state devices have a lot longer MTBF than mechanical devices, and I assume they need less power too. Both of those get rather costly.

Re:Uhh, Price?

[GungaDan]

Judging from the pictures in TFA, 1x.

Re:Uhh, Price?

[eth1]

For bulk storage, this seems very expensive...

I run quite a few Linux mail relays, whose throughput is essentially limited by disk IO ops per second... It looks a little better then, even at $30/GB:

6-disk RAID 10: ~200-300 IO ops/s - $3,000
80GB solid state drive: 100,000 IO ops/s - $2,400

I'll take the flash card :)
You get more space out of the disks, but we never use more than a few hundred meg on a relay

Re:Uhh, Price?

[God'sDuck]

What I'm more interested in is which PCIe bus it plans to use? Is this 1x, 8x, 16x ??

FWIW, you would need at least 4x to get the benchmarks in TFA.

Re:Uhh, Price?

[Anonymous Coward]

$30 a gig is more than twice the cost of most of the current SSD on the market.

Re:Uhh, Price?

[Chris whatever]

Oh no my Area 51 alienware dream computer just went from 12000 to 14500 $ now i'm never going to get it :(

i dare you to click on BUY IT!

Re:Uhh, Price?

[SnarfQuest]

At $30/GB, that would make an 80GB drive around $2,400. Hopefully the price will eventually come down, as "He even hinted that the company is looking into some gaming applications, but didn't want to give any further details."

So, a playstation 4 starting at $5,000 for the 80Gb unit, and $25,000 for the 640 Gb unit?

Re:Uhh, Price?

[edmudama]

If you never use more than a few hundred meg on a relay, why not just add a $30 stick of RAM, and get a million ops/sec?

Re:Uhh, Price?

[edmudama]

I'm guessing "Gaming application" would translate to speeding the cluster server-side, not something like a PS4 or WiiII.

Halving the number of disks they need for the same performance might allow for double the players per hardware investment.

--eric

Re:Uhh, Price?

[Kildjean]

they are charging $30 per Gb, so lets see 649Gb will be a measly $19470

Re:Uhh, Price?

[mr_mischief]

Unfortunately, you can't pop the cover off of most PCs, pull something out of a PCIe slot, and swap in a spare while the system is running. You can do that to a failed drive on lots of SANs. Perhaps someone will develop a way to have these hot-swappable too.

Re:Uhh, Price?

[Guspaz]

Except game servers are by and large not disk IO limited. They're usually either CPU limited or transit limited. You might be memory limited, but that's far cheaper to fix than the other problems, so it isn't really a concern.

Game servers only really touch the disk during map loads.

If you're building a really beefy box to use as a game server, say a quad core quad processor box with 16GB of RAM, and you're running, say, 32 game servers on there, then yes, the chance that two servers will try to load a map at the same time increases, but that's probably better solved with a cheap RAID array. For perspective, said server could easily saturate a 100mbit connection, so transit required could be somewhere in the area of 48TB/mth. I know that Cogent charges my ISP about 4 cents per gig for transit, so we're talking about $1500 for transit and $1000 for the physical connection. So internet connectivity would cost $2400 per month, and that'd be a far more pressing concern than disk IO.

On the other hand, if you start talking about MMO-type games, that's a different story. I have no idea what the bottlenecks tend to be on that sort of server.

Re:Uhh, Price?

[nuzak]

> Of course, I don't personally know if PCIe is hot swappable

It is. But so is PCI -- it's just a matter of whether the physical and electrical connections on your hardware allow it, and whether your OS is set up to handle it. I wouldn't recommend yanking out your video card while the PC is on.

Re:Uhh, Price?

[vidarh]

Because he doesn't want to lose the data when the power goes or the machine crashes.

Re:Uhh, Price?

[Ender77]

remember, this is a brand new technology and whenever something new comes out it is always expensive. The first cd burners, gig hard drives, latest video card..etc were in the thousands, you wait a few years and you can find the same technology in the bargain bin for 30 dollars.

Re:Uhh, Price?

[Anonymous Coward]

I can say from experience that RAM is NOT hot-swappable (I fried a motherboard and a stick of RAM). Given the similarities of the slots, I wouldn't try it with PCIe either.

Re:Uhh, Price?

[kiwipeso]

I was thinking about a MMO game server that was based on a peer to peer network.
I would be open to allowing dynamic hosting of servers on the grid.
This should allow for local peering on the same network (or city) and make it cheaper for ISPs.

I was thinking that maybe a best fit would be one map zone per server, running from a common master map.
I would make it Player vs Environment on the low zones, then open up to Player vs Player on tougher zones.
Another thing would be running cities as large instances on their own server, and only letting auctions be globally linked to cities of the same faction.

Re:Uhh, Price?

[WuphonsReach]

Personally, I'd still RAID them for mission-critical data. But they would indeed have a lot of advantages over magnetic disks once the price drops below $10/GB. Maybe even $20/GB...

Re:Uhh, Price?

[jack22]

Two things to consider- 1. You are assuming they are using MLC Flash (as opposed to SLC) 2. No one is going to stuff flash chips on a board like this and sell it for cost. You could point out the same hefty price about most hardware if you pick apart the raw cost of its primary components. I would imagine they plan on turning a reasonable profit as well as any future competitors. Flash pricing trends should help make this type of technology more and more affordable over time. I think focusing on enterprises with this product is the right move since what they have does indeed represent value to much more than a handfull of companies and will enable the technology to evolve and trickle down to the "regular joe" systems like mine.

Re:Uhh, Price?

[Hal_Porter]

1. You are assuming they are using MLC Flash (as opposed to SLC)

Hmm true. MLC is about twice as cheap in the most recent generation.

Presumably two bits per cell halves the die size. Actually it's more complex than that - the SLC cost / MLC cost varies with the generation.

Gb SLC cost/ MLC cost
16 2,1
8 2,6
4 1,4

But MLC has 1/10 the maximum erase cycles

http://www.datalight.com/products/download.php?type=public&resourceid=416

So it's not really suitable for a massive storage array. It would be great for an MP3 player though.

No one is going to stuff flash chips on a board like this and sell it for cost.

Yeah, I know - that's what I meant by 'That's the spot price on the chips too, a device with them in will be more expensive'

in mother russia

[Anonymous Coward]

massively large PCI-e drive stores you!

Still Expensive

[EvilSpudBoy]

$30/GB is still expensive but I bet in less than 2 years solid state drives in this capacity will come down to a few dollars/GB.

I hope this means that laptops and large capacate media players with extremely long battery life are not too far away.

Re:Still Expensive

[Mr. Underbridge]

I hope this means that laptops and large capacate media players with extremely long battery life are not too far away.

At least as far as laptops are concerned, I thought a limited number of write cycles was a problem for solid-state drives.

Re:Still Expensive

[Jeff DeMaagd]

I hope this means that laptops and large capacate media players with extremely long battery life are not too far away.

I think people expect too much from SSDs. The hard drive is far from the dominant power consumption component in a notebook. The CPU, chipset, GPU and display panel each consume more power than a notebook hard drive does. If you follow a modified version of Amdahl's law (not a law, but whatever), you want to fix the biggest problem first, and that is either the display or CPU. An LED backlit display can save some power, and running a lower power rating CPU saves power too. Compared to that, the savings of swapping HDD for SSD is negligible. On a standard notebook, I think you might add 15 minutes to battery life, which is still far from "extremely long battery life".

In media players, doubling in capacity every year is a reasonable expectation.

Re:Still Expensive

[morgan_greywolf]

$30/GB is still expensive but I bet in less than 2 years solid state drives in this capacity will come down to a few dollars/GB. The price will fall, but not that drastically. Magnetic hard drives and related technology are likely to continue to dominate for many more years due to their price being pennies a GB and their performance being deemed good enough, especially in RAID systems with large caches.

Re:Still Expensive

[duerra]

Even at a few dollars per GB, that is too expensive to hope that this technology could re-introduce cartridge-based gaming as opposed to disc-based gaming on consoles. I can't wait for the return of 3 second load times in gaming - having to wait 30 seconds for a level to load (especially in games where the action you are performing, and may want to repeat, is as long as the load time).

Re:Still Expensive

[Eccles]

Is there any sign they have a significant patent on this tech, or are they just counting on being first to market? If there are no patents, we could expect the price to go down fast, just price flash drives.

Re:Still Expensive

[Anonymous Coward]

You youngins... I remember when $30/MB was CHEAP and we liked it!

Re:Still Expensive

[adisakp]

$30/Gb is $2400 for the 80 Gb drive. That's a lot of money. Then again, I paid a similar amount of money for 64MB of RAM back around 1993 which is pretty hard to believe. The price will eventually come down but for right now, it's not terribly practical until they can get it to under $500 for your 80Gb super-fast drive. Then again, think how fast you could boot/sleep/swap at 600-800MB/s with almost no penalty for seeks.

Re:Still Expensive

[dgatwood]

The biggest power sink is quite often the DRAM. The backlight isn't even in the running anymore, typically only drawing on the order of 8-10 watts for the entire panel. From VIA's 1 watt CPU to the Core 2 Duo Extreme at I think 130 watts, there's a wide range of power for CPUs. A flash disk would be measured in tens of milliwatts, while a hard drive ranges from a couple of watts up to seven or eight, depending on the number of platters, the rotational speed, and whether it's a laptop or desktop drive.

For RAM, consumption can range from 1-20 watts per gigabyte as best I could determine from various websites. Thus, a machine with 8 gigs of the wrong kind of RAM could potentially be using 160 watts for RAM---more than for the beefiest Core 2 Duo Extreme out there. Of course, obviously you couldn't do that on a laptop; laptop RAM is more on the order of 2.5 watts per gig, or 20 watts for eight gigs. Either way, though, that's some serious power consumption, and even in a modest 1 GB configuration, your RAM is drawing almost as much as your hard drive.

By contrast, eight gig flash parts would probably draw tens of milliamps, and only during active use. When not doing something, each flash part would draw almost nothing, unlike DRAM, which has to be refreshed continuously. Thus, if we could get flash costs, performance, and reliability to the point where flash could replace DRAM, it would be a major power consumption win. Of course, that's another story for another year, but....

In terms of replacing hard drives, though, the benefit is still there, and is pretty significant. Your typical laptop could shave off about 3-5 watts of power by moving to flash instead of a hard drive. That's 10-15% of its total typical power consumption. If your laptop gets five hours on a charge, that's a whopping 52 minutes of extra battery life. It might not be "extremely long battery life", but it would be a significant improvement.

Re:Still Expensive

[AaronLawrence]

Interesting figure, but I don't think that can be true, simply because most ram runs without heatsinks. Anything over about 5 watts would need a heatsink, and 20 watts would be quite large.
I'm aware there are RAM heatsinks out there, but mostly they are not *required*, only a flash add-on for those who O/C or whatever.

Re:Still Expensive

[dgatwood]

That 20W number might be wrong. The part in question might also have an integrated heat sink---I haven't seen pictures. I think a more typical range is 2.5-10W per gig. In any case, even at 2.5W per gig, with that new VIA chipset and CPU, three gigs of RAM would be almost as much power as the CPU and chipset (including GPU) put together. At 5W per stick, the RAM would be half again more.

Also, bear in mind that this is power consumption for an entire stick, not for a single chip. If a DIMM with chips on both sides drew 20 watts, that current drain could potentially be spread across as many as 32/36 chips (parity/non-parity). Those components would then only be drawing 555-625 mW apiece. I'd be surprised if such a part required a heat sink (assuming you have reasonable airflow).

That said, yes, a 20W DRAM stick is obscene consumption....

The return of HSM?

[Wesley Felter]

With small but fast flash drives appearing on the market, it would be nice to have storage systems that can automatically migrate data between disk and flash to maximize performance.

Re:The return of HSM?

[nuzak]

> it would be nice to have storage systems that can automatically migrate data between disk and flash to maximize performance.

Vista does this now.

1x10^6

[Anne_Nonymous]

640 GB ought to be enough for anybody.

Re:1x10^6

[FinchWorld]

640 GB ought to be enough for anybody.

Please, my po.. er... some peoples porn collections are bigger than that!

Re:1x10^6

[rbanffy]

+1 funny.

Sorry. Have no modpoints today

And another question.

[AltGrendel]

What's the MTBF?

Re:And another question.

[LiquidCoooled]

Hitachi are saying that they have solved the overwrite problem (at least mitigated it by a factor of 100)

They appear to want to use normal DRAM memory for the running of the drive but then write it permanently to the NAND flash at shutdown/memory full time.
I would assume this involves charging of a small battery and dumping the data later on.

http://www.theinquirer.net/gb/inquirer/news/2007/09/26/hitachi-reckons-solid-state

Re:And another question.

[fragmentate]

The only MTBF I've found is based solely on stress testing. Estimates vary wildly from 500,000 hours, up to 1 million hours.

I think that getting this to market will improve MTBF (and, of course, give us REAL numbers). It will also spur competition in two areas. First, other companies that provide storage solutions will take it seriously, which may drive prices way down. Second, the NAND flash drive companies will try to improve MTBF to be able to sell more flash. Of course, their prices may drop as well -- spurred by competition.

I haven't personally experienced any flash RAM failures, yet. I used my flash drives quite heavily, and swap data on them frequently. So far, so good; and I'm not even a big spender (in fact, I buy the cheapest I can find).

Going to flash based storage would be a boon for RDBMS (DMBS in general, really). And people would devise clever ways of syncing to magnetic media (warehousing?), and using the flash for the hardcore querying.

I want one of these drives... NOW!

Re:And another question.

[Anonymous Coward]

Another good question that TFA didn't answer is what average seek times are. I'd imagine it's better than a standard disk drive, but it would still be nice to know.

Re:And another question.

[brunascle]

i suspect it's around 0, give or take a few nanoseconds.

Re:And another question.

[CastrTroy]

500,000 hours is a long time. That's 57 years. I've had many traditional drives die. And none of them had anywhere close to 57 years. The oldest operational drive I have is 10 years old, but it probably only has 4-5 years maximum of time that it's been on. Are you really going to care about your hard drive 57 years after you buy it? And those are old smaller capacity drives. The newer higher capacity drives tend not to last as long. Most drives I wouldn't trust past 5 years of usage.

Re:And another question.

[drxenos]

More likely it uses hold-up capacitors. We use NVSRAM that does this. It will permanently store the memory when it detects a power lose.

Re:And another question.

[Anonymous Coward]

I don't even think they could do accurate testing at this point. Once these types of drives are available there will be software that starts using these in way we can't use typical hard drives (think massive parallel streams and such). This will certainly change the MTBF results because right now the testing is done in a typical hard drive style usage where the memory based device will need to be thrashed a lot harder to represent its eventual real world usage pattern.

Re:And another question.

[norton_I]

MTBF is usually misunderstood. This is largely because it does not measure the mean time until your drive fails, despite the name. MTBF is usually quoted for a 5 year drive life. So, if the MTBF is 50 years, and you have 100 drives, you expect 2 failures/year, as long as you replace all the drives when their 5-year life is up, failed or not.

Hence, it is a metric only of direct importance to people estimating failure rates for RAID arrays and the like.

Re:And another question.

[Threni]

> What's the MTBF?

Exactly.

> the worst case scenario by using small 4K blocks and then streaming eight simultaneous 1 GB reads and writes. In that test, the ioDrive clocked in > at 100,000 operations per second. "That would have just thrashed a regular hard drive,"

How does the solid state memory perform when you perform 100,000 write operations to the same address? Let's hope virus writers don't start trashing them, huh?

Re:And another question.

[fbartho]

The firmware load leveler doesn't let you do that... Because it's no longer/harder for a flash drive to write to any location than any other, even if you try to overwrite the *exact* same location many times, the firmware can bounce your write location around the disk, transparently. No worries on that front!

Re:And another question.

[Threni]

> The firmware load leveler doesn't let you do that... Because it's no longer/harder for a flash drive to write to any location than any other, even
> if you try to overwrite the *exact* same location many times, the firmware can bounce your write location around the disk, transparently. No
> worries on that front!

So if you had a 10gig file on the drive, and then I wrote 100,000 times to, say, 50 locations at various offsets into the file it'd move the file all over the place to prevent rewrites?

Re:And another question.

[fbartho]

Basically, from my understanding of the technology, as long as you have any free space on the drive, when you write, it will try to balance the writes so any individual cell is written to equally. The first solid state drives had much more trouble in addition to having a lower max write count per cell, in that they had many fewer cells, so there wasn't much room to balance. I remember in a discussion on slashdot about this in the past 4 months where someone who was an hd engineer linked the thread to a paper describing how the load levelers worked, and how if you had an 80gb drive with 20? 10? gb of free space, then the failure rate due to write exhaustion was far outside the useful lifespan of the drive. Basically continuous writes didn't cause any reduction in capacity due to exhausted cells for >6 years. I wish I had saved the link to the paper. But it was enough to convince me that my next machine is going to have a core OS + applications partition sitting on a raid 0+1 of flashdrives and then the rest of my media will be on external large size (and cheaper) sata drives. January looks to be when Seagate will release their entry into the market, and hopefully prices will be affordable.

Re:And another question.

[Threni]

I can quite believe that there are great algorithms and firmware for regular usage - I'm not knocking them - but I just can't see how you can do anything about the pathological case(s). If you can get low level access to the drive (ie like Partition Magic does for hard drives) you can do what you want. Perhaps you can't.

Re:And another question.

[Anonymous Coward]

It's the Mean Time Between Failures, but that's not important right now.

Re:And another question.

[fbartho]

I understand what you're describing, and I don't actually know for certain there isn't a way that an application within the OS can directly write and rewrite to a specific cell. My understanding it wasn't possible with the solid state drives on the market, the firmware was embedded in the drive, and the OS had no way of addressing individual cells, it could only address specific memory locations that would be remapped to cells at the firmware's discretion. Partition Magic, with my experience with it, can't even directly access the harddrive in that manner, without booting into a custom mini-os. You can configure all your changes with the GUI tool, but for many of the big ones, it needs to unload the OS before shifting things around, and so does so on reboot.

Re:And another question.

[TinyManCan]

Wear leveling is handled in the firmware on the flash storage device itself.

It maintains a table which matches the logical block addresses to physical cell locations on the flash memory. The mapping between the logical and physical locations in handled by the wear leveling algorithms, which you can not inspect or influence from the devices data interface. Thus no program running on the host device can create a pathological case of wearing out a specific cell. Only bad firmware could do that.

Re:And another question.

[rcw-home]

Hence, it is a metric only of direct importance to people estimating failure rates for RAID arrays and the like.

Exactly. Which is why any MTBF statistic that is unaccompanied by a service life statistic should be dismissed as handwaving. People don't really believe that the coolant containment on an ESBWR nuke plant (expected to see 3*10^-8 common mode accidents per year) is going to last 33 million years, do they?

Re:And another question.

[jgoemat]

They claim 8 years of normal use. NAND flash can last to 100,000 write cycles and read cycles don't affect the life of the memory. They use an algorithm to spread the write usage around so the whole memory will wear evenly. With the 640 gig drive, you can write to it at 96 megabytes per second for 21 years before it will go bad (640,000 mb / 96mb/s / 365 days/year / 24 hours/day / 3600 seconds/hour * 100,000 write cycles). They also use error detection and correction algorithms, but I'd still recommend creating a software RAID array of these as you would any other drive.

Wow

[thatskinnyguy]

I could imagine using this as an OS drive. No sooner do you let your finger off the power button than the login screen appears.

Re:Wow

[Jerome H]

Welcome in Utopia, I hope you will enjoy your trip !

Re:Wow

[east coast]

What's the big deal about that? My Commodore 64 can do the same thing.

Re:Wow

[thatskinnyguy]

Yeah? How does WoW play on your C64?

Re:Wow

[z0M6]

Sorry, but I think you are a bit optimistic here. I recommend that you have a look at some of the videos of people on youtube that boot xp on ramdrives. You should also notice another great lag. Hint: the time spent from pushing the power button to booting the OS. http://linuxbios.org/ could fix this.

Re:Wow

[FiveLights]

You are standing in a field, facing East. A river runs East and West to the North of you and a path runs to the South and East.

Re:Wow

[HarvardAce]

You are likely to be eaten by a gnome.

Re:Wow

[TheDreadSlashdotterD]

grab rock

smash self

Re:Wow

[east coast]

Yeah? How does WoW play on your C64?

It plays very nice.

[turns to the other Commodore users] I told him it plays very nice [chuckling from users]

Now! Go away or I shall taunt you a second time.

Re:Wow

[east coast]

You are standing in a field, facing East. A river runs East and West to the North of you and a path runs to the South and East.

Come on now, this is MMO! It should read:

You are standing in a field, facing East. A river runs East and West to the North of you and a path runs to the South and East.
Also here: John lvl 12 Cleric

Re:Wow

[zlogic]

My PDA boots in about 30 seconds, and it doesn't have a harddrive. Booting isn't just loading stuff from a drive, it's hundreds of tests like
- hardware changes
- hardware initialization (e.g. loading firmware)
- searching for drivers
- applications acquiring and releasing resources and checking for stuff like library versions, user names etc.
That's why BIOS initialization often takes time, and yet it works even if the system has no drives.

The only way this would work is hibernating, but hardware would still need to be initialized.

Re:Wow

[blankoboy]

What, disk speed has somehow eliminated the POST process? =)

Re:Wow

[Actually, I do RTFA]

Come on now, this is WoW! It should read:

You are standing in IronForge, facing East. A chasm runs East and West to the North of you and a path runs to the South and East.
Also here: John lvl 12 Cleric
John1 lvl 13 Cleric
FarmerJohn00 lvl 70 Farmer
FarmerJohn01 lvl 70 Farmer
FarmerJohn02 lvl 70 Farmer
FarmerJohn03 lvl 70 Farmer
FarmerJohn04 lvl 70 Farmer
FarmerJohn05 lvl 70 Farmer
FarmerJohn06 lvl 70 Farmer
FarmerJohn07 lvl 70 Farmer
FarmerJohn08 lvl 70 Farmer
FarmerJohn09 lvl 70 Farmer
FarmerJohn10 lvl 70 Farmer
FarmerJohn11 lvl 70 Farmer
Press any key to continute...

Re:Wow

[langelgjm]

[joke]Sadly, it only scores a 4.1 on the Windows Experience Index.[/joke]

Re:Wow

[whopub]

I could imagine using this as an OS drive. No sooner do you let your finger off the power button than the login screen appears. Yeah. Vista users will start calling it "crash button" instead.

Re:Wow

[Big Nothing]

"I could imagine using this as an OS drive. No sooner do you let your finger off the power button than the login screen appears."

Unless, ofcourse, you're using Windows Vista.

Re:Wow

[dextromulous]

I could imagine using this as an OS drive. No sooner do you let your finger off the power button than the login screen appears. Good luck with that. Unless it's on an embedded device with no BIOS, you're stuck waiting for at least several seconds. Without a PC BIOS, there are some OSes that will boot sub-second if tuned properly. Get into remote monitoring equipment, and OS boot times are smaller still (measured in microseconds, usually.) It helps when you need every microwatt of battery power you can get.

Re:Wow

[phorm]

Mine (HP iPaq) usually starts a lot faster than that... if you're counting the time from the reset switch to when it asks for my password/thumbprint. 30 seconds sounds awfully long for a PDA, see cellphones (not counting the time to connect to a nearby tower).

Re:Wow

[eatont9999]

I hope you guys know that a maximum data transfer rate of 800Mb/s is 100MB/s. We don't know if it is sustained or peak so we are not even guaranteed that. Maybe it does not even matter. A few years ago they came out with this thing called SCSI U320. It is an interface that can support drive speeds up to 320MB/s per channel. If you have a decent controller and a RAID 5, you can get very close to 320MB/s sustained data transfer rates. Don't even get me started on SAS arrays! Sure it may be a little too enterprise level for some people, but before spending the money on unproven hardware, I would choose a proven robust interface. Now, if the author made a mistake and these things run at 800MB/s, then we have a new issue in which we can start talking about 4-10Gb/s Fibre SAN arrays, but I'll leave that for another time. Sorry to shit in anyone's hat, but those are the facts man.

Re:Wow

[vertinox]

Are you sure that isn't the splash screen? Mine wants me to watch the dancing Sprint logo no matter what and play their jingle. It doesn't need to check the hardware each time. Also considering my OS X and WinXP boot faster than 30 seconds.

Re:Wow

[Emetophobe]

The story submitter and/or editors made the mistake of using Mb instead of MB. This is actually one of those times when you will want to RTFA. The article is only 6-7 paragraphs on a single page and there are 2 relatively short videos. like I said, it's worth checking out.

From TFA:

So how fast is the ioDrive? Flynn said the card has 160 parallel pipelines that can read data at 800 megabytes per second and write at 600 MB/sec. He even proved it by running a Linux drive I/O benchmark

If you watch the benchmark video, it actually gets over 900MB/sec reads and 800MB/sec writes, faster than those quoted numbers.

Re:Wow

[hacker]

You mean like i-Ram?

Look at how fast that machine boots to the desktop from a cold boot. And that's XP (!!)

Re:Wow

[zlogic]

It's an iPaq hx2750 (or hx2790, I don't remember). It has a fingerprint scanner, wifi and bluetooth and it seems to be searching for drivers for the wireless modules. Icons also take ages to load, often opening a directory with a large *.exe, like Bejeweled, takes about 10-20 seconds.

write limit?

[Lord Ender]

What is the write limit for their device? It would be a damn shame to buy a $10,000 solid-state drive only to have it burn out after a month because you forgot to mount it with the "noatime" option.

Re:write limit?

[Anonymous Coward]

http://www.fusionio.com/faq.html says the following:

Doesn't NAND Flash wear out?
ioMemory(TM) uses sophisticated error correction techniques and wear-leveling to minimize fatigue and extend the service life of the NAND components within the ioDrive(TM), which results in a service life of eight years compared to the five year service life of mechanical hard drives. When an ioDrive's(TM) ioMemory(TM) does finally start to fatigue, the result is a reduction in available free space - no data is lost and the system remains operational.

Re:write limit?

[Lord Ender]

That's a standard feature these days. I want to know how many times I can overwrite the entire disk with /dev/random. If I did a "dd if=/dev/random of=/ioMemory size=[size of ioDrive]" how many times could I execute that command before the thing goes tits up?

Re:write limit?

[TinyManCan]

Somewhere between 100k and 1 million times.

Cosidering that this drive is 640GB, that means you would need to write somehwere in the region of 61 PETABYTES of information.

You'd have to write to the drive at a perfect 800 MB/s for 941 days to hit that mark.

It could last as long as 30 years, at full write speed of 800 MB/s if it can handle 1M writes per cell.

At the end of the day, semiconductors this large and high quality are certainly better than tiny bits of rust on rapidly spinning platters.

Re:write limit?

[Anonymous Coward]

you would need to write somehwere in the region of 61 PETABYTES of information

Assuming even writing. In real world devices, there are some regions which are written to rarely, and some which get constantly rewritten. The sector containing you logfile directory information starts to go to pot after about 70 days, assuming a file-access time bump once a minute.

But if they're smart (and they probably are) they'll have some dynamic sector relocation scheme going on. The net effect is you'll slowly see your total net drive capacity drop as sectors become unreliable. But again, it'll probably take some time to drop significantly.

Re:write limit?

[canuck57]

That's a standard feature these days. I want to know how many times I can overwrite the entire disk with /dev/random. If I did a "dd if=/dev/random of=/ioMemory size=[size of ioDrive]" how many times could I execute that command before the thing goes tits up?

Not as many as you would think nor as many as the manufacturers would have you believe. Those devices, while handy as all can be to copy a DVD sized image between two distrusting hosts is handy, it is not a substitute for a HD. I have seen them go in 200 writes. But usually much more is usual. And not all brands are the same.

But they are getting to a size and with a little cost reduction compete with CD/DVD-ROMs, send out the tunes in a ultra small package, and re-usable. Quick, patent it so the patent trolls don't screw up the idea!

At some future, perhaps 10-20 years we will see solid state drives, but it isn't going to happen sooner without some pain and cost. 640GB of the stuff can't be cheap. And the rewrite issue has to be solved. And Seagate and others are looking towards 2TB and up in future developments. While the gap is closing, HD shipments are safe for at least a decade.

But one interesting idea posted I saw was, put a normally read only OS in the device, write little, load like lightening and use RAM there after except for the occasional patching. Get a worm, reboot to the static image. Use HD for user storage only.

Re:write limit?

[Anonymous Coward]

Heh, I'd like to see you write on a hard drive sector 100K - 1Million times. Even worse try that on the WHOLE drive, and not a sector. A 500GB drive writing at a constant(and optimistic) speed of 60MB/sec would take over 2 hours just to do it ONCE. It would take over half a millenium just to /dev/rand the drive the first 100K times!

I don't think the MTBF is going to be a major issue.

Re:write limit?

[TinyManCan]

Certainly they are using wear leveling. They are probably (information is thin) also over allocating storage, such that 10-15% can fail before impacting the advertised free space on the device. In reality you will see your full 640GB of storage, which you could write 61PB of data to the very first sector on the disk over and over and never experience any issues. Before the last 'extra' block is used up, you'll get an alert and replace the device.

Re:write limit?

[Bazer]

I hasn't been "rust" for a long time now.
Just like those semiconductors don't amount to tiny, rapidly electrocuted bits of "sand".

Can you boot from them?

[IceCreamGuy]

I can't find any mention in the article of whether you'll be able to boot from them with current BIOSs. Surely any system they'll be in will have a decent amount of RAM for the OS, but it would be pretty cool solely for the fast boot times.

Re:Can you boot from them?

[king-manic]

I can't find any mention in the article of whether you'll be able to boot from them with current BIOSs. Surely any system they'll be in will have a decent amount of RAM for the OS, but it would be pretty cool solely for the fast boot times.

Hey,
We all use Linux right.. we never have to reboot /sarcasm.

Re:Can you boot from them?

[mathx]

at 1000x the cost of hardrives thats ALOT of $/second. Considering BIOS post and kernel device detection is currently the big slowdown in linux (who the hell KNOWS why windows is so slow to boot, its an opaque process), you wont be gaining much with a super fast disk. Perhaps 5-20% on boot, which is about 5-10s max.

$20,000 for 5-10s boot max? Cheaper to just leave the machine on longer and pay the power bills and suffer no boot (or if you're a poor sap using Metacity, the startup time for that too!)

Hype isnt really useful once the initial endorphin rush wears off. You still have a $20k hole in your wallet.

Re:Can you boot from them?

[Anonymous Coward]

Current BIOSes can do that just fine. My laptop (2 years old toshiba M50-MX2, cheapest I could find anywhere at the time) uses windows but I installed fedora on an external USB hard drive and it works just fine. I also got Ubuntu working on a 2 GB USB memory stick. If this cheapass laptop can handle USB booting, pretty much any computer should manage with an internal drive. I'm not that kind of techie but I guess that the BIOS is oblivious to the internal workings of the drive, all it cares it the communication protocol (IDE, scsi, sata etc)

Re:Can you boot from them?

[IceCreamGuy]

all it cares it the communication protocol (IDE, scsi, sata etc) That's exactly what I'm saying and why I asked. Judging from the article, it seems apparent that they're not IDE, SCSI, USB, SATA, or any of the current standard interfaces; they are attached through the PCIe bus. I remember when USB storage devices first came out that it took a little while for BIOS manufacturers to support booting from them, yet the OSs had no trouble using them. Support for storage devices these days seems to start at the software level, and trickle back down to the hardware level for things like booting. Then again, it could have it's own BIOS, like SATA controllers and PCI disk controllers - I guess I really wasn't thinking too hard when I asked the question. The points the other replies bring up, however, makes my question pretty meaningless also; I see now that it wouldn't really serve too much of a purpose booting from them anyway, and I really didn't consider the price point too much. I'm sure by the time they're affordable, BIOS support will exist for them if it doesn't already anyways.

Re:Can you boot from them?

[Helios1182]

For you, the home user, your time isn't worth those $/second. For an enterprise with I/O limited systems it could easily be worth the money. If they didn't think they could sell it, they wouldn't be wasting resources designing it. Then in a few years we will get to use the results for a fraction of the cost in home systems.

Number of write cycles?

[Reality Master 201]

Will it last as long as a standard hard drive?

Re:Number of write cycles?

[Joce640k]

Longer...MUCH longer.

Re:Number of write cycles?

[canuck57]

Will it last as long as a standard hard drive?

Not a chance. Maybe in time. But don't count the days or wait for it. The HD is tried, true, cheap and is not going outdated any time soon. The time will come, but you could become very old waiting.

Lifespan?

[rossdee]

I thought Flash memory had a limited (write cycle) lifespan. If they have solved that problem it would be great.
When are they coming out with a laptop version?

Re:Lifespan?

[Anonymous Coward]

Your concern is answered fifty times in every flash HD article here.

Expect to pay big

[PhrostyMcByte]

So how much will these cards cost? Flynn told us that the company is aiming to beat $30 dollars a GB

So the initial 80GB model will hopefully cost less than $2400. I suspect these will be limited to servers demanding immense I/O with a large amount of data.

Re:Expect to pay big

[Ant P.]

Sounds like it'd be more cost effective to just have a ton of RAM and a good UPS.

Re:Expect to pay big

[Anonymous Coward]

I suspect these will be limited to servers..

I remember hearing the same thing about the 80386.

Almost no tech is ever limited to servers.

Re:Expect to pay big

[Teun]

It sounds like a gamers dream, they are the driving force behind a lot of speed improvements and will pay about anything to have the latest HW.
On the other hand, do (present) games really benefit from this massive speed bump?

Re:Expect to pay big

[petermgreen]

ram, especailly server ram is more that $30/gigabyte.

also even ignoring the issue of power using system ram for long term storage is a bad idea because it is so vulnerable to crashes and you can't have very much of it (32 gigabytes is the limit of most server boards you see and that is only achievable by using very expensive 4 gigabyte sticks). Ramdrive cards with built in battery backup do exist but that drives up the price even more as you have to buy the adaptor card as well as the sticks of ram.

Re:Expect to pay big

[TemporalBeing]

I suspect these will be limited to servers demanding immense I/O with a large amount of data.

I could see some uses for them, especially in environments where things will get bounced around quite a bit. For example, airplanes, military vehicles, off-road vehicles, etc. could all benefit.

Hmmm....I wonder how good it would be for submersibles, perhaps they could push new depths since they wouldn't have worry about the air pressure within the hard drive enclosures.

Re:Expect to pay big

[heinousjay]

Just add the "...at first" mentally and you'll feel better.

Re:Expect to pay big

[Nicolay77]

Half-life 2 level load times come to my mind.

on behalf of all of slashdot, i would like to say

[circletimessquare]

(drool!)

$30 per gb, ouch

[dnamaners]

Its fast, but not as fast as I would have hopped with parallel access. They better get the speed up or the cost down to hit it big. Right now I'd take either direction, as they both have decent applications. Good progress though, time will tell.

Re:$30 per gb, ouch

[Anonymous Coward]

Are you friggin kidding me? At a sustained write of 600 MB/s (the article does say megabytes/s), that means i can rewrite every single bit on my 1TB raid in 2 seconds and read it back in about 1.5 seconds. Also, since this is flash, that means that we're looking at latency sub 1ms, which means you could have even the biggest apps such as Photoshop or Eclipse launching instantly (i.e. no noticeable delay to seek to all the various dynamic libraries etc).

This thing is about 10 times faster at raw bandwidth (although at $30/GB, it's 70x times as expensive as a HD and half the price of RAM)

Re:$30 per gb, ouch

[Anonymous Coward]

saw something faster at linux world

PCIe

[Gadgetfreak]

Seems like finally a newer reason to upgrade my hardware. I've never bothered with Vista, and I'm not big into gaming, so my P4/2.2GHz rig has been more than adequate for (surprisingly) over 5 years. Haven't needed a high end graphics card for a while, and only upgraded that for DVI output 3 yrs ago. When I built the thing in '02, I figured I'd get 3 yrs tops before it became a file server.

But it doesn't have a PCIe slot. Something like this would finally give me a reason to build an all new PC. Anyone else in a similar boat?

Re:PCIe

[scottyokim]

I'm sure many (myself included) are in the same boat. I built my P4 rig several years ago. You might want to consider the hot rod specs on ars technica when you make the plunge.

Re:PCIe

[Kankraka]

I'm in pretty much that exact boat. I have a P4 1.4ghz OC'd to 1.8ghz with stock cooling, no real difference in temperature, 768mb of ram, an ok video card; radeon 9550 256mb, the few games I do play on pc it runs amazingly. Basically the only short coming was the 40gb drive it used to have. I've since put just over 500gb worth of drive into it, and have no intentions of givin' that machine up. When I'm not doing my day to day activities on it, it's playing mp3's.

RAID!!

[mulvane]

Imagine a RAID6 of these. With a parallel interface between then for parity calculations..

Re:RAID!!

[kalirion]

Would it be worth a second mortgage?

Re:RAID!!

[nategoose]

and then X100, network them together, and make a Beowulf cluster of 'em

Re:RAID!!

[bigstrat2003]

I think that we'd have to come up with a different term than RAID, as this clearly is not an inexpensive disk. ;)

Re:RAID!!

[Anonymous Coward]

Ah, but they are Inedpendent. ;)

Re:RAID!!

[Emetophobe]

RAID can also mean Redundant Array of Independent Disks (see Wikipedia)... so they don't have to inexpensive ;)

Re:RAID!!

[Emetophobe]

Arg, I hate when I proof read my replies several times and I still miss something. That last bit should be "so they don't have to be inexpensive"

Re:RAID!!

[owlstead]

Yes, but I suppose they still have to be disks. Maybe we should name them Redundant Array of Independent Drives, what the hey...

Very important spec omitted

[Whuffo]

This could be a revolutionary product - or something that dies a quick death. The difference is "how much does it cost?"

Even a guess would be helpful. As a "proof of concept" it's pretty well useless - putting memory on a interface card isn't something that was invented recently. Heck, Microsoft was selling memory boards for the PC-XT way back when.

It was a good idea back then and it's still a good idea now - but what's kept it from becoming a widespread technology has always been the price. If they haven't found a way to make it affordable they haven't invented anything.

Re:Very important spec omitted

[LingNoi]

They priced it at $30 per gig but said the price may change.

Price

[Anonymous Coward]

Let's be generous and say they actually hit... 25 bucks a gig.

$25 x 640 = $16,000

There's your problem.

At 100 dollars apiece, you could snag a WHOLE lot of 500GB drives, plus cases, boards, PSUs, etc. to run an array MUCH more massive than this.

Nice trick. Try again in about 5 years.

Re:Price

[Teun]

You forget the power consumption and space factors.

Re:Price

[heinousjay]

Are you trying to tell me you don't have a room in your house set aside for nothing but storage devices, and a...

Bah, I can't even finish this. Every damn technological advance has to be poopooed by geeks. It's so freaking annoying I want to cry.

Re:Price

[stonecypher]

And when you look at the power bill, you'll make all that money back in about four months. What's your point? (By the way, this drive will still be faster.)

Re:Price

[Teun]

I think I heard a whoosh
b.t.w, who were you replying to?

Re:Price

[canuck57]

$25 x 640 = $16,000
There's your problem.

For an anonymous post you hit the nail on the head. And HD like Seagate are not sitting on their butts.

When Sandisk (or others) can come out with a 2TB billion times rewritable MTBF replacement AT THE SAME PRICE for my soon to be 4TB Seagate drive I will listen. I am not saying this tech is not inevitable, I just think it is 12-30 years out for mainstream. And there is always a possibility some new tech will plant them both in the ground.

$30 bucks a gig

[niola]

According to the article, they are looking at pricing to be 30 dollars a gig. That is pretty pricey.

That means their low-end 80GB drive will be around $2400+ or so US dollars depending on tax, shipping, retail prices etc.

Re:$30 bucks a gig

[PoconoPCDoctor]

Mod me down for sour grapes - I am fairly certain my story, which is still listed as "pending" will be rejected!

007-09-28 15:05:44 Would you buy a $2,400.00 80 gig Flash Hard Drive? (Hardware,Data Storage) (pending)

Oh the pain of rejection!

B-)

At least I got to shake the hand of the next POTUS!

Peace out!

Re:$30 bucks a gig

[jandrese]

Don't worry the weekend is coming up, prime time for dups!

Southbridge

[eepok]

Wow... adding 800gb/s to the workload of a southbridge would be quite a jump in required power, no? If they want such a possibility, would we have to accept another mboard form factor? Make room for yet another heatsink. O.o

Re:Southbridge

[wirelessbuzzers]

Wow... adding 800gb/s to the workload of a southbridge would be quite a jump in required power, no? PCI-e is usually on the north bridge for exactly that reason.

Re:Southbridge

[VoidWraith]

If it requires a lot of power, it'll probably do like graphics cards do, and have an additional power connector. They've even gone so far as to make a PCI-e-specific power connector (it's absolutely useless, as it provides the same power connections as berg and molex do).

Sounds like a nice place to put a Page File!

[The Assistant]

I would think this could help for those people who like to keep their computers multitasking! Swapping data in/out of memory would go a lot better if you send it to a faster medium, Si?

Re:Sounds like a nice place to put a Page File!

[brunascle]

a ramdisk would be much cheaper, and (if you could find a PCI-e one) should have more bandwidth too.

of course, if you're using it for a page file, when you plug it in you gotta be thinking "why dont i just use this ram as ram?"

Re:Sounds like a nice place to put a Page File!

[The Assistant]

Another thing to use this type of memory for is when the system hibernates. When system wakes up again, it should be nice and quick. Wouldn't it be nice if someone could create a utility that would automatically put the system in hibernate mode when certain conditions are met? I'm sure the local Electric Company wouldn't be thrilled, unless they're fighting with brown outs/black outs, but maybe you could justify the cost with savings in your electric bill.

Re:Sounds like a nice place to put a Page File!

[HarvardAce]

I'm sure the local Electric Company wouldn't be thrilled, unless they're fighting with brown outs/black outs, but maybe you could justify the cost with savings in your electric bill. I think you'll find that a lot of electric companies would much rather you lower your energy usage. After a certain amount of demand, the electric companies actually will start losing money -- either because of infrastructure upgrades or needing to exceed their agreements with other providers. A great example of this is in NYC, where ConEd is actually giving $50-100 if people replace their old air conditioners with energy-star approved units.

Gb or GB???

[saleenS281]

You would think being a "geek" new site, one could at least get their GB's and Gb's correct. If the drive is running at 800Mb a second, that's hardly what I would call *impressive* or *extremely fast*. That's not even as fast as most 10k rpm scsi or sata drives.

I'm also wondering how one would make the jump from 800Mb to 1GB... that would be quite the feat. I'm guessing that the B's are screwed up somewhere... it's just sad that something so glaringly wrong can be posted to a site like this...

Re:Gb or GB???

[Anonymous Coward]

RTF Article

> Flynn said the card has 160 parallel pipelines that can read data at 800 megabytes per second and write at 600 MB/sec.

this is apr 10 times faster than 10k SATA
and 7 times than 15k scsi

most importantly - seek time=0

> Flynn set the benchmark for the worst case scenario by using small 4K blocks and then streaming eight simultaneous 1 GB reads and writes. In that test, the ioDrive clocked in at 100,000 operations per second. "That would have just thrashed a regular hard drive," said Flynn.

Re:Gb or GB???

[Anonymous Coward]

WRONG. Show me a single non-RAID drive that can do 800 megabit sustained throughput please. www.storagereview.com educate yourself.

RAID Raptors can get about 100MB but not a single drive, not even close. Then you have the seek time advantage of this flash drive...

Re:Gb or GB???

[Anonymous Coward]

Even after the conversion it really isn't that impressive speed wise.

Mtron already has a drive w/ similar speed specs on the market, albeit their largest right now is 32 GB.

Re:Gb or GB???

[Anonymous Coward]

fix the fing summary

Re:Gb or GB???

[stonecypher]

You would think being a "geek" new site, one could at least get their GB's and Gb's correct.

What are you, new here?

If the drive is running at 800Mb a second, that's hardly what I would call *impressive* or *extremely fast*. That's not even as fast as most 10k rpm scsi or sata drives.

You don't know what you're talking about. Depending on fragmentation, a Maxtor 15krpm SCSI drive typically gets 42-68 megabytes per second throughput, which would indicate an order of magnitude improvement. Amusingly, the drive is running 160 pipelines, EACH capable of 800 meg per second sustained. Getting hundreds of gigabytes of throughput is three orders of magnitude performance increase. For someone getting up on the soapbox about the case of units, you sure don't seem to check your own research much.

Of course, the *real* issue here is that high load drives undergo extreme seek fragmentation. As the article points out, the major issue to corporate customers is operations per second, which is marketer for "holy crap there's no seek time, random access is just as fast as linear access." If you're thinking about this in terms of dropping the read time of a single item, you're missing the point. The issue here is that that drive can sustain that throughput no matter what random thing your thousand active users want this millisecond.

The performance impact of that kind of difference is borderline unmeasurable, it's so huge.

Re:Gb or GB???

[Courageous]

Well, somebody upstream made a units error. The capability is 800Mbytes/s with 100K IOPS.

According to their specs, that's not "per stream," but total. No "EACH," as you put it, in all caps.

BTW, that would be impossible; 800mb x 160 is 8 times greater than the bw of the PCI-4 bus this card runs on.

C//

Re:Gb or GB???

[stonecypher]

Ok, so it's only one order of magnitude faster, with no seek time. The difference is still enormous, and it's still not in the favor of 10k drives like you suggested. As far as drives saturating busses, it wouldn't be the first time that drive bandwidth had outpaced system bandwidth (if you were around for the switch to vesa local, you know what I'm talking about; if you're scratching your head trying to figure out what video standards have to do with it, nevermind.)

Re:Gb or GB???

[Courageous]

The difference is still enormous, and it's still not in the favor of 10k drives like you suggested.

Didn't suggest that, no. You're confusing me for some other poster.

I think these are neat little units. If they merely MATCHED the read/write times of ordinary hard drives, they'd have some interesting applications for other reasons (random access, shock resistance). As it is, this specific type of unit comes with the integrated performance of a highly-striped RAID array of flash drives, so it's pretty cool. As it happens, I represent one of those use cases where large amounts of money seeks to solve specialized problems. Perhaps I'll get one of these babies to try out. :)

C//

Re:Flash Memory == Vanished Data

[Anonymous Coward]

If you were my C++ Professor, then F you... Ah now I feel better

Re:Flash Memory == Vanished Data

[Teun]

He who drops nearly $20,000 for such a unit will have investigated the drawbacks.
And decided they are outweighed by the advantages, no worries professor!

Re:Flash Memory == Vanished Data

[Anonymous Coward]

should its battery fall out

The fuck?

by egg troll (515396)

Oh, carry on.

Perfect external journal device

[clawsoon]

This would be the perfect external journal device for databases and filesystems like XFS. Transactions would be guaranteed and super-fast, and could be lazily shuffled out to slower, bigger disk.

Misleading benchmark

[Chris_Jefferson]

"'[Fusion io's CTO David Flynn] set the benchmark for the worst case scenario.."

By which he means, set up a completely unrealistic benchmark which shows his flash drive in the best possible light, and a traditional drive in the worst possible light.

I still want one of these, but that benchmark is nothing to be proud of.

Re:Misleading benchmark

[Joce640k]

400Mb/sec transfer rate with instant seek is "nothing to be proud of"?

I know people who would commit felonies to have one of those for their page file.

Re:Misleading benchmark

[gbjbaanb]

You could do better - buy more RAM, set up a ramdrive and put the swap file on that.

Or, of course, just buy more RAM.

Re:Misleading benchmark

[stonecypher]

"[Fusion io's CTO David Flynn] set the benchmark for the worst case scenario.."

By which he means, set up a completely unrealistic benchmark which shows his flash drive in the best possible light, and a traditional drive in the worst possible light.

No, the scenario he set up is a classic worst case scenario for drives, one which is well known to large disk usage corporations: near-random access by hundreds-or-thousands of concurrent users. This is what SANs are built to address, this is what parallel RAID is so good at, and yes, it so happens that this new drive is also good for it, because there's no seek time.

Just because they've successfully approached a real problem doesn't mean that describing the real problem is flimsy propoganda. Knee-jerk reactions don't happen because you're a knee.

Re:Misleading benchmark

[Michael Zappe]

No, really, that was just iozone. Plain and simple, just iozone.

Re:Misleading benchmark

[Electrum]

I know people who would commit felonies to have one of those for their page file.

Or their database log disk.

$30GB Expensive?

[Anonymous Coward]

I bought my first 1GB drive back in 1992. A SCSI Seagate 1.1GB for $900.00 for my SPARC Classic. I compare that with the 750GB Seagate SATA I just bought for around $200.

We need something bigger than 32 GB in laptops

[wile_e_wonka]

These things sound great--I'm just afraid of Dell and other computer companies deciding that consumers don't want flash hard drives because we aren't buying the ones they offer. Then they attempt no development in this regard. The thing is, though, people are just not willing (for the most part) to pay extra to get a small capacity hard drive. They need to offer these things with similar capacies to their conventional hard drives--64GB just isn't that unreasonable for these things.

Then again, perhaps the computer manufacturers have not yet been impressed with the reliability of higher capacity SSDs.

Anyone have any idea why it's hard to find a laptop that comes with a 64GB or larger SSD?

Re:We need something bigger than 32 GB in laptops

[cnettel]

When the cost of the secondary storage unit is larger than the cost of the rest of the machine, it's something of a novelty item.

Re:We need something bigger than 32 GB in laptops

[Anonymous Coward]

Because you could buy 20 mid-range laptops for the same price as one of these 640 GB drives?

Of course, there's plenty of room in the middle, but the point still stands. Laptop drives aren't so terribly fragile that they give the majority of people much trouble, and for longer battery life you're better off getting extra batteries for the price rather than buying a larger SSD drive.

In short, it's simply not economical. And if it's not economical to buy, it's not economical to stock, either. If you really want it, I'm sure you can find a third party drive (if they exist) that you can use to replace a standard 2.5" drive.

Re:We need something bigger than 32 GB in laptops

[Cantinflas]

The reason you don't see a lot of laptop SSDs with more than 32GB is because it is still not trivial to pack that many flash chips into a 2.5" laptop disk form-factor (standard 2.5" drive in X and Y dimensions, 9.5mm in Z dimension).

When the next increment in flash capacities hits, you'll see a lot more 64GB drives and one or two vendors who can do 128GB.

Re:We need something bigger than 32 GB in laptops

[Anonymous Coward]

Anyone have any idea why it's hard to find a laptop that comes with a 64GB or larger SSD? Because that 64GB SSD consists of an array of flash chips. To understand what I'm saying, think of a stick of ram - which has a dozen ICs on it. The cost of those chips is still between $2 to $8 apiece.

It'd be like buying 64GB of RAM, it'll cost a lot more than $1 per GB... Probably upwards of $200 for a 64GB drive.

Re:We need something bigger than 32 GB in laptops

[wile_e_wonka]

$200 -- I could see a lot of people willing to pay an extra $150 for a 64 GB SSD.

I'm not talking about a 640 GB SSD -- sure, we'll get those someday. But right now we could seriously do 64 GB for an amount people are willing to pay (particularly if mass produced).

Airborne/ruggedized storage

[quintessentialk]

One obvious set of applicationa for large-capacity/high-speed flash storage are those which require not just greater speed but greater durability than moving media: planes, satelites, boats, military vehicles, etc. And these markets are possibly more able to pay for these benefits.

Re:Airborne/ruggedized storage

[dbIII]

planes, satelites, boats, military vehicles, etc

There have been solid state drives in those situations for years - gooogle will show you a few companies that make these. Now it's getting to the point where the rest of us can do more than drool and we can get the things on our own budgets.

Excellent

[iminplaya]

Now I can install even slower and more bloated software. Hard drives are the cassette tapes of the new millennium.

OK, this is very cool

[ThoreauHD]

Even if it's expensive now, it doesn't matter. Prices fall, but the technology solves that last problem for modern computers. The hard drive is no longer the bottleneck with this cards. They'll start selling in 2008 at 30/gig. So $2400 for 80GB's. That's the techie skim market they're looking at hitting, and not the server market. I'm sure at the middle of 2008 their prices will become more sane. 2 Mirrored SAS drives on your hp linux server may reach a grand on a bad day. Not a long way off really. I think this tech is very smart/helpful.

Wow, $19,200 for 640gb

[Acecoolco]

Damn expensive, $19,200 for 640GB......... I want it but cant afford it.. Josh

Re:Wow, $19,200 for 640gb

[glop]

There is quite are premium to pay here. Flash is at 8$/GB right now (16-17$ for 2GB etc.).
So the Flash for a 640GB SSD could be obtained for about 5 thousand dollars.
But integrating all that Flash is not that easy.

You could buy a bunch of USB flash drives and use software RAID to make a big drive. If you have plenty of USB ports on your PC, you might connect 6 flash drives of 8GB which would offer 48GB at 400$ or so.

Another idea I had was to buy CF flash and CF-IDE adapters and connect them to an Areca RAID card. But the power supplies would be a pain in the butt most CF-IDE adapters seem to require floppy power connectors, so you would have to buy N power cables and run them in your box which would be unpleasant with N above 6. But anyway most RAID cards only handle 8 drives, so you would need several of them.

All in all, this company seems to offer a pretty good deal since it is all well packaged, simple and not so easy to replicate for DIY guys.

Re:Wow, $19,200 for 640gb

[Anonymous Coward]

SUN do this with ZFS in this nice and cheesy film.

Ya, know...before WikiPedia...

[StressGuy]

We'd have said something like, "What's the Mean Time Between Failures?"

WTF?

Interesting to see

[ILongForDarkness]

Good throughput, but notice they used 10 processes to get the peak bandwidth. Assuming the ~80MB per process is the max you'd get with a single process, this doesn't beat high end disk for throughput. I have a new server with 15k SAS, the drives are "rated" at 160MB/s. They've actually tested at 140MB/s so they would smoke this device for single process.

What is interesting is that this device seems to favor situations where you can throw multiple simultaneous I/O's at it. This should be perfect for database, rendering and high performance computing. It will be a while I think before the standard user has enough going on, for them to have their nice shiny quad core or whatever chewing out a peice of this. By then, these puppies will probably be cheap enough, say 1k for 80GB. I'd pay that if I was doing a lot of video editing. You pump your current project into this device, churn and write back to spinning disk sequentially after offline.

Just as expensive as RAM

[mathx]

considering that large quad-socket boards have space for 8 Dimms per CPU we're looking at 128GB+ per machine and soon to increase. 640Gb isnt that much bigger. Since it's on the memory bus and not a PCI-* bus, its going to be faster than these drives, and it's more expensive right now. By the time they're at $30/GB ram will be alot less than that. The automatic persistence (without scheduling to back the memory to disk, like RAM would need) is the only advantage, so you're putting a high price on that ability - a database with constant accesses that need to go to permanent non-loss storage with a power out.

If you really have such an environment, I would think that fixing your HA setup would be a priority first - duplicating your servers so they can take eachother's jobs over and providing redundant power. I dont even want to know how many xactions/second a properly memory-stored database can do (once you get rid of the filesystem and driver layers, which this thing would require), Im sure many many more. While disk wont take as many xactions/sec, you can always back dirty ram to disk in huge chunks (1 meg blocks or more) to avoid having to need 100K IOops.

I just dont see any advantage properly tuning your server and process environment couldnt achieve with commodity unspecialized cheap easy to replace parts and a few brains.

Re:Just as expensive as RAM

[Joe The Dragon]

And FB-DIMMS are a $100 a gig

Re:Just as expensive as RAM

[Anonymous Coward]

Ahh another brilliant comment from the rewrite things from scratch brigade. In the real world time does equal money. The question is, will this product solve a problem (IO contention in a large normalized db for example)cheaper then re-writing a whole wack of legacy code?

Look up data warehousing and you'll find plenty of examples of ugly work arounds like report specific look up tables that need to be pulled from the main db.

Right or wrong most buisnesses are looking for more features not large rewrites of their code base. The IT departments that work on them tend to like to keep things simple as well. Exporting data all over the place just makes life harder. Data replication is a really spiffy place to have subtle problems that may not get noticed for ages.

In fact look at the intel 386 ABI, if the world works the way you seem to think it aught to we'd all be running Itaniums right now or whatever.

I won't argue that smart design is the best, but even smart designs age and some poor sod is going to have to deal with it. So calling out just be smarter is the worst sort of intellectual snobbery and blindness to practicality.

HardCard!

[jeffy210]

It's the return of the HardCard!!!! I remember having one of these with my old PC/XT. It was a 20 MB HardCard that fit into an ISA slot. The first ever hard drive i had running Windows 3.0 with DOS 3.3 on it.

Startup bloat

[IdahoEv]

Don't worry. I'm sure they'll find a way to bloat the startup to make it slow again.

As soon as resources are available, developers find a way to use them to just within the patience threshhold of the user.

Re:And another question. MTBF - HA HA

[Anonymous Coward]

Read the article and watch the video. They talk about that.

You don't actually believe that number anyway do you? MTBF doesn't mean anything. If you read the google papers on hard drives you would discover that the numbers posted don't mean anything.

Don't forget to read the information about CERN and it's data errors.

And let's not forget...

[sxeraverx]

And let's not forget, it's cheap!

Oh wait...

How much???

[drxenos]

640GB? Why don't they translate that into something more usable, like Porn/Sec.

massively fast - yada yada

[Anonymous Coward]

But how many read/write can I perform on that card before the flash goes dead ?
Also why is the card not PCI but PCI-X or better yet - ethernet base ?

last but not least - when can we get one of those card under $39.95 at Frys?

Re:massively fast - yada yada

[WarlockD]

PCI-X? REALLY? That guy is holding up as a PCIe x1, thats easy 500MB a second. Why the hell go to PCI-X when you can go to just 4x and get 2x the speed of a PCI-X 133mhz.

Hell, do you even SEE servers with PCI-X slots? I know Dell doesn't sell them new unless the customer asks for it. Even then its using a converter chip from PCIe to PCI-X.

But to be honest, I really don't see anyone using this thing on a database. The useal answers are read/writes, price and size. What people don't know is that many storage systems are self monitoring beasts. How good, for the enterprise, is this software going to be? Will it use a kernel driver? Does it offer a bios emulation mode to even see it under dos? I am all for database performance, but I don't see it any better than a RamSan. Heh, I just noticed they offer an all flash product. Funny that.

Talking to the company at demo

[shdowhawk]

Having talked to people at demo, what it pretty much came down to is this... is this a product we should be excited about? Definitely ... is it something that will do well right away? Not at all. The price has to drop before this becomes a really valid and useful tool for the GENERAL PUBLIC / Company. But there are a lot of companies out there willing to pay too much money to get these. Hopefully these big companies buy these up and fund this project as QUICKLY as possible. 7 of these side by side at 320/640 gigs a piece is a SCARY/powerful server.

That being said, a few of the guys there said that they pretty much expect these (at the beginning) to do the best sales for companies that are looking to get really really fast database servers going. NOT for scsi san replacements (it's silly to spend $100,000 for something you could get for 10,000 hard drive space wise). Eventually as the price drops... i know of a handful of people who would EASILY pay 1000$ to get one of these on a gaming rig even if it was only 100 gigs. But that right there is already 1/3rd of the price of what it currently is. (assuming it's around 30$ a gig).

Another thing to keep in mind that came up in the conversations... since these are tiny, think about the cost per server rack... and think of the cost per electricity to run. If you take those into consideration, these are actually less expensive that most people would think! A massive rack of hard drives could cost a lot of money in a co-location ... and a lot of electricity to run it all... But then again, we're talking about savings on servers, not general in home use.

When this gets to about 1/3rd of it's current price, that's when you will see these things become TRUELY mainstream both to the average company and home users (be it rich ones who need the latest and greatest).

Fusioni-io -- Link to their site.

very cool,

[Nex6]

this is very cool. has harddrives get faster, and larger. and CPUs get better...

Clearly you don't use windows

[Overzeetop]

I haven't quite figured it out, but my HD runs just about continuously for about 2.5 minutes after I log in. I even get a message that I don't have my firewall turned on because there is such a lag in the startup. I have yet to figure out what is taking so fucking long, since a restore from hibernation only takes 30 seconds or so (2GB of RAM), so it's not like it's having troubles loading everything it needs into RAM. And it's not caching a bunch of shit to the paging file, because I don't have one (with 2GB of RAM, I don't need one, even with Firefox running all morning).

I'm about 80% of the way through a fresh install on a new (bigger) HD, and it takes about 10 seconds. I wish I had a way to see what thread is accessing the HD, like I get with the processor list in the task manager.

Re:Clearly you don't use windows

[benwb]

Go to administrative tools -> Performance, switch to the view report toolbar item, click the + button. Performance Object "Process", counter "IO Data Bytes/Sec", click the "All Instances" radio, and then click add.

Re:Clearly you don't use windows

[fbjon]

Turn on the I/O read and write bytes columns in task manager. You'll see the numbers for a process go up when it reads or writes to the hd.

Re:Clearly you don't use windows

[DMUTPeregrine]

Process explorer lets you view I/O operations. And DiskMon helps monitor disk activity. Both from Sysinternals, free.

I'd get one

[FranTaylor]

I mess with VMware images a lot for work. Suspending and unsuspending them makes my raid-0 setup work really hard, the amount of data being transferred is very large. On my laptop, VMware suspends feel like they're writing to floppy disk. This drive would make a tremendous difference. They should make them in laptop form factors, too; this makes laptops actually usable for some tasks.

Indeed, right on!

[Anonymous Coward]

I can't understand why some laptop manufacturer is not selling a laptop with:
    - solid state disks (Let's face it, 60 Gb is plenty for normal Linux system with all your data)
    - ultra low power consumption processor (Doesn't need to be Intel-compatible. To hell with that! As long as it runs Linux and uses very little power.)
    - batteries which provide 16 h or more battery life
    - OK screen resolution

And yes I want to run Linux or some system which is flexible enough to work with this. Windows will never work outside the Intel processor hegemony.

I'm talking about a small, portable device, yet powerful enough to read PDF files, do net surfing, write text, read mail, program in Python etc. and whatever one does in a normal laptop. Maybe it could not play video so fast, but who cares, really... When I'm travelling I'd like to work instead of just sit still (even though that's also fun).

Re:Indeed, right on!

[Anonymous Coward]

> Windows will never work outside the Intel processor hegemony.

Except when it does, such as Windows Mobile.

But other than that, I am completely with you. Linux can run, even with X, on such lighter hardware, that I would really enjoy a ARM or MIPs based tablet running Linux just for browsing, email, media, ebooks, etc. Having a core 2 dual and 2 GB of memory is way overkill for most of what I want to do outside of work.

Could just spend $400 on a Give One / Get One OLPC laptop I suppose....

Swap on ramdrive...

[Joce640k]

"set up a ramdrive and put the swap file on that"

I'm not sure I see the logic there...

Re:Swap on ramdrive...

[jgoemat]

Well, Windows requires a lot of swap space. Let's say I have 1 gig of memory... Microsoft recommends your page file be a minimum of 1.5 times your ram size and maximum 3 times. So my swap file is 1.5gb to 3gb in size. Now let's say I purchase 3 more gig of memory, now I have to have a swap file that is 4.5gb to 9gb in size. The problem is also that Windows and it's applications are memory-greedy. So while the max I could fit in virtual memory was 4 gig before (1 gig ram and 3 gig swap), now I can have up to 13 gig in virtual memory, and WINDOWS WILL USE some of it. I would rather be able to disable the swap file altogether and just use ram. There are reasons for that, but in actuality I think my system would be faster without the swap file.

I don't think putting your swap space on a ramdisk is anything to worry about, the way virtual memory works doesn't allow for that. Otherwise you could be paging out the ramdisk to the swap file on the ramdisk. When a page is swapped out, I don't think ram can be accessed until the operation is complete, so there would be no way for the ramdisk driver to access the ram.

Re:Swap on ramdrive...

[Anonymous Coward]

The logic is that windows can't function without a swap file.

Re:Swap on ramdrive...

[fotbr]

Your flaw is thinking you HAVE to have swap space, much less "4.5gb to 9gb". Those guidelines were relevant when XP was fairly new, programs were using more RAM, but new PCs were being sold with an average of 128 MB RAM.

Now, they're not nearly as relevant, and should probably be revised (if they haven't already).

Re:Swap on ramdrive...

[gbjbaanb]

Your flaw is thinking you HAVE to have swap space I was going to agree with you but then I realised we have Vista and C# apps now :)

Re:Swap on ramdrive...

[fotbr]

Touche

Pretty pricey

[Joce640k]

I don't think their target market is home PCs.

The reduction in power/space/air conditioning alone will justify this price to a lot of datacenters.

And if you can throw in a performance boost as well ... they won't think twice about $30/gig.

Worst Power Consumers

[fast turtle]

The biggest power consumer in my Laptop (Compaq with Turion ML37/1GB 80GB 5400) is the backlight, which is listed as 40 watts of the total 60 watts output by the power brick. Now seeing at how you need a few watts reserve output in order to charge the battery, I'd say total system consumption is a meager 55 watts with the bulk of it going to the damn backlight.

Now if I could get the screen that the OLPC uses (readable in direct sunlight) I'd be quite happy as most of what I do involves simply reading/writing efforts (spreadsheets, email, word) and this would improve the battery life of my laptop to at least 5 hours instead of the 2.1 I get now.

Re:Worst Power Consumers

[Anonymous Coward]

That sounds extremely high - even with the backlight on full, the T61 I recently played with stayed well under 20W. In power saving mode, I managed to get it to 8.5W.

Re:Worst Power Consumers

[dgatwood]

That sounds like a dead short across the backlight tube or a very poor DC-DC board design or something. Typical consumption of a modern laptop panel should be high single digit or maybe very low double digit watts even at full brightness.

Expect it to drop as soon as they get a competitor

[Joce640k]

At the moment they're the only game in town so they can charge what they like (actually I don't think their prices are too unreasonable if you can look beyond raw flash prices and see the total benefits).

As soon as there's competition the prices will plummet.

So you're the source of all that spam...

[crovira]

How fast can you flood a TB switch? :-)

Did anyone notice

[colourmyeyes]

That these are initially Linux-only?

From TFA:

Linux drivers will be included and Flynn said Windows Server, XP and Vista drivers will be available three months after that.

Re:Did anyone notice

[Michael Zappe]

Glad you did. It's much easier to develop for. :-)

Yes but....

[Kildjean]

They are going to sell it at $30/per GB. Now lets do the math: 649GB * $30 = $19470....

With that amount of money I buy a Mac Pro with 8 Cores and a 1Tb Raid or a 1TB San.... I think Solid State has to grow cheaper before we consumers can jump the gun at it... but, like hard drives back when they made the jump to GB, it will be awesome to see SolidState HD in systems, better then the clunky magnetic disks we currently use...

Even for a corporation telling them a 649Gb Solution is going to cost them $20K they will flip you for it.

When in use...

[phorm]

They do when in use, but if you were just using your PC to - say - play music (not at all uncommon), then your video can be blank (or at least not using much power from the card), and likewise if you have a power-saving-friendly CPU those can be clocked down as well. Have power-saving also turn off the LCD backlight (is there a manual way to force this in 'nix through ACPI, btw?) and the hard drive might be the bigger drain.

Of course, how often one plays Mp3's on batteries is up to debate, but thinking beyond computers (and hopefully beyond the PCI-e interface), having high-capacity, high-reliability storage in a solid-state form for PC's will drive the overall price of the tech down, and hopefully make it cheaper for non-PC devices too: music servers, NAS drives, TIVO, flash-based media players, and/or other such things.

I wonder how it does for generating and handling heat... since that's always been one of the bigger hard-drive killers that I've seen with high-use systems. I have a buddy who, despite having massive cooling, regularly replaces drives in the RAID array of his high-usage servers because the combination of heat + mechanical wear kills the drives. Less mechanical parts is good in my book.

Re:When in use...

[Anonymous Coward]

Have power-saving also turn off the LCD backlight (is there a manual way to force this in 'nix through ACPI, btw?) Try: xset dpms force off

Why the product rocks....

[tempest69]

This device is all about the IO/second.. a 12 TB SAN cant come close..

If your looking to run a blast/darwin query on 50k files to find the closest match to an unknown dna sequence Either you need to recode a bunch of software to use sql, or you snag a piece of hardware that gives database level performance. 80gigs at $2400.00 is a bloody bargain.

The device is also 10x faster in bandwidth than a normal drive which is comparable to a san, but not such a power hog.

So really the tradeoff rocks for small files. It doesn't have a controller interface latency so its really quick. It should mask a good chunk of hard drive based lag.

Storm

FAQ

[ZapmanFBM]

For a lot of these questions, you can look at the FAQ here

No, they're not

[SanityInAnarchy]

Not likely, with that direct connection between them, right?

Well, maybe...

Because Windows needs a swap file.

[SanityInAnarchy]

If we were talking a sane OS, yeah, just disable swap and it'll work. But after a fairly long discussion with a guy who sells the stuff, and plenty of name-calling by him, I finally got him to admit that it was really a limitation of Windows, that if you actually disable the pagefile, certain things don't work.

Of course, I've lost the thread, and it could be entirely wrong...

Ok...

[crhylove]

And how fast exactly CAN you hop? I can hop pretty fast, but I've been practicing since I was a kid. Does height or distance matter in the hopping equation?

Re:Ok...

[Samizdata]

Neither, but you do have to hop in parallel... And remember, the Mean Time Between Falls better be pretty high, too.

$30 per Gig?

[Paracelcus]

Did I hear that right?
I guess I'll have to put off that operation, and food for the next ten years!
I guess that I can make the back of the Hyundai comfortable, I'll need a generator to run my stuff.
Well,, I can dream.

Sibling post has part of it...

[SanityInAnarchy]

If it gets fast enough, you may just want to use it as your main system memory, rather than swap. Then, suspend-to-RAM = hibernate, and your persistant storage could be something like tmpfs.

Actually, the best solution is a system with Orthogonal Persistence -- that is, remove the distinction between disk and RAM.

In an ideal situation, this means that no program ever actually closes. The closest it gets is the entire system being suspended to save power. All of the reasons this might not make sense to you are based on the way software works now, not on any limitations inherent in hardware or in the concept of software.

It's hard to explain abstractly, but maybe try an example: Google Documents. While it's not instantaneous, because it has to go over the Internet, Google Docs does autosave every document, along with a change history. You can still revert, not to the saved version, but to any version since it was created. If you want to play around with it without saving, you simply copy the document and work on the copy.

Or, consider the concept of a savegame -- for the most part, rather than explicitly saving your game and closing the program, you'd simply pause it, minimize it somehow, and move on to whatever else you were doing. After awhile of being paused, the game might automatically release the audio/video hardware, so it won't be using any resources except RAM. And when that RAM is needed, it'll get swapped out.

Does it have any advantages over a traditional savegame? Well, it's easier to program, but imagine the whole game is coded like that -- rather than having to "load" a level off the hard disk, you simply access what you need. No more "loading" screens.

So yes, probably the first thing that will happen is, someone puts a Windows swapfile on it. And probably, it will be a LONG time until legacy concepts of filesystems and virtual memory go away. But if you get something performing even close to RAM speeds, with persistant storage, you can go a LOT farther than simply emulating a spinning disk, and you could do a lot better than trying to store your swap on an emulation of a spinning disk that runs on something resembling RAM.

Mod parent up

[Anonymous Coward]

Agreed. Does he really need to be a snob about asking about mean time between failures? What a waste of an acronym and link...

OH HOW IVE MISSED YOU

[Anonymous Coward]

LOVE YOUR WORK ESP ON USENET

Sandisk Cruzers

[rapidweather]

I've been looking forward to these fast solid state drives, wondering when they would be ready for market.
I run my Rapidweather Remaster of Knoppix Linux (See Screenshots, below) on a little 2 GB SanDisk Cruzer drive, it is a little slower booting up than an ordinary hard drive, but so portable! Just plug it in, and boot up, and all your files are there. I have 4 partitions on it, even a swap. When I run Debian 4.0 from the hard drive, that swap gets picked up too. Often, Office Depot has these drives for about $25.00.
I understand that Dell is offering some solid state hard drives as an option now, but they are very expensive.
They did say those drives are faster, which I just gotta see, since the Cruzer is slower. Could be that I use a PCI USB card, rather than a "real" USB interface directly into the motherboard. I'm using an old Gateway 2000 PII, with 512K cache, so it's able to run both OS's just fine, even Rapidweather Remaster from the USB drive. To get the SanDisk running, it is necessary to have these files on the hard drive, in a MSDOS or Windows 95/98 partition:
http://www.rapidweather.com/download
(free download)
When there, pick the usb tarball. Everything necessary to boot from a usb drive running Rapidweather Remaster is there, and a detailed readme is included. Look at that to see all the details.
The older PC's can't boot directly from a USB drive, so the files are necessary to provide a menu, and get the SanDisk running. Any old small hard drive will do, as long as you can install MSDOS on it.

800MB/s and 600MB/s NOT b/s

[master811]

I hate to nitpick but the TFR uses MB/s not Mb/s, remember there is 8x DIFFERENCE between the two, otherwise 800Mb/s sounds a lot less impressive (its pretty much the same as current raid-0 (ish)

What about hypertransport?

[Anonymous Coward]

When is someone going to mount one in a cpu socket and use hypertransport? This would effectively eliminate any throughput limits of pci, correct? Would significant architecture changes need to be made to cpus to treat mass storage like ram?

Re:What about hypertransport?

[kybur]

True, a hypertransport connection is faster than a PCI-E connection. They probably chose PCI-E x16 because many-many more motherboards have that than have HTX (Hypertransport Expansion Connector) Slots. Although PCIE might be, say 20% slower than HTX, the speed far exceeds the performance of this flash storage device (which is only 800MB/s). The reduced latency of HTX might help you, but in the real world, one has to rely on empirical results. There are some papers about HTX vs PCIE online that can be found with a quick google search.

If Slashdot is the place for nerds...

[MBMarduk]

..then why the fuck do we always get these blurbs where people mix up "Gb" and "GB"??
THERE IS A BIG DIFFERENCE I.O.W. THEY ARE NOT THE SAME, FOR FUCK'S SAKE!
IMNSHO IF YOU DON'T KNOW THIS BASIC TIDBIT OF UNIT NOTATION YOU HAVE NO BUSINESS BEING ON SLASHDOT.
...so tell me, what the fuck's a "gb" gerbil brain? There's no such thing as "gb".
http://en.wikipedia.org/wiki/Giga
And don't tell me it's a limitation of /.'s headings like Wikipedia's, the word "PCIe" next to it
has three capital letters in it just fine.

Coyotos + The Hurd

[Verte]

Eros performed fantastic as a persistent system, and Coyotos is a big improvement. That way we will be starting with a secure, semi-persistent system and emulating *nix on top of it.

/troll

swap

[gumbi west]

No de facto--my desktop has been up for 32 days and has paged out a few million pages. Also, Apple and MS (if one of the other repliers is correct) have figured out ways to use swap to change performance. If there were a good reason not to do this, I'm sure the OS writers would figure out how to do it, but with HDs very reliable and cheap, why not optimize performance?