OCZ Couples SSD, Mechanical Storage On a PCIe Card

J. Dzhugashvili writes "We've seen some solid-state drives on PCI Express cards before, but OCZ's RevoDrive Hybrid may very well be the first solution to combine solid-state storage and a mechanical hard drive on a single PCI Express x4 card. Using Dataplex caching software from Nvelo, the RevoDrive Hybrid uses its solid-state component (a RAID 0 array of SandForce-based SSDs) as a cache for an onboard mechanical hard drive. The caching scheme is reportedly so effective that "a 5,400-RPM drive can be used without sacrificing much performance," according to The Tech Report's coverage. OCZ hasn't hashed out all of the details yet, but it expects the RevoDrive Hybrid to start at $350 this July. The base configuration should couple 60GB of solid-state storage with a 500GB mechanical drive."

Comments

Reminds me of hardcards

[yuhong]

Do anyone remember the old ISA hardcards?

Re:Reminds me of hardcards

[adolf]

Do anyone remember the old ISA hardcards?

Yes. They were made by Quantum, and later by some other manufacturers.

It kind of made sense at the time, since hard drives were non-trivial to install back then. I still remember performing a series of dark incantations in MS-DOS debug to initialize an MFM hard drive on an XT.

At the time, I thought it was pretty cool, getting my fingers dirty like that. But I think most folks would have preferred to die in a fire than get involved in their hardware to that extent.

And at the same time, I felt it was a lousy idea to integrate everything since it also increased the number of single points of failure in the storage system. (This so-far vapor offering from OCZ suffers the same problem.)

Another issue with the OCZ product: What problem does it actually solve which cannot also be solved by a good OS, a competent admin, an SSD, and a spinning disk?

I feel spoiled, these days, when I pull the side off of my desktop, plug in a new SATA drive, and it just works -- immediately, without even turning the box off first.

(I also remember 8-bit memory expansion cards populated with six dozen individual DIP RAM chips. I remember soldering pins onto SIMM memory to make them fit into my SIPP motherboard. And I remember caching hard drive controllers, stuffed with as much RAM as you could afford. And I remember hardware data compression cards of at least two general variations. I remember the And I remember when sound cards actually did something, and themselves had SIMM sockets. And I remember squeezing sixteen 30-pin SIMMs into four 72-pin sockets on a Socket 5 board.

I even remember an 8-bit ISA card, called the Copy II PC Option Board, which existed only to facilitate copying software on floppy. I even found a Gopher source for the reference just to show how full my beard is, and how long I've been in Mom's basement.

Now, get off of my lawn before I start lamenting about how under-appreciated a common 8-bit parallel port is.)

Re:Reminds me of hardcards

[Anonymous Coward]

No, you're alone here, old man.

Re:Reminds me of hardcards

[Anonymous Coward]

I have a 127MB hardcard downstairs. Still works.

Re:Reminds me of hardcards

[Anonymous Coward]

Sure, the hardcard's integration was a liability for desktop machines, but it had its benefits, too.

Had an old 286 luggable (a Sharp, IIRC) back in the day, and you only had one full-height drive bay, one internal ISA slot, and an expansion port for a dock (wired straight to the ISA bus, nothing fancy, nothing hotswappable). If you put in a HDD, you had to pick whether to put a half-height 5.25" or a 3.5" in the remaining half of the drive bay, so either you wouldn't be able to share 3.5" disks with (most of) my desktops, or you wouldn't be able to load games off a 5.25" at a friend's house. But if you put in both floppy drives, and then drop in a hardcard, and you wind up with more machines than most people had in their desktops, and portable to boot. (With a horrible, horribly slow 1-bit LCD screen.)

Finding the corresponding connector for the docking port, and wiring it to an ISA connector desoldered from an old motherboard to add a Sound Blaster, that's another story.

Re:Reminds me of hardcards

[Anonymous Coward]

No, No he's not.

I have mod points too, and would have modded him up but there's just no category for "epic nostalgia".

- Dan.

Re:Reminds me of hardcards

[geedubyoo]

I put a 32MB hard card in my Amstrad PC1512. It was the first hard disk of any type that I ever owned; I guess this was around 1988. Before that the machine just had twin 5.25" floppies.

Re:Reminds me of hardcards

[Anonymous Coward]

The reason hardcards were a good idea is that you often needed both a slot for the hard drive controller and a place to put the drive. If you only had a single drive, you could put the drive on the controller card and not use up the extra space.

In this case, though, the SATA port just isn't fast enough to separate the controller and the drive. The only way to max out the speed of the SSDs is to put them on the PCIe bus. A Revo can already max out a 6GB/s SATA 3.0 link, so what's the point in separating them? The problem is that fast SSDs are expensive, so you can't get much storage out of a super-expensive card. The solution is to put an HD on the same controller as the SSDs to enable both the "HD is a slower backing store for the SSD" and "SSD is a huge cache for the HD" scenarios.

dom

Re:Reminds me of hardcards

[hairyfeet]

But how long does it ACTUALLY last. I'm not talking those MTBF numbers that the manufacturers pull out their butt, but some cold hard "lets see how long it'll go" kind of in the trenches numbers.

Because from what I've seen while SSDs may be satanically fast they also seem to die pretty damned quick. Even Jeff Atwood at coding horror has posted you need to use a "Hot/Crazy scale" for SSD, as in how much money and data/downtime are you willing to risk for the crazy speeds.

Frankly between all the horror stories and watching my two gamer customer blow several hundred on drives that barely lasted them two years I've been telling my customers unless there is a specific reason for needing SSD, such as mobile devices that are gonna get slung around a lot, not to bother with the SSDs at this time. Frankly the HDD tech has gotten so good that often I'm pulling perfectly working drives as people upgrade for increased space long before they kill the drive, hell I have a drawer full from 20Gb on up to 200Gb, all working perfectly.

Also if the article I linked to and the gamers I worked for are any indication SSDs don't "fail gracefully" or give you plenty of advanced warning like HDDs do. With every HDD I've had fail short of being dropped there was plenty of time to get the data off as SMART gave warning long before the point of no return. With both of the gamers it was "flip the switch and its gone" no warning at all.

So does anybody out there have some real world experience and not just the MTBF numbers? I'm sure there are plenty of geeks here at /. that have pounded the hell out of SSD looking for boosted performance, how did they hold up? Are they still running? If not did you get warning before they died? Because if the drive can't be counted on to last at LEAST 3 years reliably in my book it isn't worth messing with or giving to my customers.

Re:Reminds me of hardcards

[TheRaven64]

With a hard disk, you didn't really need the second floppy disk. I had a 40MB hard disk in my PC1640 (128KB more RAM than you! Ha!), replacing the 20MB one that it shipped with. I rarely wanted to copy between two floppies, and when I did it was pretty easy to copy from one to the other. Oh, and I had to have an 8MB C: drive and a 32MB D: drive, because DOS (FAT-12) didn't support disks bigger than 32MB back then...

Re:Reminds me of hardcards

[Anonymous Coward]

Another issue with the OCZ product: What problem does it actually solve which cannot also be solved by a good OS, a competent admin, an SSD, and a spinning disk?

Um I think you answered your own question. It solves the need for a good OS, a competent admin, an SSD, and a spinning disk. Most technology isn't revolutionarily new. It's simply an abstraction of old more basic concepts and this is no different. However if I can do the same thing cheaper or simpler (from the people perspective) it's almost always "better" from an economics standpoint.

That doesn't of course make it the most efficient or technically elegant solution from an engineering standpoint so "GET OFF MY LAWN, I CAN DO THAT WITHOUT ALL YOUR FANCY GEEGAWS!!!" is a pretty normal response from those who value the elegant over the economic.

Re:Reminds me of hardcards

[wangerx]

Oh yea, that takes me back. I checked off all of the boxes as you went down the list. Controllers separate from the drives... replaced the controller and the drive was hosed. The whole debug to format was a joy. You were really cooking if you had RLL or ESDI. You were somehow "overclocking" your drive if it could take a 1:1 interlace (or non-interlaced in this case). You'd have to experiment with 2:1 or 3:1 to find out which worked best. You felt cheated if you couldn't get at least 2:1 working. It took hours between tests because you forked over big bucks for a 20 MB Seagate! Lest we not forget the two cables with those marvelous tab connectors.

Re:Reminds me of hardcards

[Ed Avis]

What problem does it actually solve which cannot also be solved by a good OS, a competent admin, an SSD, and a spinning disk?

None - but you'd be lucky if you manage to get all four of those prerequisites at the same time!

Re:Reminds me of hardcards

[Muad'Dave]

Me too. g=c000:5 and all that. I remember interleave optimization. That's 1987-ish stuff. Back when I also had a Perkin Elmer 7350A system running MicroXelos SYS V in my living room dialing into UUCP at 2400 baud. I even found an old UUCP entry from 1991 (thanks, internet!).

#N gnd0
#S Concurrent Computer Series 7350A, Running MicroXelos System V
#O Home of Mass Destruction and Woe (Ground Zero).
#C David Tiller, n2kau
#E gnd0!davet
#T +1 201 222 6753
#P 35 Sternberger Ave. Apt. H, West End, NJ 07740
#L 40 16 52 n / 73 59 00 w
#R Home of RF, Lasers, Computers, Dave-TV, suffering, pain and woe.
#U tsdiag
#W tsdiag!davet (David Tiller, n2kau) ; Mon Apr 22 16:00:00 GMT 1991
#
gnd0 tsdiag(DAILY), ka2qhd(DAILY)

Re:Reminds me of hardcards

[GaryOlson]

...an 8MB C: drive and a 32MB D: drive, because DOS (FAT-12) didn't support disks bigger than 32MB...

I must not be awake yet; the nightmares of technology past are still continuing to play in my head.

Re:Reminds me of hardcards

[drinkypoo]

Even at that time some vendors had software to help you do the install, mostly just to kick off a low-level format. My Xebec controller card, of course, did not. If I wanted to low-level format my 30MB full height Quantum disk I had to use debug like everyone else. But I lived right next to Seagate so I knew some vendors had solved this problem. If only they had solved the problem that forced me to whack my drives with a screwdriver we wouldn't have called them "Seizegate".

Re:Reminds me of hardcards

[Skater]

I still remember performing a series of dark incantations in MS-DOS debug to initialize an MFM hard drive on an XT.

I don't think I ever had an MFM drive, but I did have an RLL drive I had to do similar things with (a Seagate ST-238R, 32 megs!). I remember a little cautionary jingle: "Oh what a mess we weave when we amiss interleave!" I remember being happy when we could use 1:1 interleave because our 286 was fast enough to deal with it... LOL.

Re:Reminds me of hardcards

[SIGBUS]

Also if the article I linked to and the gamers I worked for are any indication SSDs don't "fail gracefully" or give you plenty of advanced warning like HDDs do. With every HDD I've had fail short of being dropped there was plenty of time to get the data off as SMART gave warning long before the point of no return. With both of the gamers it was "flip the switch and its gone" no warning at all.

I'd agree - don't use SSDs to store important data, or, if you feel that you must, be obsessive about backing it up. My own experience with SSDs have been mixed. I've used them in two OpenSolaris (actually now OpenIndiana) systems. They were identical OCZ Vertex 20GB drives, used for ZFS L2ARC (caching). One of them failed in operation without warning after about a year in service, and the other is still working after a similar amount of time, but it will undoubtedly fail at any moment. Since the fried SSD was a cache, there was no data loss, but it still gives me pause to think of using one as a main drive.

The bad drive is completely dead; it is not even detected during POST. It has been replaced by a 60GB Samsung. I'm through with OCZ, not just for the drive, but I had one of their 700W power supplies go "pop" on me last November (shortly after the warranty expired).

Re:Reminds me of hardcards

[hairyfeet]

See? That is EXACTLY what I'm talking about! To me the problem with SSDs isn't so much that they fail, although if they can't give you at least 3 years of reliable service I'd say the price is too high, but its the fact you get NO warning that something is about to go it is just "Pop, no drive" and that is it. With the gamers neither they nor myself could even get their game saves off the drives, they just worked one day and the next they didn't. No warning, no SMART heads up, nothing.

So I could see SSDs being used in a case like yours where it is simply a high speed cache and thus doesn't matter when it dies (although again if you can't even get 3 years? price too high for what you get) but to get ZERO warning that you are about to lose your data? To me that is simply unforgivable.

If someone wants the speed I'd recommend what I ended up doing with my gaming customers and just putting a couple of Velociraptors in RAID 0 with a large 5400 RPM for backups, but to not give the user ample time before dying hard without the drive even being abused to me says this tech simply isn't at a mature enough stage to make it usable by the masses. While there may be a few "speed at all costs!" types that are willing to let their drives be a money and data sink, most folks are not gonna be willing to shell out $$$ for a drive that might not even make a year. Frankly that is just nuts, especially when fast HDDs with plenty of cache are so cheap and RAID comes on just about every board made nowadays.

Re:Reminds me of hardcards

[ShakaUVM]

I recently upgraded to a SSD from a old, full RAID0, and a new HDD as well.

Old RAID0: Read 92MB/s (12ms)
Hitachi HDS7230: Read 115MB/s, 16ms
SSD (C300-CTFDDAC064MAG): 301.13 MB/s combined, 360 MB/s read, 70MB/s write, 33 us

It's a pretty cheap SSD, too, but too small for any long term use. I bought it as a stopgap until the new generation of SSDs comes on the market. After a lot of dickering around with my upgrade to Sandy Bridge, I finally bit the bullet and switched to a SSD for Win7 and HDD as a data drive (which is fast, but also sucks since the SSD is too small to do anything with it), while keeping a cloned copy of my system ready for any old apps I need.

With no apps installed on the SSD, or any data on it, and daily backups, I'm not overly concerned about it dying on me, though it'll obviously be a PITA.

Re:Reminds me of hardcards

[Henk Poley]

What problem does it actually solve which cannot also be solved by a good OS, a competent admin, an SSD, and a spinning disk?

The problem of not having a competent SSD, or good OS. Possibly also, lack of remaining HDD slots or SATA/power connectors.

Re:Reminds me of hardcards

[KnownIssues]

Another issue with the OCZ product: What problem does it actually solve which cannot also be solved by a good OS, a competent admin, an SSD, and a spinning disk?

The dependency on a good OS and a competent admin?

Re:Reminds me of hardcards

[asdf7890]

Another issue with the OCZ product: What problem does it actually solve which cannot also be solved by a good OS, a competent admin, an SSD, and a spinning disk?

It solves the "competent admin" bit somewhat, and removes the need to rearrange things when the data most commonly in use varies over time.

I'm surprised I've not seen talk of a Linux block device that merges two block devices (one an SSD and the other linked to larger spinning disks) and uses one as cache for the other. Such a software solution would have the disadvantage of having no non-volatile RAM to keep indexing structures safe through power fails so they would have to be synced to the cache block device very regularly (and in a journalled manner so these writes were safe from power-fail initiated corruption) adding a little extra inefficiency (unless these cards also have no faster NVR and are relying on the SSD to always hold the latest indexes of what is cached where on the SSD, of course), but should be quite practical to implement and certainly not beyond the abilities of many a kernel coder.

Re:Reminds me of hardcards

[asdf7890]

Or it could be done in the filesystem layer of course, like ZFS (not directly available under Linux ATM due to licensing issues, though there are ongoing efforts to port it in a license compatible manner) can.

Maybe the existence of ZFS is why it hasn't appeared at the block device level: the interested+capable parties are working on a ZFS port (or something similar) instead.

Re:Reminds me of hardcards

[Nefarious Wheel]

Well, I've got an OCZ SSD (non-hybrid) booting from a PCI-E slot and I'm rapt with it. Deadly fast. And you can disable swapping or move the swapfile to another drive if you can't add sufficient RAM (I'm running comfortably on 16gb, high-end gamer, no real point to a swap file) then you're limiting the overall write activity your SSD incurs. This is good. O/S and game content is largely read-only, so it's not going to write itself to death as many early SSD's tended to do.

Re:Reminds me of hardcards

[Nefarious Wheel]

The PCI-E OCZ I'm running does effective 750MB/s RW. From the bios boot it looks like it's an internal RAID stripe set of 55GB chunks. Clever. The fast R/W I believe is mostly due to its bypassing the SATA bottlenecks. There are times when I wonder if it isn't just RAM with an internal battery, though. Dunno, but the combination is lightning fast.

Re:Reminds me of hardcards

[Nefarious Wheel]

Youngster! TI ASR emulator dialed up to a Sigma 7. Get off my lawn!!

(Come to think of it, I think my watch has more RAM than the entire installed base of Xerox Data Systems machines.)

Re:Reminds me of hardcards

[Cimexus]

Just got my first SSD (Corsair Force F115) two weeks ago. Remind me about this thread in 3 years and I'll tell you if it's still alive ;) Given that the drive it is replacing was a mechanical HDD which only lasted 3.5 years, the way I see it, if it equals or beats that I'm happy with it (especially considering in 3 years time I'll be able to get a far larger and faster SSD for the same price).

Reliability is a concern for most computer users who are less than fastidious about their backups. People like my sister or parents who aren't very tech-savvy. For me though it's not too much of a biggie (apart from the irritation of having to drive to the store, RMA the drive and get a replacement). I used Clonezilla to make a full drive image of my fully installed and updated OS with the settings the way I like them. So it's really just: install new HDD (10 mins), restore image to it (another 10 mins), and reinstall bits of software on an ongoing basis as I require them.

Re:Reminds me of hardcards

[Cimexus]

YMMV I guess. Last hard drive that failed on me was a mechanical drive and it literally just died instantly. Was working fine one minute, then my PC freezes up for 20 seconds ... followed by a bluescreen. Reboot, and uhoh, no boot record. Data recovery software showed a bunch of bad sectors, conveniently located right in the middle of the master file table! It said it could recover files for me ... and according to it, it 'succeeded' at doing that - but all the files it extracted and wrote to the good medium were scrambled to hell. (i.e. MP3s that weren't seen as valid MP3s, executables that wouldn't run). Even a low level format/zero-fill of the drive would fail.

Yet, despite all this, smartmontools would continue to report everything was A-OK with the drive ;) No indication of failure at all.

The replacement drive I bought is an SSD, so we'll see how this one fares in the long term... :)

Re:Reminds me of hardcards

[Cimexus]

Not that you can ever generalise about any brand of SSD or HDD, but OCZ is fairly well known as making the less-reliable (but damn fast) drives. Halfway up the scale are Corsair, Crucial, Mushkin etc. who make drives that are a little slower on paper than the OCZ (but still stupidly fast, really) but are a fraction more reliable. At the top you have Intel who easily have the most reliable SSDs by far, but they'll cost you.

Re:Reminds me of hardcards

[Muad'Dave]

My first computer experience was in jr high - a dumb terminal dialed into an HP-3000, (or was it 2000)? Anyway, 110 baud, then 300 baud, both via acoustic coupler. I got to visit the Math/Science Center where the beast was housed, and used a teletype and paper tape to store my programs (and play adventure, where a ctrl-D sent in a message to another adventurer would turn off their teletype!).

I got a TRS-80 model I Level II when I was 15-16. Wow!!! The Perkin Elmer box was from my first job out of college. Am I still a young 'un?

Re:Reminds me of hardcards

[GameboyRMH]

You know the other day I got an Acer Aspire One that had a bad SATA controller. I installed Debian w/ LXDE to a plain ol' 8gb USB flash drive, and it's not fast but it isn't painfully slow to use. I've seen many off-the-shelf PCs that are slower overall. You can get 16GB or 32GB flash drives for cheap (and you can get "high performance" drives that are faster too), so if you need a shockproof computer on a tight budget, I'd say do that instead.

And if you're not afraid of hardware modding, you can move the flash drive inside the PC either by using the internal USB headers many laptops have, or by sacrificing an external port.

Re:Reminds me of hardcards

[GameboyRMH]

When I was building my gaming PC in early 2009 I put twin VelociRaptors in RAID0 in it just because the cost of SSDs per GB absolutely made no sense. At that time at least, SSDs were only an option of you needed just tens of gigabytes (you still need hundreds of gigs on a gaming machine) or if money was no object, all reliability issues aside. Same reason I went with a high-performance hard drive instead of an SSD for my laptop in early 2010, I figured I could settle for 60GB of space but it still didn't make sense. Although if I were building a laptop today I think I'd go with an SSD in the 80-160GB range - with frequent backups of course.

Re:Reminds me of hardcards

[hairyfeet]

Did you try Spinrite? I have had a few over the years that died on customers (usually they got a really cheapo drive like the Maxtors and then abused it by putting it in a lousy spot, or crowding their PC into a corner with lousy ventilation and slamming the hell out of it) and I found that while Spinrite couldn't save the drives themselves it would get it to go at least long enough to pull the data. the only one I found it couldn't was one where a girl had her cat knock her laptop off the desk and it suffered a head crash. But I have found Spinrite is one of those apps you won't hardly need, but when you do you'll be damned glad you have that CD in your toolbox.

But again a situation like that is pretty far from the norm when it comes to HDDs, as I have seen plenty of drives get worn out by years of use and abuse, especially those i built for construction sites, and as I said you got PLENTY of warning before the drives died. you'd get SMART errors, the classic "Windows has had a delayed write fail" error, etc. But the problem with SSDs is apparently the norm is to die hard with ZERO warning or chance to back up data!

Now I don't know if it is the switch from SLC to MLC, as I've heard that when a SLC SSD dies it will usually become read only thus allowing you to get your data, or whether they are trying to cram too many chips in too small a space, or controller issues, frankly I don't care. the fact that they can't even design it so the drive fails gracefully is simply inexcusable in my book. if you want to take the chance go right ahead, but I have customers working on projects that cost several hundred thousand and if they even lost a few hours worth of work it could really screw with a deadline. Even my gamers weren't too happy about it being 3 days for one and 6 days for the other since they ran a backup, so all that gaming was lost.

Call me weird, call me old fashioned, but when it comes to storage I expect at LEAST 3 years of reliability barring abuse. that isn't too much to ask for is it? If I can get several times that off my flash sticks I don't see why they can't at least give me a drive that lasts 3 years and gives me a heads up before it dies. Speed is nice, but unless you have the money to have a spare sitting in a drawer waiting for the first to die along with backups every 20 minutes to me the risk is too high compared to the price and rewards.

Re:Reminds me of hardcards

[Eil]

Another issue with the OCZ product: What problem does it actually solve which cannot also be solved by a good OS, a competent admin, an SSD, and a spinning disk?

That of not having a good OS or competent admin.

Re:Reminds me of hardcards

[Lennie]

http://bcache.evilpiepirate.org/ you mean ?

Re:Reminds me of hardcards

[StikyPad]

I'm buying an SSD this week, and it will have one purpose and one purpose only: To boot the OS and to run apps and games. My personal folder(s) are mapped to an actual magnetic hard drive with a redundant copy, along with an image of the boot drive (which will henceforth be the SSD) with the OS, drivers, and essential apps freshly installed. If it fails, I just restore from the image, and of course being a nerd, the SSD is just one of several bootable drives anyway.

Anyway, with TRIM now standard on SSDs, I expect we'll be seeing much longer average TBFs in real-world scenarios. Yes, some people's drives are DOA or die within the first few days of use, but that's the typical failure pattern for any electronic components. And now that prices are approaching $1.50/GB -- even as speeds are already saturating 6GB/s SATA connections -- it just makes sense to take the plunge. Even with a 2 year old system, there's nothing else in the $200 range that could even approach the performance gains to be had by an SSD. Sure, I could get 20x the storage for less than half the price, but I already have as many drives as I have SATA connectors, so I'm pretty much set for capacity. What I need is speed.

Still, I don't think I'll be buying this, or any other hybrid technology, any time soon. I have a better idea of what items I want to be on the SSD than any algorithm or usage pattern. Just because I don't fire up Photoshop every day doesn't mean I want to wait when I finally do. Likewise, just because I listen to the same songs over and over doesn't mean I want them taking up precious space on the SSD. If I felt like being exceptionally lazy, I'd rather just map a folder to a magnetic device.

Finally, Intel's enhancements to its Rapid Storage Technology (onboard RAID) in the new X79 chipset are reported to enable hybrid functionality across discrete devices, so there's good reason for anyone interested in such a device to hold off for a few months.

Re:Reminds me of hardcards

[asdf7890]

That is exactly what I mean.

I'd somehow managed to be blind to its existence. Not ready for production use yet it would seem, but something I'll certainly be keeping my eye on.

Re:Reminds me of hardcards

[yuhong]

To be more precise, the reason the debug worked was that they integrated the software to do the low-level format into the ROM. debug was just used to invoke the software.

Re:Reminds me of hardcards

[WuphonsReach]

But the problem with SSDs is apparently the norm is to die hard with ZERO warning or chance to back up data!

At the moment, I blame that on shoddy firmware design.

It's still a very young technology - and all those corporations that have jumped into the SSD market don't have much experience. Which means you have a lot of people writing firmware with the approach of "if it works, ship it".

Intel, I mostly trust to write good firmware. Some of the other folks, I don't.

(There's no technical reason why MLC shouldn't also go read-only at end-of-life.)

Re:Reminds me of hardcards

[drinkypoo]

To be more precise, the reason the debug worked was that they integrated the software to do the low-level format into the ROM. debug was just used to invoke the software.

Well, that's mostly true. Some vendors had an additional low-level format utility. IIRC you had to jump to a slightly weird address to invoke the utility on my MFM controller but it's been a long time.

Re:Reminds me of hardcards

[RockDoctor]

See your

Oh, and I had to have an 8MB C: drive and a 32MB D: drive, because DOS (FAT-12) didn't support disks bigger than 32MB back then...

and raise you a (darn, I've forgotten) ...

... still forgotten ...

Oh, got it. Compaq "Portable" - my first portable.

OK, Work's first portable, which I had to repair after the technical department fucked it up on the rig.

(And I've got a 40MB MFM hard drive on an RLL controller, giving 60MB. Top that [GRIN]!)

OCZ Fail.

[Anonymous Coward]

Too bad OCZ has already lost myself (and apparently a whole lot of others) as a future customer(s) due to their terrible Vertex2 drive that bricked after shortly after install.

I've been waiting for these

[IICV]

I've never understood why they make those 200 - 300 GB SSD drives, when ~30 GB of SSD cache will hold your operating system and your most frequently used applications. It's not like everything on your hard drive needs to always be immediately available at SSD speeds, and yet recently that's been the only option.

Caching isn't some mysterious arcane technology, why has it taken so long for them to make a hybrid drive like this?

Re:I've been waiting for these

[im_thatoneguy]

I hear rumblings that Windows 8 is going to finally solve this once and for all at the OS level. Designate a drive as a cache drive and it'll fill it up with frequently used files and the current cache. Then when you go into standby it can just dump all of your RAM to the SSD. Start back up and it reads off the SSD.

Re:I've been waiting for these

[mlts]

Windows Vista and newer had the ability to use a technology called ReadyDrive which can use the SSD as a cache, and the spinning platter as the HDD.

So far, there have been very few drives using this technology. It seems like it would be useful, although what would make SSD and conventional HDD pairings more useful would be a hard drive controller doing what most SANs do -- autotiering. Data that is read/written to all the time gets moved to the SSD while stuff that isn't used gets put on the platters.

Re:I've been waiting for these

[davester666]

Isn't this crap patented by Microsoft? I remember reading some stupid warnings saying you are only licensed to use it with Windows Vista or newer, or else you need to pay Microsoft additional money to use it.

Re:I've been waiting for these

[Sir_Sri]

Ya, it seems like a readyboost, but for SSD's would be the way to go.

My problem is that, while I have a pile of data, even the OS has a lot of data I don't want (language packs for example). Because regular disk space is so cheap that wasn't a problem. But caching would seem to solve, rather directly, the problem of having a lot of crap I usually don't want, for the relatively commonly used data I do want.

Even the jumplist in Windows 7 would a viable half assed solution to this problem.

Re:I've been waiting for these

[Anonymous Coward]

I've never understood why they make those 200 - 300 GB SSD drives, when ~30 GB of SSD cache will hold your operating system and your most frequently used applications.

You must not be a PC Gamer then.

Crysis 2: 7.62 GB installed game folder
Portal 2: 9.78 GB installed game folder

Now quick, how much space is left on my 30 GB Windows 7 drive after installing these two games? That's right.. not enough.

Re:I've been waiting for these

[IICV]

You're not using all ten gigabytes of Portal 2 every time you play the game. An intelligent caching scheme would see that you're reading a lot of data from a contiguous section of disk, and copy more stuff from the area into your SSD cache than you actually ask for.

And anyway, double the size of the SSD - 60 GB, like these guys are saying - and even that complaint goes away.

Re:I've been waiting for these

[Anonymous Coward]

Some high-performance servers require that amount of storage, but require more speed than any sort of RAIDed hard drives can provide. I believe AOL (might've been someone else) recently had to commission a server with several terabytes of SSD for an ultra-high-performance database. Expensive as fuck, but cool.

Re:I've been waiting for these

[mlts]

From what I recall, MS operating systems can use the feature the best. ReadyDrive didn't seem to go well, because it is "neither fish nor foul".

SANs do autotiering, using SSD for either a large persistant cache, or move data up and down the hierarchy as needed. Most HDDs would be too expensive with the drive controller intelligence to handle this. So, until recently this has been a dead feature.

Re:I've been waiting for these

[Dahamma]

I just built a new machine, did a clean install of Windows 7 64-bit with a few apps/drivers (just the ones that came with my motherboard, mouse/keyboard, BD drive, graphics card).

My (240GB SSD) HDD now has about 40GB used. I haven't even installed Office on it (I think that's 5GB+ alone these days!?), let alone a Linux instance under VMWare. Or a whole bunch of random apps/utils I used. Or a few games...

Modern software has grown (ridiculously) to fill modern storage.

Oh, and it hasn't taken quite as long to make a hybrid drive as you think - you can get a 500GB Seagate Momentus XT drive with 4GB SSD cache for $100. Same storage as the OCZ PCIe card, and given how good cache hit rates are in practice (ie. your point), probably 2/3 the performance at less than 1/3 the cost.

Oh, and the Intel Z68 chipset (on the motherboard I just bought) supports using any 20+GB SSD drive as a cache for your HDD. Haven't tried it, but it's supposed to be similar in performance to something like the Momentus...

Re:I've been waiting for these

[Datamonstar]

You don't have to store applications on the same partition you have Windows installed on. I currently have Windows XP install, page file, Program Files, and My Documents all on completely different partitions. As soon as I get around to caring enough, I'll get a SSD and sequester whatever blend of Windows is current at the time to that.

Re:I've been waiting for these

[kevinmenzel]

Frankly I could use about a 1-2TB SSD for all of the samples for my virtual instruments on my studio machine, though a 250GB drive would probably handle the most complex and most used instruments...

Re:I've been waiting for these

[Neil Boekend]

The Eve Online database server uses ramsan, becasue SSD's are too slow. They have 2Tb of network attached RAM.

Re:I've been waiting for these

[adolf]

You're not using all ten gigabytes of Portal 2 every time you play the game. An intelligent caching scheme would see that you're reading a lot of data from a contiguous section of disk, and copy more stuff from the area into your SSD cache than you actually ask for.

No, I'm not using all ten gigabytes of Portal 2 every time I play the game.

But an intelligent caching system won't see that I'm reading a lot of data from a contiguous section of disk, because the installation is likely not contiguous on disk in the first place. Without knowledge of the filesystem AND the application, "intelligent" sector-level caching is a waste for such applications.

Meanwhile the OS will likely, while I'm running through the single-player mode of that particular game, only request new data once per session: Even if I play a level over and over again, it's still going to be in the OS's RAM-based cache on any respectable machine that isn't otherwise burdened. The extraneous "intelligent" SSD caching system won't help a bit.

And even if I'm particularly good at Portal 2, and never replay a level, the data for the next level will still need loaded from a spinning disk...because until I do play them, unplayed levels will appear to an "intelligent" caching scheme as data that has no business being pre-cached because nothing has ever used it before (aside from the first time it was written out to disk at installation).

All it will help out with, consistently, with common use of Portal 2 is initial load time of the game itself...and only then if the game is used often enough that its data isn't flushed from the SSD cache in favor of more recently-used data.

Just to be clear: I'm not sure what point you're trying to argue, exactly, when talking about games and this particular sort of technology (which is not, by any means, a new concept). I can, however, see that you are wrong.

Re:I've been waiting for these

[im_thatoneguy]

I think ReadyDrive has failed mostly because it was left to the drive controller to handle the caching.

My understanding is that with Win8 they're moving the logic to the OS and divorcing the hardware from the equation freeing you to buy any old spinning medium and any old fast SSD/Raid to act as your cache.

I like this idea since I can 'upgrade' my existing drives to ReadyDrive by just buying a SSD and I can still have my multiple disks but just the one SSD between them.

Re:I've been waiting for these

[Anonymous Coward]

Well if you have an iPhone and want to back it up, you need at least iPhone memory sized space on your C drive (in my case 64GB). I'm not buying another HD just to backup my iPhone when I have 4TB of SAN.

Fuck iTunes.

Re:I've been waiting for these

[synthesizerpatel]

Seagate's offered a similar product that has the SSD built into the same form-factor that contains the mechanical for a while. There have even been 'caching' solutions that front-end mechanical NASs with memory or SSD drives. OCZ just paid for a little marketing to get their name at the front of the queue for a couple days.

Re:I've been waiting for these

[Anonymous Coward]

Just to be clear: I'm not sure what point you're trying to argue

Likewise.
What are you trying to say? That because of your high demands you will stick with slow mechanical drives?
Good for you I guess.

Re:I've been waiting for these

[adolf]

What are you trying to say? That because of your high demands you will stick with slow mechanical drives?

Since when is running a game a "high demand" for a storage system?

*facepalm*

Re:I've been waiting for these

[artaxerxes]

There is also marvell hyperduo on motherboards and HighPoint RocketHYBRID 1220 which allow you to choose your own SSD and HDD to combine aftermarket. Toms hardware had a good rundown on the pros and cons.

Re:I've been waiting for these

[Pieroxy]

I hear rumblings that Windows 8 is going to finally solve this once and for all at the OS level. Designate a drive as a cache drive and it'll fill it up with frequently used files and the current cache.

This is stupid. What about, my windows crashes and is beyond recovery. Now all my data is spread on a "cache" SSD drive and on the physical drive. How do I get that onto one drive again?

For all its pitfalls, OCZ's solution is ONE DRIVE, hence having the benefit of ONE DRIVE. You can take it out, bring it to a friend, put it in an USB enclosure, etc...

At the OS level, I guess it is fine for some purposes.

Re:I've been waiting for these

[yarnosh]

For one thing, good SSDs like OCZ and Intel put out are not cheap. I'm not sure you'd gain a whole lot by using one of those cheap SSDs as a cache.

Re:I've been waiting for these

[imsabbel]

You dont seem to grasp what "cache" means.

Re:I've been waiting for these

[imsabbel]

Sorry, but the cache of the XT is besically useless in most cases, as its FAAAAR to small.

The only thing it really would get 2/3rds of the performance of anything SSD is when you have a benchmark repeated for the 2nd run.

Re:I've been waiting for these

[Pieroxy]

So please enlighten me. Don't forget write-cache in your explanation.

Re:I've been waiting for these

[GigaplexNZ]

For all its pitfalls, OCZ's solution is ONE DRIVE, hence having the benefit of ONE DRIVE. You can take it out, bring it to a friend, put it in an USB enclosure, etc...

You can? I don't know of any PCIe x4 compatible USB enclosures...

Re:I've been waiting for these

[Pieroxy]

Uhhh, true enough. I guess they should get their act together and release SATA drives with this.

Re:I've been waiting for these

[deroby]

I beg to differ.
I've been using one since september last year and it beats the crap out of the standard HDD. In my team we're all having pretty much the same machine with the same software installed. Because I do a lot of database stuff and also like to have my music collection and some games to carry around I got fed up with the limited capacity of the standard drive and out of frustration bought my own 500Gb Momentus XT. The thing boots *much* faster than all the other machines around me. Outlook takes seconds to start, it takes about a minute for my neighbour. Same for Visual studio.

There's plenty of Youtube vids around that show the impact of the Momentus XT and I can only confirm them. more cache might have been better, but 4Gb sure does a great job !

Re:I've been waiting for these

[TheRaven64]

I very much doubt that it's patented by Microsoft, since Sun shipped this support in ZFS - and were sued by NetApp, claiming that they owned it (not sure what happened to that in the end). With ZFS (on Solaris or FreeBSD), you can designate a drive as a level 2 adaptive replacement cache (with L1 being in RAM). There was also some talk from Oracle about using hard drives as L3ARC and using tapes as persistent storage (just writing the entire change history out, so you can always roll back to any previous point), but I don't think that's finished yet.

Re:I've been waiting for these

[Anonymous Coward]

Considering they are talking about the readyboost equivalent in windows write caching isn't relevent to the discussion as that is for read caching and volatile memory only.

Re:I've been waiting for these

[xanadu-xtroot.com]

I just built a new machine, did a clean install of Windows 7 64-bit

That's you're problem right there. M.

Re:I've been waiting for these

[Jeff DeMaagd]

I think ReadyDrive has failed mostly because it was left to the drive controller to handle the caching.

I recall it being promoted, but I don't remember exactly what happened. Was there a particular problem with that method? Was it a cost problem? Or did it not work well? Drives have already been doing caching for as long as I remember, so I think it's understandable to assume that it would work.

It sounds like this OCZ device is the same way.

Re:I've been waiting for these

[drinkypoo]

Caching isn't some mysterious arcane technology, why has it taken so long for them to make a hybrid drive like this?

Piss on the magic PCIE drive, you can do the caching without any special hardware. You could even do it on Linux if someone would pick dm-cache back up. (Please?)

Re:I've been waiting for these

[GaryOlson]

Write-cache by definition does not include data in left cache. So, since nothing is left in the cache, all is write. No problem.

Re:I've been waiting for these

[Anonymous Coward]

Couldn't it be handled the same way a virtual memory read/write is handled? When Windows does a write it goes to the spinning platter drive. When Windows does a read it tries to read from the SSD. If that fails (hard fault), Windows reads from the platter drive and writes it to the SSD. The next time it reads that data it comes from the SSD. It could take advantage of the extra space by tracking access frequency and keeping the most heavily accessed data on the SSD (think Java style garbage collection, multi-generational partitioning?).

Write-cache acts at a lower level of abstraction to this.

Re:I've been waiting for these

[Pieroxy]

Well, one of the point of caching on an SSD-type memory has to be the fact that your cache is non-volatile. So when you power up your drive the next day, the cache is still there. Why not doing write-caching then?
It really is one of the advantage of SSD over pure RAM. The downside being that its a lot slower.

Re:I've been waiting for these

[LordLimecat]

Cache is not a place you move data. Cache is a place data is copied so that subsequent reads are faster; the data remains on the old drive. Presumably you could also use it as a write-cache as you mentioned, but that isnt necessary and I doubt they would go that route--I would assume that writes would be cached to the main drive's built in cache, and then flushed to disk (there really isnt a reason to have a 60GB drive for write cache when you have 32mb built into the HDD, under a normal desktop load....).

Writes are rarely ever a bottleneck on desktops under any workload I've heard of. Its mostly reads that slow the computer up, AFAICT.

Re:I've been waiting for these

[LordLimecat]

If theyre smart theyll have a SATA drive connected to the pcie board with a mounting bracket and a sata connector, and presumably you COULD remove the drive.

Re:I've been waiting for these

[GigaplexNZ]

But then you're back to the same situation where (as per Pieroxy's scenario) if Windows crashes beyond recovery you may have data on the SSD that isn't yet on the HDD. Assuming the two options share the same cache writeback policy, there's little difference between this OCZ solution and a custom one where the SSD and HDD are separate units - negating Pieroxy's claim of it being "ONE DRIVE" being an advantage.

Re:I've been waiting for these

[LordLimecat]

It might be a read-only cache (that is, not a write-cache). Absent further details, it seems rather unnecessary to automatically assume the worst, least reliable implementation when there are other implementations that offer benefits with no drawbacks.

Re:I've been waiting for these

[Anonymous Coward]

I hear rumblings that Windows 8 is going to finally solve this once and for all at the OS level. Designate a drive as a cache drive and it'll fill it up with frequently used files and the current cache.

This is stupid. What about, my windows crashes and is beyond recovery. Now all my data is spread on a "cache" SSD drive and on the physical drive. How do I get that onto one drive again?

Err, writing data to a cache doesn't delete it from the source.

Re:I've been waiting for these

[spire3661]

Is the XT a SSD, no. Can it increase performance over pure HDDs? YES. ITs a hybrid device and so it acts that way. Of course SSDs are faster, they are also ALOT more expensive or a 2 drive solution. The biggest strengths of the XT are very fast 'time to usuable desktop' that wont fade over time, large storage, realtively cheap price point ($99/500 GB), and single drive solution. If you go in understanding this, you will be alot happier. Its a good product that fills a niche at a decent price. Im very happy with mine in my desktop gaming rig.

Re:I've been waiting for these

[Anonymous Coward]

your correct, all caches are terrible and unstable by design. lets get rid of them all.

And btw, its MUCH more likely that one of the two drives in your "ONE DRIVE" system will burn, killing all your data before windows crashes to the point of unrecoverability.

Re:I've been waiting for these

[Anonymous Coward]

umm. . . you do realize that the SSD cache updates the platter HDD regularly. . . right? As in, up to as fast as it can spin, or as slow as seconds later. So worst case, when your windows crashes catastrophically and irreversibly. . . somehow. . . . you lose seconds of work.

Oh the humanity!

Let me guess, you were against RAM when it came out as well. "This is stupid, now my data is all spread out over RAM and my HDD".

Re:I've been waiting for these

[nabsltd]

Well, one of the point of caching on an SSD-type memory has to be the fact that your cache is non-volatile. So when you power up your drive the next day, the cache is still there. Why not doing write-caching then?

Write caching to an SSD would be slightly dangerous, but not really if you have a UPS.

Basically, the write would happen to the SSD, and then eventually be copied to the hard disk. That "eventually" doesn't have to be a very long time, though, as even a gigabyte can likely be written in the background in less than 30 seconds. Once that has happened, the cache would be consistent, and you don't have "data spread on a 'cache' SSD drive and on the physical drive" like you are worried about. The hard drive would have all the data, while the SSD would merely have a copy of some parts of the data, so moving the hard drive would be no problem.

Re:I've been waiting for these

[nabsltd]

OCZ just paid for a little marketing to get their name at the front of the queue for a couple days.

One of the advantages of this OCZ solution using a PCI-e slot is that it can bypass the SATA bottleneck. For this particular setup, that might not matter, but for their SLC-based SSD Revo drives, you can get unbelievable performance.

Re:I've been waiting for these

[Dahamma]

Ok, I think someone needs a tutorial in how caches work...

The most effective cache size for an application depends on a lot of factors: implementation (LRU, associative, adaptive), dataset, relative latencies, etc. But based on those if you get a 90% hit rate at N MB, and a 92% hit rate a 2N MB, making it bigger is usually a waste of money. For example...

Mechanical HDDs usually have a RAM cache of 8-32MB depending on HDD size. That's almost 3 orders of magnitude smaller than the XT's SSD. Is that RAM cache useless?

Most operating systems use extra RAM as a buffer cache - that's a few GB for possibly TBs of disk - is that useless?

Your web browser uses a couple GB on your disk to cache the entire Internet - is that useless?

A modern CPU usually has tiny (32-128K) L1 cache, a small (256K-1MB) L2 cache, and sometimes a somewhat larger (~8MB?) L3 cache. Are those useless? The L1 cache is really small, is that more useless than the L3 cache? In fact, for CPUs the diminishing return if the cache is fully associative is about 64K - if it's direct or 2-way associative it may be more like 256K-1MB, but beyond that it's not worth the additional cost.

Oh, and as far as my guess at 2/3 the performance of the OCZ for 1/3 the cost... RTFA :) This was comparing the OCZ hybrid with 60GB SSD used as cache for a 5400RPM HDD vs. the XT which uses 4GB SSD as cache for a 7200RPM HDD. There was no comparison to a direct SSD, and the OCZ hybrid is going to have the same performance issue "on the first run".

Re:I've been waiting for these

[StikyPad]

That article is 3 years old... NAND SSDs have come a long way since then, both in performance and price/performance. SDRAM prices and performance have increased as well, of course, but the limiting factor (according to TFA) is still the network -- even using Infiniband -- so I suspect modern NAND SSDs might suit them just fine.

Re:I've been waiting for these

[StikyPad]

I got fed up with the limited capacity of the standard drive and out of frustration bought my own 500Gb Momentus XT.

Similar story here, though I expensed it along with the packaging it came in, which happened to be a new computer.

Love for OCZ

[mcrbids]

Recently, after reading performance reviews on the Vertex 3, I bought one. The speed is simply amazing! I've been using it as a data-intensive development database server drive. Shortly after buying it, I discovered that there were numerous complaints about the Vertex 2 being unreliable.

To this, I can only say that after about 6 weeks of extremely heavy use as the data partition on a PostgreSQL 8.4 server I've had no issues so far.

Re:Love for OCZ

[Anonymous Coward]

And that hard drive looks permanently fastened to the border, so no worries about pesky upgrades later on. Just throw it away and buy a new one!

Re:Love for OCZ

[Dahamma]

Well, I just bought a Vertex 2 (SATA 2 rates are plenty for me and I couldn't pass up a crazy deal - 240GB SSD for just over $300!) so I hope it's reliable ;)

Only been a week but so far but (in a desktop) it's SO worth it in boot time alone (and the fact that it's totally silent doesn't hurt).

Re:Love for OCZ

[MightyMartian]

Yeah! I bought one and my penis just doubled in size, my girlfriend suggested a three way with her hot girlfriend and my boss gave me a 30% raise. And I haven't even plugged it in yet!

Re:Love for OCZ

[Dahamma]

Wow, congrats, good things really do come in threes... 30% raise, a 3 way, and now your penis is 3" long!

Re:Love for OCZ

[yarnosh]

Hell, the Vertex 1 is pretty amazing. I installed one of those in my 2nd gen. Macbook Pro and I go from spinning boot icon to functional desktop in like 3 seconds.

Re:Love for OCZ

[dbIII]

I tried to mirror a vertex1 and 2 - they use different types of memory and the 240GB vertex2 drives are actually quite a few GB smaller and unfortunately the data I was going to put on there would thus no longer fit. I ended up striping the things but it took three attempts to get a stripe size with performance equal to or better than a single drive. Attempt 1 and 2 were dog slow with writes and not good for reads, third attempt outperformed both single drives for reads and isn't too bad for writes The moral of the story is the obvious one of using identical drives, but at the time of the original drive getting two of them would have been a bit too expensive.

Re:Love for OCZ

[Anonymous Coward]

Cmon guys this ain't a review site

Re:Love for OCZ

[TheLink]

Keep up to date backups of any important data:
http://www.spinics.net/lists/pgsql/msg122280.html
http://www.codinghorror.com/blog/2011/05/the-hot-crazy-solid-state-drive-scale.html

I'm tempted to get an SSD (or two), but the failure rates seem rather high, and the failure modes too often are worse than normal HDDs (drive totally dead or even "time warp" drive rolls back to a state X days ago : http://www.ocztechnologyforum.com/forum/showthread.php?83778-Time-warp-drive-vanishing-after-3-days-data-gone-on-reboot...I-need-3-to-5-users-with-this-issue-to-help ). The time warp failure mode doesn't puzzle me that much, I can think of reasons why it would happen - but it worries me from a design perspective.

Yes normal spinning disk drives do suddenly fail completely too, but that normally is a result of abuse. So far what normally happens is you have bad sectors and/or the SMART reporting stuff starts giving you some ominous signs).

If the SMART stuff worked well for SSDs, even if SSDs failed every 3 months or so, many would be happy enough - they'd buy a stack of SSDs (yes really) and swap them in whenever they see the SMART warnings.

huh?

[Anonymous Coward]

With HDDs so cheap, why shouldn't the base configuration be 1TB?

Re:huh?

[AHuxley]

They seem to want to back ~10% of the storage with SSD. Thats the OS, select files and applications?
Your game/photoshop/browser will load SSD fast, the large video clip last watched 4 months ago may not.
How smart will MS be about fitting the 50 or so gigs of used files and sorting 'todays' work?
Will you hit a part of your 500 gb @ 5,400-RPM sort mid game?

Re:huh?

[swilver]

Actually, all you need to cache is meta-data and small files (say anything less than 100 kB). Anything larger (assuming it is unfragmented) can be streamed from a traditional harddisk at speeds comparable to SSD's anyway.

Large files are almost by definition rarely accessed randomly as they are usually some kind of media (image, music, video).

Also, at today's data density, even traditional harddisks can saturate a link as long as the reads are sequential.

Re:huh?

[adolf]

Interesting, indeed. I can't disagree with anything you said.

But back in context: Does this OCZ device do this sort of selective caching? It seems to me that, in order for it to do so, it'd have to be aware of (at least) the filesystem. That's not so far-fetched, given today's low-cost CPUs and FPGAs (we've had various NICs running Linux available for years now, for instance, to wring out the nth degree of latency), but again: Does it do any of this?

Or is it just a somewhat-larger SSD-based multi-gigabyte caching system of the same sort that common and modern operating systems already supply with RAM caching?

Or is it worse? (I suspect this, but without a product, I can't say...)

Re:huh?

[dbIII]

As always, it entirely depends upon what you are doing.
While at home I've got a relatively cheap 30GB SSD purely for game files I needed a pair of striped 240GB SSDs for files that are continuously getting accessed at work by about a dozen people. It's not a real database so an hourly snapshot to spinning disk does the job of redundancy reasonably well. If the SSDs die the snapshot is always in a usable state even if individual files are up to an hour out of date.

Re:huh?

[gman003]

Gaming. Most games use some sort of single-file archive for all the textures, meshes, animations, and sounds. With modern games, those are usually 4-10GB. And the access patterns are only rarely sequential. Hence why some gamers invest extra on 10krpm hard drives or (now) SSDs - loading times in some games can be ridiculous otherwise.

Re:huh?

[MrNemesis]

You're mostly right, but SSD's have been outstripping platter HDD's in sequential performance for at least a couple of years now; my comparatively ancient and pedestrian OCZ agility could do sequential reads and writes faster than the 150GB raptor it replaced (across the whole "surface"), and random access patterns were already a couple of orders of magnitudes faster. There are still very few SATA discs that can saturate a single 150MB/s connection on a sequential read, whilst SSD's were bandwidth limited to ~270MB/s until 6Gbps SATA arrived earlier this year in the form of new AMD and Intel chipsets (I don't count the 6Gbps marvell SATA controllers cos they're crap).

I picked up a 256GB crucial C300 last year that's currently used as scratch space for video editing - reads at 350MB/s, writes at 270MB/s, to get anywhere near that with platter based hard drives you're already talking spending comparable amounts on just a decent RAID card. Newer drives out this year are already pushing 500MB/s, to say nothing of the PCIe based SSD's.

If you need the space, use platters and a RAID card. If you need the speed, use an SSD. If you need both...

Back OT, this OCZ thing is nothing terribly new, and I suspect its thunder is going to be stolen in the consumer space by Intel's "SRT" (essentially software/driver based use of an SSD as either a read-through or a write-through cache) with a plan jane HDD (doesn't have to be a comparatively weedy 2.5" either) on the backend. Sadly it's windows-only and (stupidly) restricted to only the Z68 chipset at present. Enterprises are already using things like ZFS-with-an-SSD-cache (or some other SAN/NAS with built in flash cache) or, like ourselves, Fusion-IO in between the SAN and our blades.

Anand did a fairly thorough comparison of intel's caching technology here http://www.anandtech.com/print/4337 and I suspect the OCZ card won't be any better performance wise (and don't get me started on OCZ's QA...!). Well worth a read if you're thinking about building a computer in the next six months.

Re:huh?

[Hatta]

Actually, all you need to cache is meta-data and small files (say anything less than 100 kB).

Are there any file systems that manage this transparently?

Not new

[Anonymous Coward]

There are at the very least two other solutions that do the same thing, that were out there before this one:

HighPoint RocketHybrid: exactly the same, an expansion card with connectors for one ssd and one hdd.

Intel Smart Response Technology: Software on top of the Z68 chipset that uses an SSD of up to 60GB as cache for a different drive or raid.

really bad technology

[slashmydots]

I looked into existing hybrid SSDs for notebooks lately and was scared off by piles of negative reviews. Most now are 4 GB flash, 250+ GB mechanical and it intelligently determines what you open most often and transfers it to the flash section. Apparently that causes poor performance in certain circumstances, unpredictable performance in all circumstances, and lots and lots and lots of crashes and blue screens. If they got all that taken care of and upped it from 4 GB to 60 GB this would be a nice solution but for $350 I'd just get a 160GB or bigger SSD! What an insane price point even for a PCI-E version!

Also, the price difference in motherboards is huge. I have a graphics card in PCI-E 1 of course and my board has just one x16 slot. Needing a 2nd 16x or even 8x PCI-E slot means around double the price on average. So if it went from a $90 board to a $180 board, you might as well have just bought an even higher capacity SSD for the same price. It would work with non-graphical computers but with SSDs, quite a few are gaming computers. I think I'm better off getting one of those 400MB/sec+ 32-60 GB SSDs and installing my games on it. Most new ones support that. My boot time would still be crap but my 3D models would skin ultra fast and the load times would be really quick. Now that's a hybrid system.

Flash as cache means a short life span

[erice]

I would hope these devices have a good warranty. Flash has limited write endurance and writes are slow too. Cache operation is all about concentrated writes and rewrites. Cache is just about the worst application possible for flash.

Re:Flash as cache means a short life span

[Anonymous Coward]

have you personally written a decent SSD to death? why not use your fast drives to make a system faster? I personally think that these types of arguments belong about three years ago, with CF-IDE adapters, 4GB netbooks and crappy jmicron SSDs. turn off journaling and put your page file on a RAM drive while you're at it.

Re:Flash as cache means a short life span

[SpazmodeusG]

You're are correct, SSDs don't have any issues with read/write cycles. But this following line is hilarious;

put your page file on a RAM drive

Think about it. You're talking about creating a virtual hard drive using RAM which is used to store page files that you don't want in RAM.

Just turn your page files off altogether. In this scenario there's no disadvantage to doing so. The total available RAM (real+page file) will be the same either way and you save yourself from going through a completely pointless layer of indirection.

Re:Flash as cache means a short life span

[Anonymous Coward]

I have. SSDs are barely more reliable than what passes for reliability on shitty consumer hard drives these days.

Re:Flash as cache means a short life span

[Anonymous Coward]

lol, I was joking... but, seriously have a look at these "tweaks" on the ocz forums. there really are people advocating this deviant behaviour.

http://www.ocztechnologyforum.com/forum/showthread.php?47244-XP32-64-SSD-Windows-Registry-Tweaks

Re:Flash as cache means a short life span

[Anonymous Coward]

Don't disable your pagefile unless you know you will never exhaust memory. Instead, put your pagefile on the slowest possible media so you'll know you're paging. When that happens, go to the task manager and choose to kill a process that you can afford to lose, instead of losing the (random) process that had the misfortune of exhausting memory.

Re:Flash as cache means a short life span

[Anonymous Coward]

even more reason not to try to use them as long term storage, as cache they'll hopefully do less damage when they die.

Re:Flash as cache means a short life span

[Anonymous Coward]

The Arachne web browser actually had/has this as as a standard method of speeding it up. Of course, it is a DOS application.

Re:Flash as cache means a short life span

[Anonymous Coward]

That's what I first thought as well. However, CPUs also have a solid state cache, so they're probably using a similar technique.

Re:Flash as cache means a short life span

[dbIII]

The way MS Windows swaps makes perfect sense in some situations but appears utterly braindead in others. That's why some corner cases with things like some games on machines with a lot of memory can actually benefit from doing things that appear to be utterly stupid - eg. virtual drive on ramdisk. Without a page file at all you can find the bugs where there is an assumption that there is a pagefile :(
Maybe MS Windows 7 acts more like a server OS and handles virtual disks in a more sensible way than XP.
On a real server OS I assumed the users wouldn't chew up the 64GB I had on one machine but gave it 8GB of swap anyway. It needed all of that for one task and in hindsight it looks like I should add a resizable swap file in addition to the swap partitions. Once you put something new on a machine everything could change - for instance on an MS Windows machine it could take something as simple as a tabbed web browser to go from the requirement of not needing virtual memory to needing a lot of it.

Re:Flash as cache means a short life span

[yarnosh]

Depends on how aggressive the cache is. It doesn't need to be very aggressive if it is non-volatile. Say for example you are caching some oft used OS file. With a smart cache controller, it shouldn't be a whole lot different than having two separate drives, one for system, one for apps. Except you dont' have to go out of your way to manage that. Nor does the OS.

Re:Flash as cache means a short life span

[Anonymous Coward]

> Flash has limited write endurance ... and hard disks have limited life-span. So what? "limited" is a meaningless word without actual hard numbers. And the hard numbers for flash aren't that bad if you take wear leveling into account, plus the possibility to use an SDRAM-based cache next to the flash for small writes.

> and writes are slow too

Wrong. Clearing flash is slow; writing to flash is very fast. Clearing free spaced is done in the background, so you'll only ever notice the slowness of clearing the flash when you repeatedly overwrite the whole flash. Which isn't going to happen, even with large bulk writes, if the cache controller is properly designed. At worst, a too-large bulk write would degrade the speed of the disk to that of a regular hard disk.

Re:Flash as cache means a short life span

[GigaplexNZ]

Solid state cache: yes. Flash based: no.

Garbage Brand

[Anonymous Coward]

OCZ just recently swapped their NAND for cheaper, denser, slower NAND. They didn't even change the model #. When enough complaints came in, they were forced to RMA everyone's drives or face a bait&switch lawsuit.

Re:Garbage Brand

[Mashiki]

Citation Required.

Re:Garbage Brand

[the_jone]

I don't know if they're a garbage brand, but OP is certainly right. You can read OCZ's own announcement, or read Anandtech's analysis in their Vertex 3 review. Storagereview did a comparison of the 32nm and 24nm Vertex 2's which is also worth a read. .

Re:Garbage Brand

[tgeek]

Citation Required.

http://www.anandtech.com/show/4256/the-ocz-vertex-3-review-120gb/2

Not sure this incident should brand OCZ as a "garbage brand" though. If anything it highlights the necessity for the consumer to do their homework and not rely on just the summary specs of a drive before making a purchase. (That's not meant to defend OCZ's mistakes - just pointing out there's more to selecting an SSD than just the summary specs you see printed on the box or in the results of a Newegg search)

Re:Garbage Brand

[Confusador]

GP's vitriol was a bit over the top, but his facts were basically right.

After a dose of public retribution OCZ agreed to allow end users to swap 25nm Vertex 2s for 34nm drives, they would simply have to pay the difference in cost. OCZ realized that was yet another mistake and eventually allowed the swap for free (thankfully no one was ever charged), which is what should have been done from the start. OCZ went one step further and stopped using 64Gbit NAND in the 60GB Vertex 2, although drives still exist in the channel since no recall was issued.

Re:Garbage Brand

[anethema]

They did the change way before their announcement, in secret.

They swapped the chips from 16 lower density to 8 higher density chips without telling anyone at all. It had to be discovered, and was noticed basically when peoples performance was no where near what others real world performance was.

How were the first few thousand supposed to do their homework? There was no model number changes at all. Not even a "-a" or "-25".

Re:Garbage Brand

[spire3661]

OCZ is definitely shady. They come out with bleeding edge stuff, but they skimp on QA and they do bait and switch. I would like to point out that the newest sandy bridge boards can do this behavior with discrete parts now too (SSD cache + mechanical)

Re:Garbage Brand

[adisakp]

You also lose storage space with the newer NAND - the 60GB drive formats to 55GB and the 120GB drive formats to 118GB. So it's a combination of being slower and being more expensive per GB. The storage space is lost because smaller flash is less reliable and they need to set aside more storage for error handling when blocks fail. One of the main reason the drives are slower is that they use half as many storage chips so there are less lanes in use. If you get a larger capacity drive that still populates the whole board, they don't slow down as much but the slowdown is noticeable on the 60GB model.

Re:Garbage Brand

[Anonymous Coward]

I agree, OCZ has aways been a shady, "enthusiast", overpriced performance brand. Often slapping heatsinks on memory chips and selling them as "overclocked" and other skeevy things. If they want to be taken seriously as a corporate infrastrucuture hardware provider, they better get their shit together and quit messing around.

Re:Garbage Brand

[StikyPad]

That's not bait & switch -- bait & switch is when an ad baits you in with one product but there's little or no stock on hand, so the buyer is switched to a costlier product. It's also perfectly legal (though scummy) in the US if the ad makes it clear that there's limited stock.

Similarly, the internals of a product are completely irrelevant so long as they're not mentioned as a selling point in advertising, or if there's a disclaimer that specs are subject to change at any time without notice (provided there are no local laws that prohibit this practice). And unless you were buying direct from OCZ, it's the retailer who would be liable. If OCZ accepted RMAs, they were going above and beyond their responsibility.

As you say, they may well have faced a lawsuit if they didn't, but anybody can be sued for anything; that doesn't mean there's any merit to the claims. And it doesn't hurt that the good publicity of exchanging the drives was probably money better spent than by fighting customers who felt slighted.

That's simply packaging.

[obarthelemy]

NewsFlash: SSD are much faster than HDD, but much more expensive, so it makes sense to use an SSD to cache HDD content. For some obscure reason (probably the need for something new in Win8); MS refuses to do that (even though they can ReadyBoost off of a USB stick...), so it has fallen to third parties to implement it in software of hardware, even though it should really be the OS doing it.

How it's actually done is of no real import, all are kludges anyway because MS is, once again, letting us down.

Re:That's simply packaging.

[PhilHibbs]

I disagree, for the same reason that I won't use RAID-0. If the OS goes down and I have to recover my data on another machine, I just want to be able to plug the drive into another PC and have it readable. If I have RAID-0 or if I have some "smart" dynamic moving of files from one device to another hidden behind a single drive letter, are the devices going to be individually readable? I've had too many drives fail to trust "smart" systems like that. I know, backing up is the only way to avoid data loss, but I'd rather have the intermediate option available for the drives that didn't fail.

Having said that, I'm tempted to get a Drobo, which is RAID, as a backup device. But that's

Re:That's simply packaging.

[obarthelemy]

one does not prevent the other: to take the example of ReadyBoost, it was pure intelligent caching (of small OS read-only files): you could take your USB key away, the OS still booted (not sure if you could actually unplug while running, but that's probably not very important). The cache can simply duplicate files, and, implemented at the OS level, it can do so very intelligently: cache preferrably small files, that are only read, and frequently; never cache frequently-written files, use a fancy format with as few writes as possible (timestamps...)...

Uncle Joe reports... Really?

[Anonymous Coward]

> J. Dzhugashvili writes

That's pretty offensive for a large part of the world population!

Just for the record: Josif (Vissarionovich) Dzhugashvili was the birth name of Stalin, the soviet communist dictator, who was more interested in sending people to the Gulags for hard labor, to give them bad memories, rather than hard drives combined with memory chips. Around 25 million did not return, in total.

Wonder if an austrian painter, Adolphus Shicklgruber will report next time about the new USB 4.0 draft standard or so.

Re:Uncle Joe reports... Really?

[GottMitUns]

Hitler never had any name like that. His name was always Adolf Hitler. Shiklgruber is a creation of British war propaganda. Please do not proliferate the lie.

Re:Uncle Joe reports... Really?

[shutdown -p now]

For that matter, it's also a fairly common Georgian name. Are you suggesting that Georgians are therefore inherently offensive to "a large part of world population"? ~

Ampersand

[Anonymous Coward]

When concatenating a list of just TWO items, please use the AMPERSAND which is intended for this purpose.

The comma is reserved for multi-item lists.

OCZ Couples SSD, Mechanical Storage -> OCZ Couples SSD & Mechanical Storage

You know it MAKES SENSE

Re:Ampersand

[Anonymous Coward]

You know, you're a waste of oxygen.

Re:Ampersand

[Anonymous Coward]

... and you should always end a sentence with a period. I love how these grammar nazis ALWAYS make one typo themselves, then think "holy shit it can happen to me too!" before finally realizing that IT DOESN'T FUCKING MATTER. If you have nothing to contribute to the discussion, shut the fuck up and move along. Yes, you are an oxygen waster.

HSDL?

[creat3d]

Their HSDL drives have transfer rates of 720mb/s, how about evolving from there instead of regressing?

Honestly underwhelmed

[bughunter]

I'm honestly confused. And still open-minded.

Explain to me why I need this instead of spending that $300 on a 7200 or 10000 rpm drive with two to four times the capacity and far, far more reliability and service life.

Is it just to extend the time I can operate on battery power? A second battery will double my time at far less cost.

Re:Honestly underwhelmed

[creat3d]

It will be worth it when you can get a 1TB SSD for that 300$, rather than the current, overpriced mess of some great drives in a sea of overhyped duds.

Re:Honestly underwhelmed

[Anonymous Coward]

Why should I buy a 400hp car when a 80hp car gets me from point A to point B. Answer: some people like performance.

1 SSD vs two 10k rpm drives in RAID 0 is still night and day difference, not just due to the mb/sec, but due to the IOPS/sec and near zero latency.

With SSD boot times of 15 seconds are normal, vs sometimes over a minute before a regular HD is done fully loading, and you can open every program on your desktop at once and they pop up like popcorn... regular drives do not multitask well at all and when performing multiple tasks just fall over flat.

Re:Honestly underwhelmed

[frovingslosh]

And before you spent that absurd price on a 10,000 RPM drive I suggest that you look at it's transfer rate as compared to a basic Seagate 7200 RPM drive. My understanding is that the Seagate actually beats out expensive drives like the Raptor. The reason? Because the 10,000 RPM WD drive still uses traditional recording while the Seagate uses perpendicular recording, packing the bits much closer together and thus allowng more to be read in the same time, even at the lower rotational speed.

Now, to be honest, this is the way things were a couple of years ago. Maybe WD updated their technology and I missed hearing about it. But there were plenty of people who thought that they were buying a faster drive when they shelled out 3X as much for a 10,000 RPM drive, and I expect there are still plenty of people who buy based on that 10,000 RPM number alone.

Re:Honestly underwhelmed

[bughunter]

Why should I buy a 400hp car when a 80hp car gets me from point A to point B[?]

Using your analogy, I have about 250 horses under the hood, and if I had more, I really wouldn't use them 90% of the time.

And inverting it, my CTO has a Lotus Exige S and a Chevy Corvette Z06, and drives one of them to work nearly every day, both very nice toys. But he doesn't need over 500 HP to drive the 1.5 miles from his home to the office.

So again, I ask, why do I *need* one of these hybrid drives?

Re:Honestly underwhelmed

[bughunter]

Actually, a Hitachi 7200 is what I have. Boots 10.6.7 to the desktop in under 20 seconds.

As you suggest, I didn't see any value in paying the 100krpm price.

And I had a series of Seagate drive failures, so I thought I'd try Hitachi this time.

Re:Honestly underwhelmed

[GigaplexNZ]

So again, I ask, why do I *need* one of these hybrid drives?

You probably don't. That doesn't mean others won't appreciate the option.

Re:Honestly underwhelmed

[adisakp]

And before you spent that absurd price on a 10,000 RPM drive I suggest that you look at it's transfer rate as compared to a basic Seagate 7200 RPM drive. My understanding is that the Seagate actually beats out expensive drives like the Raptor. The reason? Because the 10,000 RPM WD drive still uses traditional recording while the Seagate uses perpendicular recording, packing the bits much closer together and thus allowng more to be read in the same time, even at the lower rotational speed.

There are multiple parameters in drive speed. Sequential operations may be faster on a drive that uses perpendicular recording but random operations will be much faster on a higher RPM drive. In fact, sequential performance on HD's may be faster than many previous generation SSDs. Then again, if you are doing lots of random operations, a good SSD will kill ANY mechanical hard drive for speed.

Re:Honestly underwhelmed

[Mia'cova]

The perf of an SSD blows away traditional hard disks. For me, my killer app is outlook. On my laptop, I boot it up once every day or two. On my old HD, it would spend an hour downloading/indexing the mail. That made the machine pretty unusable. The ssd flies through indexing and other bulk operations. Also, you never get those system hangs when multiple apps want the hd, causing 'disk thrashing' where the head is flying back and forth and spending all of its time on seeks, rather than throughput. One really really nice benefit of an SSD is that the system is completely usable while the disk is under load. No more slowness while installing an app for example.

But yes, it does extend battery life. The hybrid approach also lets you get many of those perf and battery benefits without spending tons of money. If a disk goes up $50 to massively improve your system perf in common scenarios like bootup/resume, and outlook while also improving battery usage (cuts idle usage down by about 1.5 watts), that's a huge win without sacrificing too much in cost or disk size.

Re:Honestly underwhelmed

[bughunter]

OK - thank you for an honestly informative explanation. Speeding Outlook database management might indeed be something I would need...

It seems like bcache could do the same thing

[LukeCrawford]

with your own hard drive and your own SSD. It'd probably cost less, too. http://bcache.evilpiepirate.org/

Re:It seems like bcache could do the same thing

[elh_inny]

Or ln-s on linux /Unix or mklink on windows.
Basically if you can afford to micromanage the caching and designate only specific files/folders to SSD, you will get the best performance characteristics.
No need to have your mp3 on SSD even if you listen to them often, likewise, even if something evicts your kernel files from the SSD cache, you will need it again soon enough...
That's why I have SSD and HDD separately and manage caching myslef to an extent.

Re:It seems like bcache could do the same thing

[LukeCrawford]

yeah, the advantage of bcache is that it will automatically cache the stuff you use often (which yes, can be a disadvantage, too) - I'm just thinking for something like a database server (not that I'd use bcache in a prod database server until it was more stable and until it supported mirroring the SSD) you have large files where parts of the file need caching and other parts of the file don't. A block cache helps with that more than manually symlinking files.

No Thanks

[frovingslosh]

Wow! The very high cost of a solid state drive coupled with the power hunger and mechanical constraints of a hard drive. I find it hard to believe that using SSDs as a cache is wise, but I also question if the money couldn't be better spent adding more DDR3 RAM to the processor and letting it cache the drive (with the obvious options and benefits that brings). I think I would just stick with the 2TB drives that I've been buying recently for well under $100 each rather than this overpriced compromise, even if I had a 4x PCIe slot available (which I don't and I expect few who would have an interest in this do).

Re:No Thanks

[Anonymous Coward]

bigger volatile write caches mean more data to lose in a crash/power loss situation.

Re:No Thanks

[yarnosh]

If you've ever run a good SSD like OCZ or Intel, you'd know that there is much to be gained by storing often used files on there. I can't speak for desktop systems, but installing an OCZ Vertex in my laptop made a huge difference. And if I needed teh space, I'd definitly consider an SSD/HDD hybrid. I guess the problem here is that they're selling to desktop users who already have fast HDD options (if you include RAID). Either way, if 2TB is your target, I doubt speed is a huge concern.

Re:No Thanks

[frovingslosh]

Things can go wrong in a crash, period. If you care about them don't use the OS most likely to crash. There are software strategies for minimizing the time data is in write cache unprotected, yet still giving most of the performance benefits. One shouldn't be running a computer without some form of power backup, and I expect that few willing to pay $350 for a drive do, So better to spend just a fraction of the cash on a UPS. And while I'll admit that power supplies might fail, and have even used dual supplies in mission critical applications, I have yet to see one of my own fail (I have seen someone else's bargain basement supply cause crashes), so my thinking is still that there are better ways to spend the money than on this turkey. People buy solid state drives to get away from the issues of hard drives, but this drags back the hard drive concerns at SSD prices.

Re:No Thanks

[Anonymous Coward]

Actually it's more the best of both - the fast speed of SSD due to the cache but with lower cost because it's much smaller, combined with the cheap space of a HDD which doesn't consume as much power as normal since a lot of stuff is served from the cache so the HDD can stay in standby a lot of the time.

Of course it's useless for anything really disk intensive such as video editing, since the cache won't be any help there.

Re:No Thanks

[Christian Smith]

I find it hard to believe that using SSDs as a cache is wise, but I also question if the money couldn't be better spent adding more DDR3 RAM to the processor and letting it cache the drive (with the obvious options and benefits that brings).

The benefit of using FLASH as a cache is that it is much denser than RAM, and therefore you can get much more for your money. While your CPU spends most of it's time working out of it's cache and RAM, the data not being referenced can be stored on FLASH in the knowledge it is a millisecond away from being usable. While not low latency in computer terms, it is blindingly fast in human terms, and more than makes up for the swapping that isn't required to disks.

Of course, this assumes your working set of data is bigger than available RAM, which even on modest desktops today, is unlikely for personal use. Where it really comes into it's own is on server machines, where the working set can be significantly bigger than RAM. Buying 60GB of RAM is still relatively expensive, whereas buying 60GB of FLASH is not, and against network stack latency, the latency of FLASH is not so significant.

Re:No Thanks

[Anonymous Coward]

I work 'in the biz' and I agree that this seems silly at least for enterprise systems (which is what I work on). It seems a better option is just to load the hard-drive array with both SSDs and HDDs and let the system handle sending the 'fast' bits to the SSDs and transferring them to the 'slow' tubs at their leisure.

For home users, it makes some more sense, but then again, what home users really care that much about HDD access time. The power usage of SSDs is what makes them great for laptops and all, and while this would have a 'low power slow motor' it would still have a motor, which is the main problem.

Re:No Thanks

[Anomylous Howard]

I think you are advocating using SSD for swap. If you're running a very swappy system you'll wear out your SSD pretty quickly.

Re:No Thanks

[CSMoran]

the money couldn't be better spent adding more DDR3 RAM to the processor and letting it cache the drive (with the obvious options and benefits that brings).

The catch is that this RAM cache is lost upon reboot and you have to populate it anew. With the SSD approach you get (at least in theory) faster boot times because the files are there already.

Re:No Thanks

[Anonymous Coward]

It is a committed cached storage, not volatile RAM (ie, if you turn off your computer, you loose whatever was stored on DDR3 RAM that hasn't been flushed to the slow disk). It makes sense for database servers (and most servers in general). Get 2 drives and set them up in a RAID 1, and you get full redundancy. Near full SSD performance with mechanical drive capacity at a fraction of the cost.

Re:No Thanks

[maraist]

You're implying write-intensity. I don't think SSD's have any faster write speeds than HD's (at least linearly). Further, most write operations can be journal'd / buffered. It's the random read speed or read-modify-write that's killer.
The question is begged, if 4GB of dedicated RAM cache (e.g. on a 6/8GB RAM installation) is insufficient for your RMW critical work-load, then perhaps 64GB is enough (e.g. if no more than 1/10th of your data is critical path). This seems like a reasonable assumption for today's common problems.

However, it personally is far from enough for any workload I've encountered in 6 years. My critical volume is on the order of 100GB to 500GB (random disk-seek on a multi TB of indexed data - e.g. hashed indexes). Though if I KNEW a specific table only had 64GB of fast-path retrieval, I could possibly specially allocate it's index. Just seems like a lot of work to fine-tune. But it would certainly be nice to know that, unlike a pure SSD solution I COULD grow beyond the size-constraints of the SSD. I think THAT is the target market (along with fast boot laptops, etc).

As for write-wearing on the SSD. At least intel had some intelligent write-leveling - didn't see whether this solution did it as well. Given the need for block-remapping, I'm assuming the answer is yes.

Re:No Thanks

[maraist]

Letting the OS do it is great in theory. I think ZFS supports this concept. Brtrs maybe in 10 years. Maybe MS will support it soon (as a reason to upgrade).
The problem for me as a user of the HW.. What if the SSD went bad? what data did I lose? What if I want to upgrade the disk? Previously the RAID HW or bare OS was very explicit about what went where. Do I trust that I can just clone a disk with a newer partition.. Or relocate the disk to a new machine? If you think of it like RAID-0, then obviously it's no different. But to the casual user, it seems to be more complicated - since the HD 'seems' to store all your data, and thus the SSD is just a 'cache', but in reality it's not necessarily fully committed after hard shutdown.. I can just see the user complaints down the road.

For the initiated, I'm sure it's fine though.

As for the laptop 'low power mode'.. My HD's were usually powered-down most of the time.. If the HD detects wake-up triggers and moves those data blocks over to SSD, then over time the need to wake up the HD reduces over time. So while statistical in nature, I do see value-add.

Re:No Thanks

[maraist]

Depends on the working set size. A 4-way RAID-10 controller is like $200. So with 4x2TB at roughly $150/disk, that's:
4TB of storage at $800

For RAID-10 on 500GB hybrid SSD platters with 64GB SSD cache, you'd need a 16-way RAID-10 controller: $1k

16x $300 + $1k = $5,800
That 16-array only gives you randomly allocated 512GB of SSD (maybe a little higher if you interleave which half of the RAID-1 and thus separately allocate SSD mappings - though this can't be controlled by the OS so it's largely random).

IF, instead you explicitly allocated indexes / hot tables onto a quartet of 128GB SSDs (say $300 each). Then we could have:
RAID-4 controller with 4x2TB drives: $800
+ RAID-4 controller with 4x128GB drives: $1,400
Total: $2,200

Less cost per HD failure, less power consumption, and more targeted index disks (SSD) and journaled disks (HD).
So indexes are guaranteed more efficient than on the mixed-mode storage, journals are equal in performance, and data-disks are slightly slower on pure HDs. We have slightly less SD capacity, but there's no guarantee the firmware will properly guestimate what data goes into SSD (same issue as cache-pollution).

You're still going to want 8GB to 128GB of RAM on any DB server. And if you're up to 128GB of RAM, then the SSD is of slightly less usefulness (especially on read-mostly nodes).

For non DB loads, RAID-5/6 might be a more cost effective solution, and thus those 64GB sections can start to add up to a usable working set. Hot regions of video metadata files can speed up drastically, while the HD's can potentially be leveraged for large spanning reads/writes (though don't know if you'd have cache-pollution).

Lastly, something like WAFL on netapp isn't going to leverage the hybrid at all, because it's specifically journaling all data (e.g. translating random writes into remapped linear writes, with explicit fixed-sized SSD mapping tables / journals).

Redundant

[Anonymous Coward]

RAID 0 array?

Didn't Seagate already do something similar?

[samael]

So, not that different from the Momentus XT - which is a hard drive with a bunch of SSD as a cache...

Re:Didn't Seagate already do something similar?

[Henk Poley]

This one is less convenient (not a standard SATA format) but faster.

Re:Didn't Seagate already do something similar?

[stubob]

The Momentus has a 4GB cache. This will have a 60GB cache. I haven't noticed much improvement in day-to-day use with the Momentus over a normal SATA drive. I think the cache is just too small. But maybe launching Windows, Firefox, Photoshop, VirtualBox, MediaMonkey, and a few other big apps isn't its designed usage and is causing cache swap.

Now, if I could tell it to *just* cache Windows, that would be an improvement.

Re:Didn't Seagate already do something similar?

[Anomylous Howard]

I've been hesitant to buy a Momentus XT, because of early flakiness and a firmware fix that seriously degraded performance. Anyone here have more info on this?

Re:Didn't Seagate already do something similar?

[spire3661]

Mine was acting up in TF2 with slight delays (all the voice loading during gameplay), but not BFBC2 on firmware SD23. I flashed to firmware SD25 and it works like a champ.

Re:Didn't Seagate already do something similar?

[deroby]

I read about quite a bit of issues on Apple hardware (MBP) with the drive (freezes, bouncing ball, etc)... Personally I experienced none of that on my Dell laptop (running Win7). although I had the same HDD-BIOS version when I initially installed the drive. As things just work I'm hesitant to upgrade (I'm at SD23) because it seems they've rather stripped options like powermgmt out of the drive than actually fixed the issue. Hence causing higher power-consumption in order to get it working on MacBooks, bit strange way to handle things as I'd expect the non-Apple user-base to be the bigger one ...=/

I *may* have been lucky in that my Dell BIOS already had Acoustic DIsk Management (or something) disabled which also seems to be related to the issues. There's a google spreadsheat on the Seagate forums that clearly shows the issue is almost entirely on MBP.

Anyway, I sure hope the *concept* didn't get too much bad credit due to these issues as I'm pretty sure this is the way to go until SSD becomes "as cheap" as normal harddisks, even for big capacities.

My take on the OCZ PCIe board here : it might benefit from the enhanced throughput as it avoids the SATA conversions, but unless you're a super-I/O consumer (random IO that is), I doubt you'll notice it much. That aside, pc's usually only have so many PCIe slots, but sport plenty of SATA ports !

PS : building said PCIe board into my laptop would be quite a challenge too =)
PS: yes, downgrading the BIOS is an option apparently, but still, why fix what isn't broken ? =)

Best and worst of both worlds...

[Tatarize]

You'd figure with a few gigs of SSD even as much as a silly thumbdrive worth you could get read speeds on par with SSD. SSD are crap for massive amounts of data, for now. Though, I think tying your data to die on a thing with moving parts, vs. and SSD gets the worst of both worlds too. You get the speed of SSD without getting rid of the potential to no longer have the silly things crash.

I'd prefer if people just really understood they are two different technologies and you should have one of each. You should have 20 gigs for your OS of SSD and you should have 2 TB of data for whatever else.

Somebody should invent a form of RAID for drives of radically different sizes and read speeds that just works a bit like stripping but as a cache. Make this silly doohicky here pointless.

Re:Best and worst of both worlds...

[IICV]

Actually, it looks like SSDs are currently less reliable than hard disks in practice (though of course that link is to Jeff Atwood saying he'll buy them anyway).

Re:Best and worst of both worlds...

[cynyr]

"/dev/root 92G 79G 8.7G 91% /"

This is my gentoo ex-desktop/now server install from 2 years ago, there is a /home and a /Storage on separate drives as well. 30GB of / is /usr/portage/distfiles/ and /usr/src/ is 10GB that still leaves 40GB of stuff just for the OS. I should probably track that all down and fix it...

Re:Best and worst of both worlds...

[tepples]

Somebody should invent a form of RAID for drives of radically different sizes and read speeds

Is ZFS anything like that?

Re:Best and worst of both worlds...

[Anonymous Coward]

There is -- create an array with mdadm and set the "write-mostly" flag on the conventional HD. That tells mdadm that that piece of the RAID is slow and to prefer reading from the faster mirror, but keep the thing up to date at its own pace (or something like that).

A UPS is not always the answer

[dbIII]

Where I am I get occasional power weirdness that flickers the lights and panics every UPS into a hard shutdown but everything on mains stays on. You can get it too no matter what part of the world you are in if somebody is using an industrial crane or similar big bit of gear within a kilometre or two. That's after spike filtering etc for the entire site. It's really annoying when the only things that go down are the switches, file servers and phone systems. A UPS is pretty well certain to fail once every three years or so anyway - many properly with plenty of warning but even some from expensive brands just do a battery test, fail it and turn themselves off and that's the first you know that it's time to replace the batteries. That's not how they are supposed to do it according to their manuals and some of the same model may behave properly, but they do it anyway.
Battery backed controller cards save my arse in that situation because the cached writes are still there ready to be done when the power comes back on.
As for redundant power supplies, I have a failed redundant power supply chassis on my desk because the redundant modules are still OK but the non-redundant chassis for them is dead. Everything dies. The OS won't save you completely from hardware failures.

I f-ing hate OCZ.

[rolfwind]

I was burned once. Paid good money 2 or so years back for 2 fairly pricey SSDs. A couple months down the line, halfway full, they started stuttering like failing harddrives or harddrives waking up. It was painful. There were few reviews online at the time of purchase, but when the problems cropped up, I found out that the controllers in them were considered complete crap and that OCZ wouldn't do anything about screwed customers other than replace failing ones with the same exact model even though they had a new one in the works.

Since then, I've stuck with intel and haven't regretted it. While I realize OCZ probably have their stuff together now, I don't look favorably upon companies that are willing to ship obvious crap. (And they must have realized the problem early in testing already.)

But yes, I do love SSDs. Still use HDD for long-term, rarely accessed storage, but for working on? No comparison.

bcache

[Anonymous Coward]

i wonder if this is better than doing it in software (like bcache). software would seem far more flexible, you could cache whatever large disks you already have, and can quickly swap out the SSD if it fails. but hardware may be faster.

Can I exclude some files or dirs from cache?

[cheekyboy]

Some files I never want cached, or some dirs.

ie. such as all my mp3s/avi's dont need to be cached.

What about the connectors?

[gr8_phk]

Are motherboards and the PCI connectors going to like the vibration levels of having the drive on the card? I'm not saying it will fail, I just don't know. It's not the normal use case.

Re:What about the connectors?

[drinkypoo]

It wasn't a serious problem with ISA slots and bigger, goofier hard disks, so it probably won't be a problem here, either. It does seem a bit daft not to float the disk just to reduce noise, though.

Re:What about the connectors?

[adolf]

It's not so abnormal: Witness modern video cards, with their sometimes-multiple jet-engine fans that are anything but balanced, yet are held in place with a single screw and a card-edge connector.

The vibrations from a hard drive, while having more rotational mass but at least intended to be balanced, seem miniscule to me.

you can force readyboost on SSD drives

[cheekyboy]

Googling i found this;, havent tried, but here it is;

Go to:
HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows NT\CurrentVersion\EMDMgmt\[drivekey]

Change:
DeviceStatus to 2

Let's see how well the caching works...

[Metabolife]

when I fill the entire storage space and randomly select files from across the drive to work with.

what...

[Gravis Zero]

no floppy disk?

5.25" and 3.5"

[tepples]

Anonymous Coward wrote:

you had to pick whether to put a half-height 5.25" or a 3.5" in the remaining half of the drive bay

TheRaven64 wrote:

With a hard disk, you didn't really need the second floppy disk.

Unless you wanted to exchange data with other machines using both sizes, or to use applications that used "key disk" copy authentication. Or did they have drives that accepted both 5.25" and 3.5" floppy disks in the era of sub-0.1 GB hard drives?

Re:5.25" and 3.5"

[drinkypoo]

By the time 3.5" disks became commonplace, so did 3.5" floppy bays. Macs were pretty much the only significant installed base before 3.5" floppy drives were 1" tall.

Re:5.25" and 3.5"

[TheRaven64]

Check the context - you missed a post in the middle. The poster I replied to had an Amstrad PC1512, which had two 5.25" floppy disk drives (double-sided, single density - maximum 360KB per side). I had the slightly more expensive PC1640 model, with a hard disk drive instead of the second floppy and 128KB more RAM. If I wanted to exchange data with computers that had a 3.5" floppy drive (like my father's laptop), I had to use a serial cable (later I got a parallel cable and Laplink).

Eternal sunshine of the HDD mind?

[Xeranar]

What is with the SSD hate? Eternal sunshine of the HDD mind? Seriously, SDDs aren't perfect but are usually still salvageable by a restoration firm because the card blows but the memory is intact. I agree they're not entirely reliable and they have shortened lifespan from what people seem to deal with (my 64 GB SSD going on 2 years and working just fine). I would though for $120-150 easily justify an OS boot drive that holds only programs and the OS while having a larger 1TB+ drive for other work. This just combines that into a standardized system which for the price is actually fairly cheap compared to large 64 MB cached 1TB hard drives if you look at the combination cost and the PCI-E x4 speed.

Course this is all relative and real world performance of SATA II is fast enough to handle Win 7 and most other items currently.

Oracle/Sun 7000 series

[Anonymous Coward]

Oracle/Sun(/StorageTek) sells entire SAN's based on the same concept. Their 7000 series is setup with a large chunk of SSD for caching reads/writes and then large, "slow" disks (2TB, 7200 RPM) to fill it out for capacity.