Slashdot Mirror
RAID for Zero-G?
Cujo asks: "In all seriousness, I need a RAID that supports at least level 3 and stores > 500 GB, and I need to it work in zero-G (but not in a vacuum), and be able to take a fair bit of vibration and noise when turned off. I don't want to spend huge sums: I'm thinking well less than $50,000. I've looked at Apple's XServe/XRaid products, and they look great (about $10,000), but are they rugged enough and who is their competition? Some people make hardened RAIDs for military use, but I'm unfamiliar with the best candidates in that field (and do I really need mil spec?)."
How do you plan on getting this equipment into a zero-g enviornment? That will problable determine if you need hardened/milspec type equipment. If its going up on the shuttle, with those G forces on it during launch, then yeah, you probably do need 'milspec'. If its going to reside in a plane, that does zero-g free-fall testing, you can probably get away with something less... YMMV.
"The natural progress of things is for liberty to yield and government to gain ground." - Thomas Jefferson
They do use ThinkPads on the shuttle/ISS after all, so they must have the drives capable of this kind of thing. RAID cabinets and controllers have no moving parts (or maybe a fan or two), so I doubt they would be affected by zero G anyway.
UNIX? They're not even circumcised! Savages!
Hey just because NASA has a tight budget doesn't mean you guys can use Slashdot for your R&D!!!
__________
Love conquers all... except CANCER
Remember that a lot of mac kit is specifically designed to use convection to move air and therefore heat through the box. For example an old-style iMac will probably melt as it relies heat rising. Not something that is gonna happen in zero-g. You might need to be changing the fans and such like.
China has announced it intends to accelerate its Mars program, using experience and expertise from its fledgling lunar program.
They didn't mention China's use of highly reputable sources of expertise, such as Ask Slashdot!
How are you going to keep them down on the farm once they've seen Karl Hungus?
The heat is a concern, and the lower the power dissipation, the better.
The RAID performance in orbit doesn't need to be top drawer, but when it returns to Earth, I want it to perform well and not be a hassle to administer or set up, since there's a lot of data to analyze.
Helium balloons want to be free.
- Heat
- Air pressure
- Radiation
Testing tip:Convection cooling gets assumed into almost everything, so you'll have to make sure the gear gets some air forced over everything to keep it cool. Inside the hard drives, you've got those nice platters pushing the air for you, so that should be ok.
You indicate that there will be air, but not the pressure. You should test your system at the operating atmosphere and pressure for an extended amount of time. This is critical because the hard drives typically float on a cushion of the ambient atmosphere.
Since you're outside the 50 or so miles of air which filters out most of the radation common in space, make sure you have hardware ECC RAM, etc. It would also be good to make sure there is a hardware watchdog in place to protect the OS from hanging do to an induced CPU error.
I'd suggest you test the unit, then run the same test with the unit operating upside down, and on each of it's other 4 faces, as a minimum.
You've got an interesting project, good luck!
--Mike--
Please, for the love of God and all that is good, do not use an Apple product. We'll never hear the end of it.
compact flash memory is not very expansive (~200$ for 1GB retail)
it is pin to pin compatible with IDE so you can build a standard linux raid
if you buy bulk I think it will be in your price range
As of Postgres v6.2, time travel is no longer supported.
Boy, those budget cuts at NASA must be getting bad if they're coming to us for advice!
"Lawyers are for sucks."
- Doug McKenzie
For anything operating on the shuttle, you're gonna have to consider heat dissipation (convection cooling won't work!), outgassing properties (closed environment), vibration and mounting (not so much how the drives are affected, but how the drives affect everything around them), and gyroscopic forces, (There may be real issues with mounting a rack of 10k drives with all spindles on the same axis), size, weight, and power consumption, just for a start. You really need to provide a more complete spec to get recommendation,
What's your experiment budget? If you have the option of going solid-state (i.e. flash), that may simplify things - you mentioned write performance was not critical. You clearly want to use the largest, slowest (rotationally) disks possible to minimize space and power consumption. Perhaps a hardware ATA or SATA raid controller in a chassis with e.g. 8 180-250gig drives in a 0+1 configuration?
-Isaac
I am not a lawyer, and this is not legal advice. For Entertainment Purposes Only.
I have worked with these guys before and they are a great group. They have a lot of experience building rugged mass-storage solutions for airborne and military applications. In addition, they are a relatively small company, with a lot of engineering capability, so they should be able to give you personal attention and help you work through the various issues involved in this type of system.
I've heard it costs about US$10K per pound to put an object into orbit.
If that's true, why isn't weight more of a consideration?
I presume your project's individual cost limit is preventing you from investigating solid state disk solutions, which would probably be less susceptible to shock than platters in a magnetic disk hard drive.
"Provided by the management for your protection."
Paul, hello, I've just sent an e-mail from my work address. I've got some contacts for you in the Balitimore area. If you would like a sales rep to call, please reply with your contact information via e-mail and I'll pass the info right along, and pass you some names / numbers / e-mail addresses as well.
Michael C. Hollinger
IIRC, Harddrives are already vacuum sealed.. so it shouldn't matter if the surroundings have gravity or not.
The vibration is more of an issue; however, if the drives are parked.. it shouldn't matter too much.. I mean, they *do* go through UPS and the USPS often enough without too much damage.
You are sending up an experiment on shuttle mid-deck which I would hope implies that the experiment is worth the significant risk to human life and great expense that this requires.
But no, you are looking to cheap-out on the drives that are undoubtedly critical to the success of the experiment. That's pretty damn penny wise and pound foolish.
Having vented...
I suspect that NASA has specs and requirements for experiments on spacecraft if not for protecting the integrity of the experiments then at least to protect the astronauts and shuttle. What do those specs dictate?
If you can use COTS hardware then I suspect that laptop drives are your best bet for not only ruggedness but also weight, power draw, and size.
~~~~~~~
"You are not remembered for doing what is expected of you." - Atul Chitnis
Which brings up a few other questions besides technical limits of hard drives in space...
On airplanes they want electronics shut off during takeoff and landing. I would expect NASA to be no less stringent about 'spurious radiation' during takeoff and reentry, though probably more technical and perhaps more flexible if you shield carefully.
I also wonder what they think about 'little embedded gyroscopes' (hard drives) on the shuttle. Do they have to know about every one so they can account for it, are they just negligible, are they cumulative, or can you mount every other drive in the rack upside-down to cancel out righ-hand-rule effects?
There used to be a "Space Shuttle Operators Manual" for publicity/potential customers. I believe the local library has one, so I should check out the 'electronics restrictions' section.
The living have better things to do than to continue hating the dead.
I want to know where you're getting Zero-G from!!!
; -- the corruption of government starts with its secrets. a truly free people keep no secrets. --
You are asking for a zero G drive system. Zero G is no gravitational forces... so what is the problem?
If a drive can be mounted in ANY direction, it can take 1 G in any direction, and certainly zero Gs which is easier than the 1Gs most drives experience. If youre asking for zero Gs, youre launching the drives into space arent you? If so, shouldnt you be worried about the Gs during takeoff??? So shouldnt you be asking about say 100Gs while off, instead of 0 Gs which ANY working drive in the world can take?
The sames true of RAID hardware or system with software raid. Generally ciruit boards can take as many Gs as its fibreglass structure can take (assuming the heatsinks and capacitors arent too big. Make sure you screw every hole in the motherboard down and the case should be rigid enough not to bend, but not rigid as the Titanic.
And like another poster commented, you should be worried about the radiation levels and be making sure theres about 1 atmospheric pressure on all components unless you want exploding capacitors. Worry also about the temperature and make sure theres no humidity in the chamber and theres good insulation all around. I would strongly suggest you go with a system-on-chip with flash drives unless you need space because you want to boot windows 2003 with sql 2000. Worry also about vibrations during takeoff unless your zero Gs are in a 747 in a nosedive.
"Give orange me give eat orange me eat orange give me eat orange give me you." -Nim Chimpsky
Have you looked at a solid state solution. Not really cost effective, but you can certainly guarantee that any launch/re-entry gee's will be catered for.
sic transit biscuitus
If you want RAID-3 you pretty much have to go SCSI. There may be a way to do it with ATA drives, but I haven't heard of it.
The other reason you want SCSI is reliability. That's one of the reasons SCSI drives are so much more expensive. I've seen more than one SCSI drive get dropped on a hard tile floor and still be usable for a year or more (These are half hieght Seagate and IBM, 7200 or 10k RPM, YMMV).
If you do decide to go IDE, try to use laptop drives. They have MUCH better g-force tolerance than the standard 3.5 inch IDE drives. However, I've still never seen one survive getting dropped on a hard tile floor. Shock and vibration are different things, though, so the laptop drives still may be a better choice. You can
You could go flash, and that would take care of the vibration/shock issue, but at 1GB each that's an assload of IDE controllers you have to somehow get working together. Assuming 4 per controller, that's still 125 controllers. Even if you solved the IRQ problem, where would you put all of them? Space is a precious comodity on these missions. Plus at $200 each that means $100k for 500GB, which seems to be out of your budget range. A custom motherboard with 125 PCI slots is certainly out of your budget range.
What I would do is talk to standard RAID vendors like EMC^2 or Ciprico and see what they've got. I know a company that would be happy to design and build a shock-mount for a standard raid chassis for you for probably under $10k. You could also go somewhere like Musicians friend and buy a road case, which will certainly have some anti-shock measures, for a few hundred dollars if your needs won't be too severe.
I very much doubt that zero-g will be an issue at all. The things that will be problems have already been mentioned by other posts.
Under capitalism man exploits man. Under communism it's the other way around.
If you don't need random access, tape drives might be more robust. You can still do a RAID of them if you wish.
Why is everyone assuming that zero-G implies a space shuttle ? Its terribly unlikely that NASA tech's are onto slashdot- are they ??
I guess he needs it for something simpler - such as skydiving- or anything that involves free falling - maybe a laboratory experiment - something to do with air/fluid flow perhaps ?
be that as it may, I concur that standard IDE's should do for 0g, but of course - u need to investigate u r requirements better.
Your helping the marxist dictator state of China by answering their questions. Don't.
It's not like it's going to put any stress on them :) Fluid or ball bearings should work fine.
You will have to pay carefull attention to vibration and shock considerations though. Are you going to be accelerating quickly? If so, you will need some kind of packing that can handle a lot of displacement. For vibration, I would personally think it would be easiers to isolate each drive from the chasis using some kind of foam or something, and also cushion the rack. It would require much more work to get the rack isolated alone. As long as you don't have any high frequency vibrations getting to the parts, you should be good since the acceleration will be very low.
sig is false in this case
I"m thinking redundant arrays of compact flash, microdrives, or laptop drives. i.e., get about 20 laptop drives, build like 5 mirrored arrays out of it, so that even if half your drives fail, you still got your data...is that a good idea??
Let's see, you have been entrusted with finding RAID storage that is going to be floating around in space, or perhaps mounted in diving or falling airborne devices, and you are unfamiliar (scary in itself) with what's out there, and with this kind of risky work you are reaching out for help from... Slashdot?
Let me just say, YIKES!
ok..not the shuttle but a rocket (far more hostile than the shuttle i'd imagine). experiment worked for 3 days, data returned safely. equipment :
2 x 533Mhz Alpha 21264As (164LX boards) with 1Gb RAM (ECC) each, RAID-5 using ICP-Vortex boards with 128MB ECC cache RAM each and 7 x Maxtor 120GB HDDs with hot swap PSUs. systems mirrored each other, so there was 400GB of usable space (roughly, 2 hot space + 1 checksum drive). total cost was around $10K including custom parts (boxes, power distribution, batteries, etc). All usable space in the inside of the boxes and RAID towers was filled with DuoFoam (Its the spray can thingy which expands into rubbery yellow foam in contact with air. Systems were shutdown on the ride up, switched on in freefall and shut down for the ride down.
2 hard drives failed during the entire trip, one of the alpha CPUs has its heatsink warped by heat and some foam had melted off. all the data was ok, though.
How about a RAID of IBM/Hitachi Microdrives? They're IDE compatible with the right cabling, use less power than any other rotating storage, and are super shockproof. Of course, given their size, they weigh very little relative to other rotating solutions.
The only problem I can see is that they're only good for 1GB each right now, but 4GB is coming soon.
Of course, you'll need some kind of off the shelf motherboard and some IDE controllers - but you can GLUE that together to avoid stuff shaking out of slots. But I guess you'd have the same issues with any RAID solution.
Jonathan
Level 5 RAID has both high performance and high reliability (striping with distributed parity), at the expense of less capacity. I don't think you can do it in software, though, so you need a hardware RAID controller (prob. not a big expense). One thing I like about the Apple controller is that it's dual redundant.
I'm not working directly with NASA (don't ask. NO - don't.), but I don't believe anyone's flown a RAID before. Alternatives ARE being considered, but I speculate that /. folks have more RAID experience than all of NASA put together. They have hardrives on the laptops they have used in the past to support Getaway Specials and other experiments, but these are not up to speed for the application.
Helium balloons want to be free.
yea ... go with one of those xserves. that'll be like $10k. then, pay me the other $40k and i'll hold it for you
vodka, straight up, thank you!
With all the work done on the Deep Space Network, I would hope you find a better way to store data. I'd probably design a DSN listner to store data on a large, fast RAID on Earth and have (if any) a RAID on the Shuttle that would be a "backup device" of sorts where data that did not make it through the DSN gets corrected/validated upon shuttle return.
The Xraid would likely be tought enough, but sadly, the Xserve itself my not be able to handle launch. It's so thin, that many users have bent the tope covers, or had the entire unit warp on them.
Cooling the units shouldn't be an issue, the fans are powerful enough that in null gravity they'd be able to propel the unit:)
And, i know that at least one company was planning on using G4's in space, so there should be some studies on radiation effects and such on those processors floating around somewhere....
I doubt it's an issue any more, I jsut remember it so well because I thought at the time it was interesting. I thought I might find an article mentioning it, but no luck. I did, howewver, find this little bit...
IBM announced that NASA had successfully used its one-inch diameter 1 GB Microdrive to store digital snapshots of deep space and bring them on home to Earth, during recent Atlantis and Discovery shuttle missions. The drive costs $499, the shuttles cost billions.
Not exactly 500GB, but it appears they are already well tested in your application. 500 drives x $500 ain't exactly cheap, but perhaps you could get a sponsorship agreement and an opportunity to employ their more recent, higher capacity drives.
It's ALL ball-bearings these days..
Everybody dies frustrated and sad and that is beautiful
Huh? That's strange...mine isn't. Mine are fluid..and BOY can you tell the difference...sooooo quiet. Have fluid bearings in my laptop(at home) and desktop (at work).
Go watch Fletch...
Everybody dies frustrated and sad and that is beautiful
Comment removed based on user account deletion
The chokepoint is the downlink, and the SHUTTLE deosn't really use DSN per se. Anyway, DSN generally tapes all the data they get (without unpacking it), but passes it along ASAP to its destination. There the responsibility for efficient archiving and retrieval lies.
Helium balloons want to be free.
Although it's from AC, sounds serious to me.
In Soviet Russia, articles before post read *you*!
Otoh, non-rotating disk memory is fast and only uses a disk during startup and shutdown (whether failure or otherwise). Normally such solid state disks have ECC and so on and they are much faster than conventional disks. You can also RAID them for additional reliability.
See my journal, I write things there
I don't care what flag a human being brings to Mars, so long as one of us goes there. We've been squabbling down here long enough.
Their List of Drives by Heat Output shows the Maxtor DiamondMax 16 (160 GB ATA-133) to be the coolest at 13 degrees Celcius and the next coolest/largest coming in at 200Gb and 19.2 degrees C. I highly reccommend going in and doing a head to head comparison because they break the temp readings down even further. If you're not going solid state, StorageReview can held you make up your mind as to the best drives. (not affiliated w/SR even though i sound like their pimp)
[Fuck Beta]
o0t!
Years ago when I had some experience with shuttle operation there was always the issue of testing and certifying equipment as being safe. For example, that plastics don't outgass volitile compounds, things won't catch on fire, etc. Such certification can be expensive and involve a ton of paperwork. Say a $5 object requires $10,000 worth of assurance.
So unless you want to pay for this process, find equipment that has already been certified. Sorry but I can't help you here, but NASA probably can.
You need to reread your books. RAID 5 has basic reliability (can afford to lose a single drive), excellent read performance, dismal write performance and offers the highest usable/raw capacity rate of any RAID type (where the R is meaningful).
according to the Software-RAID howto RAID 5 in software is just peachy.
Need a Catering Connection
Apple's RAID is probably the best off-the-shelf product. Or if you want to DIY try:
seven or eight three-disk 80G RAID 1s using laptop drives, or
three or four three-disk 160G RAID 1s using SCSI drives.
Stripe them if the space has to be contiguous (ie: RAID 10).
Statistically speaking the less drives the less chance of failure, but laptop drives are rated for higher levels of shock resistance than desktop drives. Laptop drives also have the advantage of being physically smaller and weighing less.
How well you want to assure data safety is really a matter of personal taste. However, I imagine one extra drive for each mirror costs a lot less than having to do the whole thing again :).
You'd also want to consider things like radiation hardening for the case. I'm sure the people at NASA could help you with that.
You'd probably also want something a bit more performance oriented with less redundancy to transfer the data to once it gets back. RAID 5 would probably be tolerable, but again RAID 1 or 10 is preferred. Faster disks would mainly be the advantage here (7200 or 10k RPM IDE drives would most likely give the best bang per buck).
You can fit the hole half-a-terabyte (500 GB) in a single build to order G5. Plus they use SATA drives so they'll be fast, add that with the blazingly fast G5 DVD Backup and quiet 9 fan variable independent cooling in an aluminum enclosure running a rock solid UNIX based OS with an easy to use interface and you've got a winner!
To bad Beleaugerd Apple is dying. (tee hee)
Hard drive are totally not vaccum sealed. Most even have breathers. They're dust sealed (which is why there's that little snake looking thing leading from the breathing hole to the internals)
Hard drive may have had a vaccum back in your day but today that's utter crap.
You might mount every other drive upside down as to cancel out the torque from spinning up/down. That would eliminate some stress on the disks, and then you can allow the drives to float more freely, reducing the shock from thruster action.
-twb
I built a project that flew on the VomitComet 20+ times in the early 90's. One of the main goals of the project was to use off-the-shelf equipment. At the time I used a Connor hard drive. Powered it up after take-off, and power-down on landing. I accumulated nearly 2 hours 0G running the system with no problems at all. With today's drive technologies, you should have no problems. We just made sure that we used a lot of vibration absorbing material. Syntroxis ps: ball mice won't work in 0G, a trackball with a captive ball works great (there are -G's too, and I'm not sure about an optical mouse.
Wherever you go, there you are.
All bow to his Noodliness!! His Noodle Appendage has touched me!
I'm so glad I have a huge email archive
See the article here. Called "2001: A Space Laptop", they discuss what's involved in using computers in space. It's a little dated, and we can hope they're using better computers as the spec at that time (2000) was a IBM 166MHz Pentium MMX Thinkpad.
When you sympathize with stupidity, you start thinking like an idiot.
That's an easy enough one to solve. Install an even number of hard drives. Mount them in pairs. In each pair, ensure that the axes of rotation for the platters lie on the same line, but mount the hard drives facing opposite directions. The hard drives should have nearly-identical platters that spin at nearly the same rate and are well-balanced, and the platters will be spinning in opposite directions, so the gyroscopic forces will cancel each other out. (There may be other symmetrical arrangements that would solve the problem too, hmm...)
Not that it matters, necessarily, since presumably the equipment will be bolted down or otherwise mounted to the shuttle's structure. It's not like they just load everything in, take off, and then wait for it all to float around in a big mess...
Other free advice for the original poster: Do the RAID right. You can hire consultants cheap enough these days. Get an IT guy in there who has set up numerous real production servers that go in server rooms. Someone who has set up a fully-redundant fibre-channel system that spans multiple physical sites. Someone who has designed storage configurations for mission-critical databases whose downtime can cost hundreds of thousands per hour. In other words, hire a real IT consultant, not someone whose experience with RAID consists of plugging in some 7200 RPM drives into a hardware RAID controller they bought at Fry's and then being surprised when it dies on the first power outage.
Such a consultant should easily be able to come in and suggest stuff that will suit your needs. Presumably, that'll include ideas such as:
I think my point is, overall, you need someone who really has experience doing this kind of thing, because there are traps for the unwary. At least, if you want true reliability.
Hey, I remember in the extremely old day of computers the number of G forces the drive could take was clearly posted on the drive. For example, my old Fujitzu MPU drives had clearly on the label "Do not subject this device to more than 40G's."
I don't really thing the environment will harm things too much, just make sure you have this thing buckled down really well, and that you have it lead shielded (those little dark plastic bags work for electrostatic dissapation, would work great because the shuttle will be highly ionized coming through the atmosphere).
I think if I were designing this though, I'de go solid state. If anything, they run cooler and take up less space, not to mention weighing much less. Like someone posted earlier, it is pin compatiable with IDE, so just stick a bunch on a raid card and you are set.
"Victory means exit strategy, and it's important for the President to explain to us what the exit strategy is." G.W.Bush
Half of you examples tell me to buy IBM.
The problem with RAID5 is that the performance when degraded (due to a lost drive) is usually far worse than when the array is healthy. If you design your experiment to use the peak performance of your array then a failure will cause it to fail even though you won't lose any of the data you've gathered so far.
First off, this states that you explicitly need it not to work, i.e. "fail" in a vacuum. I doubt this is your intent. If you need something to fail in a vacuum, one idea may be to have a membrane holding back some sort of acid at air pressure, but which bursts and destroys your media at a sufficiently low pressure. If you merely don't care about failing in a vacuum, then leave that out of all future problem statements.
Secondly, to all the posters before me, it appears as though everyone assumes this person is going into space. While Cujo is from John Hopkins' Applied Physics Lab (looking at the e-mail address) which is apparently sending a craft to Mercury, it is possible that Cujo is working on a different project which is terrestrial. Perhaps a roller coaster type environment or sky diving type airplane bourne freefall. I'm not sure why you need 150GB of storage on a craft that will forever have a 9600 baud connection home (if even that fast)-- so I suspect that the poster has less lofty goals. But maybe 150GB is necessary because of the volume of data to record and the craft will return.
To combine RAID type reliability and resistance to vibration and probably some attention to power dissipation, solid state is a good possibility, especially if you have few write cycles. You can apparently get 512 MB of compact flash for $128 but someone might give you a discount if you buy 300 of them. If not, that brings you to less than $40k, which still leaves you with ~$12k to pay a student to integrate them into a smart FPGA based controller.
For pre-existing solid state solutions, google for SSD or solid state drives.
For writing, we don't need anywhere near peak performance. We estimate peak rates of 4.5 mB/sec, and average more around 0.5 mB/sec.
Helium balloons want to be free.
not really an issue. The shuttle flies at 400 km altitude, which is not a bad radiation environment, and the RAID would be inside a midddeck locker, which enjoys a fair bit of protection form radiation. That an the usual aluminum cases provide plenty of protection. Also, we'll be near solar minimum when we launch.
Helium balloons want to be free.