Slashdot Mirror
Hard Drive of the Future: Ram Drive
benzick writes "3d Retreat has posted a hands on look at a 2gig ram drive called the Rocket Drive. Article blurb: Overall the rocket drive is the best in I/O performance I have seen. It outperforms U160 SCSI drives by almost a factor of two. Yet there are some drawbacks to the Rocket drive, foremost is the price, although listed at the end of the review is some alternative pricing options to make it less expensive. And the rocket drive can not act as a boot drive. Also, if you have some extra money to spend, you can use multiple rocket drives in parallel."
It's another bullshit "hardware site" test.
The "benchmark" was a Photoshop filter on an image. It was twice as fast as with the SCSI disk.
This tells you very little about the relative performance of the drives since image processing is typically not disk bound.
For that price, I'll wait the extra 7 seconds for Mozilla to load, thankyouverymuch.
mstyne: real name, no gimmicks
And if someone trips over the cable, there goes your 2 gigs of data!
Thanks, but no thanks, I'll stick with mah good ole Winchester disks.
The hard drive is some ancient technology that is the *easily* #1 cause of all computer failures. Other than the cooling fans, they're the last moving parts, and the most critical ones too... A fan dying may cook your computer, but a hard drive kills your *data*. It's high time that something came along to replace those damn things. I'm typing this on my PC with a 2 drive RAID because I can't afford downtime or data loss. That really shouldn't be necessary any more. Bring on the alternatives!
Keeps that data after the pc is turned off, which i bet your amstrad didn't do.
It has an external supply that keeps the card powered.
And i believe this is the whole point of this card, its pretty much useless otherwise.
Also the xfer speeds are limited to PCI (66mhz) speeds, that is why "its only" 2x as fast as a U160 scsi.
This is really pretty interesting. The device has it's own power supply that actually allows you to save data when you shut down your computer. It doesn't seem like it would be too reliable, but it does provide a reason as to why this is better than a traditional RAM drive (provided you have it hooked up to a UPS).
1. A DRAM "drive" suffers the fundemental problem that if the "external" power source is lost, you lose everything on the drive.
2. 80-100 MB/sec sustained performance is nothing to write home about for DRAM performance. A RAID 0 stripe across 2 ATA drives could give you this same performance for about 1/4 the price without the power issue.
Although its a long way off, MRAM offers a much more promissing application in the area of high speed RAM drives.
Someone you trust is one of us.
Fundamentally, you're always better just to use caching. Essentially, this amounts to a 2GB dedicated disk cache, except that the power supply ensures that the contents survive boots (though I don't know how it would do in a power failure). Anyway, how often do people reboot their machines nowadays? Stuff stays in my computer's cache for months at a time.
So, why not just add the 2GB to your main RAM? Then the OS could use it as a disk cache if it were so inclined (and you'd be right where you are with a ramdrive) or else the OS could use it as actual RAM if you needed it.
In short, RAM is just like a ramdrive except more flexible.
Patrick Doyle
I mod down every jackass who puts his moderation policy in his sig. Oh, wait a sec....
- The OS wants to store something in RAM, trys to allocate RAM and fails.
- The OS decides to swap out some memory based on an analysis of page usage statistics, so it grabs the handle to the swap file.
- Then, the OS streams the offending pages through file I/O subsystem, worrying about waits and updating page counters and such. This includes dragging all the data over a busy system bus.
- The receiving device stores it in RAM.
Cool! What operating system could aspire to such levels of efficiency?Here's some neat facts. No real solution though. Reply with one if you have it...
Ram = Fast
Cache = faster ram
On Board Cache = faster
1. Processors used to not have a quick way to get to ram so cache was created (faster and expensive ram) and put on a chip.
2. Cache proved to be too expensive so they put it off the chips (pII)'s. Celereon's even took off some of the Cache.
3. Now that ram drives will be created, it's just another link in the chain to the HD.
now it'll be:
HD->Ramdrive->Ram->off chip cache->onboard cache
Each one of those levels cost more as you move to the right. This just puts another link in the chain.
The official website lists the capacity as 4 GB.
The bios is set so you can only put in the amount of memory that was specified by the card. So If I purchased the 2Gig empty card. My only option is to put in 4 sticks of 512 pc133 memory. If I wanted to upgrade my card to 4gig. I would have to pay for an upgrade, and then replace all the sticks of memory with 1024Meg sticks.
I use Macs to up my productivity, so up yours Microsoft!
The article mentions that multiple RocketDrives could be used in parallel. That would seem only to be practical on 64-bit PCI buses. One RocketDrive transfers 80MB/s, which is close to the maximum sustained bandwidth for 32-bit 33MHz PCI. 132MB/s is the burst bandwidth, and cannot be sustained for very long.
In fact, I would think this drive would interfere with other devices that rely on the PCI bus. I doubt you could get 100Mbps (~12.5MB/s) on the same PCI bus.
HD->Ramdrive->Ram->off chip cache->onboard cache
Each one of those levels cost more as you move to the right. This just puts another link in the chain.
Sure about that? For the price of the Ramdrive, I could easily get 2GB of DDR. Hell, for the price of the Ramdrive I could get a motherboard that supported 64GB and fill a moderate chunk of it. That thing has lower speed and greater access time than main memory and costs more, so just using RAM as disk cache would appear to be more useful under the majority of circumstances.
This tells you very little about the relative performance of the drives since image processing is typically not disk bound.
Actually, the author of the article made it disk bound, by forcing Photoshop to go into swap space with an image much larger than the available memory.
And you missed the HD testing pictures, measuring high throughput (sp?) and unbelievable low latency.
If these guys had made this compatible with scsi, like emulating a scsi card and 1 device, so it could boot it would rock.
...
From power off to up and runnig in seconds !
This would be ideal to store an OS, even for a server, and have the HDs configured to copy the os back to the card and reboot in case it fails to boot from the card for some reason
Hook the power source to a ups and you can probably keep the info for more than a week without external power should you need. (i mean, how much juice can this need?)
What if you could boot from this drive? If it were Flash Ram then presumably you could store your operating system on a RAM Drive and get it to boot instantly cutting down on the boring and annoying wait for the computer to start up.
Another drive could be used for storing files and such like (if it were thought to be somewhat dangerous medium to *store*) but in anycase it would be lovely to get instantly booting computers.
In addition, rather than booting, couldn't these drives make possible a kind-of ghost-like save boot-up state. ie Copy into memory very quickly the state of the computer just following boot up. Now that would surely speed up the booting process.
And I know everyone might say, oh that doesn't matter, but even though I use a Powerbook which with OS X stays up almost forever - I *still* need to boot occassionally. And it is *painful* to wait for.
Once it is built into my laptop I will be well happy... I hate chugging drives... and silence is golden
---- The Open Source Record Label : : LOCARECORDS.COM
"Sustained performance" is only a practical measure in a few uses (eg: multimedia). For most other things, latency (ie: seek times) has a far greater impact on performance. Even the fastest harddrives have seek times measured in milliseconds. With DRAM we're talking about nanoseconds.
The fundamental problem of "power is lost" can be solved easily by adding a battery on the drive.
Okay... I'll do the stupid things first, then you shy people follow.
[Zappa]
Anyway, how often do people reboot their machines nowadays?
I'll answer the related question, "why would anybody need to reboot a computer?"
Some slashdot readers are lucky enough to live and work in an environment that primarily uses BSD or Linux. But unlike some slashdot readers, I, Damian Yerrick, live and work in an environment that primarily uses Microsoft Windows. Therefore, I have to use Microsoft Windows.
Some slashdot readers are lucky enough to be able to afford personal copies of PC virtualization software such as VMware, so that they can run other operating systems within a window on their computer. But unlike some slashdot readers, I, Damian Yerrick, can't afford a VMware license. Therefore, I must run Microsoft Windows on the bare hardware, and if I want to run an operating system other than Microsoft Windows, I must reboot my computer to access it.
Some slashdot readers are lucky enough to be able to afford to buy additional hardware to shield their other computer from exploits of newly discovered vulnerabilities in its operating system. But unlike some slashdot readers, I, Damian Yerrick, can't afford a second motherboard, CPU, case, and keyboard on which to run BSD firewall software. Therefore, if I want to keep my computer connected to the network, I must keep my computer updated with patches from Microsoft Windows Update. Those patches often require a reboot of the computer.
Therefore, I, Damian Yerrick, cannot afford to own a computer system that does not have to be rebooted.
Will I retire or break 10K?
Most people complain about how this drive or RAM disks are completely wiped out by a power loss, but couldnt that also be used as a great security feature. If you have alot of data that you dont want someone (read: The Government) getting their hands on or taking as evidence, just keep it on a drive protected by a UPS, then if you need to wipe it out quickly and completely, just pull the plug. Or if a few lovely FBI agents come and take your computer, they do it for you.
"There's no way to rule innocent men. The only power any government has is the power to crack down on criminals."
This is just sad...
3,000 dollars for something that has 2 gig of ram. I could get 2 gig of ram for a fraction of that amount... In fact, for the speed its giving, I could fill the thing up with sdram or edo ram...
This is something I could imagine being useful with my hard drive... Why don't they make a standard plugin for hard drives... Make it where you can add cache directly to the hard drive.
But wouldn't it be better to just have RAM instead of this?
Not if you're going to go over 4 gig. You'd then need a 64 bit solution for that... If, on the other hand, you could add MAJOR amounts of cache to your hard drive, it wouldn't matter if you only had 4 gig of actual memory. You could run IA-32 as long as you like. You could have potentially gigantic databases without worrying as much about disk thrashing...
It's going to be potentially a long time until a 64 bit winner in the PC world is declared... As time goes on, something like this may actually be viable. And as memory prices go down, we're going to be seeing a lot more 4 gig systems around...
Or am I on crack?
/^[A-Z0-9._%+-]+@[A-Z0-9.-]+\.[A-Z]{2,4}$/i
n.
:)
Most modern Fighter jets use ramdrives exclusively because the high-G manuvers, shaking, ect would cause a normal drives heads to go smashing into the platters. I know the B2-stealth has some, so does the F-16.
When you're the goverment, and you want reliable killing machines, 250k for a 250meg ram drive back in 1987 isn't even an afterthought. It's a requirement. Now that it's dirt cheap to produce memory, it can finally trickle down in price enough to have become feasable for a civilian to afford one.
I know originally the early drives were used to hold map data for the HUD, which was basically just the data gained from our digital elevation modeling sattelites launched in the 80's. Given the advances in 3d modeling and sattelite map aquisition, I would guess whatever system they're using today probably contains terrabytes of map data.
I'm sure the old green vector graphic huds have been replaced with something a lot better to utilize the new data too. I'd suspect that in the actual control area there must be a LCD display, much like what you see on a modern 737.
It must be one helluva video game for them pilots
Furthermore, RAM drives are really meant for servers. Such a server will most likely use a fast internal memory (like RAMBUS) and cheaper, slower & much larger SDRAM 100 for the RAM drive.
The Raven.
The Raven
I've had the pleasure of playing with RAM based storage devices at work. They're a lot of fun. The particular models we used were external, SCSI, memory based, with a battery backup and internal disk for long-term storage.
The nice thing is that they can drive as many reads/writes as you can get out of the SCSI channel. Seek time isn't a factor at all. We're talking disk performance through the roof here! There are also UltraSCSI and other types that'll go even faster. But the only real limitation is the link between the drive and the computer.
Still, the data rate doesn't approach what this PCI based solution has going. But what I use is a more 'enterprise' solution. You've got internal battery backup and disk backup. If the unit is off the mains, the battery kicks in. After 30 minutes, it stops all IO, dumps to disk, and shuts off. When power is restored, it loads back in from disk.
This PCI solution is way cheaper than the external drives that I've used. Just it doesn't appear to have the reliability... or the commodity standards that'll let it hook up to just about any type of system. Still, quite neat!
While profiling a high-volume qmail server with fast mirrored drives, I noticed that I could get at least an order of magnitude sustained mail throughput by eliminating the fsync() system call, which essentially forces the disk subsystem to stop whatever else its doing and get a few specific blocks all the way onto disk. You can't run it in production this way, as the SMTP RFC specifies that the mail must be actually on disk before the server can claim that its done.
The problem is that magnetic-media drives can only seek a few hundred times per second. Regardless of their claimed sustained throughput, if you are writing a bunch of small files to disk, you are completely dependent on the seek time of the drive.
But mounting a magnetic-media-based ext3 with data=journal and the journal on an NVRAM block device would essentially use this as a trusted write-cache. Linux will return from the fsync() system call as soon as the data is in the journal, which could happen instantly on an NVRAM disk as there is no seek time. It then reads from the journal in its spare time, sorts it to minimize seeks, and writes the data out to disk.
I suspect that this should offer roughly the same speed as eliminating the fsync()s entirely.
I was looking into ordering a similar product to test this. I found:
Aaron
Putting your swapfile on this doesn't make much sense. You'd be buying memory, putting it in a special card that makes the memory act like a hard drive, then making that hard drive act like memory. It would be cheaper to just buy more regular memory.
Tim
Omnia vestra castrorum habetur nobis.
First off, run ext3 and put all of journal log file on. Poof, now you got a disk that has no latency pently for syncing the data. It will reliable be there when you reboot, so if you crash the log file is still there. You know all those benchmarks that Moshe Bar does where he turns off fsync() so he can push the CPU and memory to it's limits instead of the disk. He wouldn't have to do that so much any more.
Some Oracle DBA's would trade their soul to get Oracle's transaction logs written to something like this. A drive that has no latency is very, very good. No it's not as fast as RAM because it's behind a PCI bus, but in a lot of ways, no latecy permanent storage is the holy grail to a lot of problems.
If filesystems and OS's supported this, it's like getting a very flexible configuration for very high end SCSI cards. You know those really highend SCSI cards that have battery backed up RAM in them? The ones that sit behind that pokey SCSI bus? By putting the no latecy storage out in a place where you can get your hands on it w/ OS tools, you can custom configure it just the way you need it. You can upgrade it, you can add more. You can do a lot of things with this, that are much more flexible then any SCSI card will let you set up.
No they aren't the end all be all of permanent storage, however they have very specific usages, in specific high end situations that make them extremely valuable. If this one doesn't have an internal battery that can hold it's contents for say 400 hours without power, I'm not terribly interested, but as soon as it can do that without external power, I'd pay for it in a heartbeat for the Database servers and the high performance filesystems we run at work. A number of the ext3 people have talked deployment of devices like this will really improve the performance of a number of filesystems.
Kirby
This is potentially interesting, but there are several problems:
1. The pricing model sucks. The entry price of $399 is too high for a card with one chip and four DIMM sockets. And that only supports 512MB. To go to 1GB, 2GB or 4GB you have to pay hundreds of dollars more - even though the only change is a BIOS setting.
2. The RAM pricing is absurd. These guys need a reality check, pronto.
3. The board takes standard PC133 NON-PARITY RAM. No way in Hell would I trust my data to something like that. Honestly, this is just plain stupid. The board is too expensive for the home market and no-one sane is going to put non-ECC memory in a server.
[As a side point: Even using standard DIMMs, you could do some sort of block-ECC at the driver level (or in the controller chip) and use the fourth DIMM as a parity device to recover from on failures, like RAID-3. Alternately, you could treat each DIMM as a 48-bit device and use the remainder for ECC and Chipkill. There's nothing on Cenatek's site to suggest they do anything like this, though.]
There's a few other things that annoy me: the lack of specifications (while they have a list of approved memory modules, they point-blank refuse to provide the required memory specs on their support forum). Also, the board appears to require four identical DIMMs, which is a royal pain in the bum. Expandability? What's that? Low entry cost? Don't got one of those either.
So this board appears to be worthless for its target market and overpriced for anything else.
One-word review: Sucks.
Score: 3/10.
Memory is absurdly cheap, and a properly thought out board (even one that implements the ECC in software) at the right price has a market waiting. I know a lot of people doing embedded Telco apps would love something like this. This card isn't it.
The Platypus card is also over-priced, but it does support (indeed, requires) ECC, and also goes up to 8GB.
I don't know why a hobbyist website is reviewing this unit - the target market is not for hard-core gamerz or other home user types. Solid state disks are primarily used as database accelerators. Although the throughput of a solid state disk like this can be easily beat by a reasonably small raid, it takes a much larger raid to beat the io/s rate. If you have indices that don't fit in ram, you stick them on the solid-state disk and watch your database speed up by an order of magnitude.
Alternatively, as at least one other poster has already mentioned - if you use a journalling filesystem like ext3 or rieserfs, then putting the journal on a seperate solid-state disk is a huge performance gain.
When information is power, privacy is freedom.
I don't get it.. let's assume, in the name of generosity, that U160 actually pushes 160mb constantly. Then the rocket drive pushes 320mb constantly. Now we've got PC3200 ddr ram that can theoretically push 3.2 gigs per second, while this rocket drive only does a tenth of that.
What's the damn point ? It has to go through some sort of slow-ass bus anyways.. it's not like it has precisely timed local access to the northbridge, like traditional ram.
It would be better for motherboard (and chipset) makers to accomodate buckets of ram sticks, so we can allocate a real RAM disk using 8 or 16 sticks of 512mb/1gb ddr.
-Billco, Fnarg.com