Slashdot Mirror
IBM sets another disk-drive world record
Anonymous Coward writes "IBM has set a new computer data storage world-record of 35.3 billion data bits per square inch on a magnetic hard disk -- a 75 percent increase over the 20-billion-bit milestone the company achieved less than five months ago. " The press release goes on to talk about that this is expected to lead to drives with three times the storage of today, quite soon. Just think - a MP3 server laptop!
The area that I see this technology being most useful is in reliable portable storage. In particular, IBM's microdrives (1" disks, not the old Sinclair endless tape loop!) could be turned into a very nice, very small RAID array. This was suggested when they were first announced, but each disk was limited to 170MB IIRC, which didn't lend itself to many practical uses. With this new technology, they're claiming over 2GB for a 1" disk -- 5 of those would give an 8GB RAID 5 array in a *very* small physical space.
"The invisible and the non-existent look very much alike." -- Delos B. McKown
Huge HD is nice.
Huge HD that gone is bad.
I used trust the manufacturers' MBTF number for their HD, but with more and more HD failures before their time, I do not know anymore if bigger and bigger disk is the answer to go.
Why not make disks that last longer, rather than bigger disks that go belly-up before its time?
Muchas Gracias, Señor Edward Snowden !
USB is 12 mbs, not 1.2.
USB 2 is 360-480 Mbs.
mlk
Wow, I should not post when knackered.
The 77 hours of music, as said at the bottom of the article, is only 3.2 days of music, not 6.5
Of course that's per "^2, your new HDD would be more than 1 square inch.
the pun is mightier than the sword
But seriously, is it just me, or does it seem like HD technology is actually outpacing the needs of the average desktop user for once? I can remember when a gig seemed like an unreachable depth, but at least it was conceivable. I'm still using a 10Gb HD, which I've never come close to filling. I don't want to sound "Bill Gates"esque here, but who's really going to need 50 Gb *per platter* any time soon? Of course, if I had the bandwidth to fill this monster up with dirty pictures, look out!
Anyway, let me know when they go on sale so I can buy one.
It's great that we could very easily have terabyte RAID5 arrays within, say, $10k soon. I'll be able to copy ALL my CDs up to my hard drives without converting them into mp3s.
BUT
until more people start buying tape backup units and driving down their price/gb, this is a scary trend.
One of two things need to happen, preferably both:
1) RAID 5 arrays need to get affordable, perhaps in the scenario one person posted about the microdrives. The ability to not have to worry because you have an affordable redundant drive waiting to be used is a tremendous thing. Mirroring is nice, but it just doesn't give that warm'n'fuzzy feeling.
Unfortunately, I don't see affordability happening anytime soon. RAID5 technology just hasn't saturated enough yet to get prices down. Also, the standards wars that are occurring w/r/t EIDE, SCSI, USB+, Firewire, etc. are hampering things, too.
Maybe IBM can come up with a 5 platter hard drive that has RAID as a function per-platter, with one platter set for failover.
2) Tape drives need to drop in price, and increase in storage capacity. A great deal. This has become apparent this week as I search for a backup solution at work. The best bargain is the Onstream unit, with SCSI 50g backup for about $550, with cartridges at about $50 (i think.) 50g won't make a dent in the drives that are coming soon. And the unit only has a 2mb/s transfer rate. The best tape drives are DLT, which just work, period, but the 35/70g units are way out of the reach of the average consumer at over $2k, with cartridges at about $90. I don't keep up with backup technology, but I can't imagine there are any streaming tape breakthroughs coming anytime soon. I think the fact that DLT units are still priced in the stratosphere points out that there's a lot of unprotected data out there.
Long term analysis: I think tape backup will die off, and RAID arrays will become ubiquitous. There's no other way to protect data easily. In the meantime, I continue to be scared.
stored on computers from birth to the grave
The slowest part of a disk subsystem is still the drives. Especially in the case of IDE, where drives have only just recently begun to broach the maximum theoretical speed of *Mode 4* IDE (ca 1993). SCSI is a bit different, since RAID is actually a real option and it's not hard to envisage attaching enough fast hard disks in a RAID to swamp a U2W controller.
However, the bottleneck is not really in the interfaces, it's in the drives. This is, as I said, especially true of IDE.
although the idea of more mp3 storage is nice we still have to remember other factors. what about transfer rates? i sort of doubt that this new technology is going to be run on ide/scsi. i think before we get huge hard drives developed a cheaper-than-scsi, yet very fast transfer method needs to be developed.
I fear the size of apps to come.
Has anyone else out there had to install, say for instance, M$ Office 2000? It is huge, and for no apparent benefit in function or ease of use. Linux can fit an entire server and all the required apps on one CD(usually), but M$ Frontpage 2000 comes on two discs by itself!
Remember before you had a HDD? Huge programs then ran on two, sometimes three, floppies. Remember the 640Kb RAM cap? Some very impressive software ran in a very limited space. All of these limits hae been effectively removed, allowing the mind-bending size of modern applications.
I understand that the prices of RAM and HD space or plummeting, but does that require more crappy software to fill the void? Wouldn't you rather have 50Gb worth of useful stuff?
Larger drives will inevitably lead to larger programs, but I'm not sure I want them.
Computers can only simulate determinism. ~Hermetic.
UDMA/66 won't put SCSI under, I'm not familiar with UDMA/66 but if it's just got a faster bus... well big whoop. SCSI is meant to be able to do multiple operations simultaneously, hang lots of drives off of one controller, etc. Until something other than SCSI/fibre channel can allow me to hang 15 devices off of it, be able to do multiple operations simultaneously, allow me to share drives (I can hook-up 2 boxes to the same drive using differential controllers, for HA use only). Try hanging a terabyte of UDMA storage off of one box (we've got something like 20 controllers in one of our monster SGI boxes). If anything is going to put down SCSI, fibrechannel will (not in the near future, but sometime), SAN's are really sweet. I can hang 150+ devices onto one SAN and gigabit FC-AL is almost finalized...
You completely lost me in your logic of USB... I don't know anyone of connecting phones, speakers, scanners, etc. to SCSI. SCSI is meant for drives, USB for periph. What you said is like saying my monitor is my visual connector of choice over printing everything on my printer (move mouse, print, move mouse, print).
SCSI u2w - 80 megs/sec
SCSI u3w - 160 megs/sec
sure you take a u2w hard drive and you wont get 80 megs/sec, same with a UDMA/66 drive. point is the bus can handle it, so you could have say 4 20 meg/sec u2w drives before the controller becomes a bottleneck.
Also, dont forget the whole 15 devices a u2w card can support, compared to what, 2-4 for UDMA/66?
I wont begin to get into RAID implementations.
Drives with this sort of capacities, coupled with our new lovely broadband connections, could truely mean the start for piracy of movies. Even with the crappy quality VCDs, 1 gigabyte a piece is to much to keep any collection worth noting, even on modern pcs with 24-36 gigs of space.
When I bought my Celeron based PC last year, my old pentium got to move to the bookcase where it became my resident mp3 player (gotta love using crontab as an alarmclock). I know I'm not the only one who has done this.
The logical extension is that when my current computer gets tossed out for a Merce.. I mean Itanium, Opteon, K8 or whatever, my current PC will move into the TV room to feed movies and generally do everything that those TV-cache machines do (only a whole lot better).
With these sort of drives, that would seem very likely.
-
You see I'm not in the IT industry anymore, I was for a few years, but to tell you the truth, it kind of sucks, and I stopped enjoying using my computer(s) at home. At the time, I had built a P233 with an Adaptec SCSI controller, a SCSI CD-ROM 4x, and a 2.0 gig SCSI drive. All of this is now dated technology, and it is not upgradeable to my next system. How much did I spend on this fiasco? $150 for the controller (plain 40 Mb/sec SCSI) $200 for the drive, $200 for the CD ROM. That's a total of $550 for a bunch of equipment that is not worth $100 only 2 short years later... Now had I been smart and opted for IDE, I would have spent a lot less, gotten more storage space for my money, and felt a lot less foolish now that I had wasted all that money on equipment that I can not transfer to the new workstation I am now building.
Sure if you're some hot shot WAN admin, you're used to throwing around someone else's cash on a daily basis, and for you buying massive Ultra SCSI 3 RAID arrays is no problem. You know that the whole thing will be upgraded in a year or 2 on the companies dime and written off as a business expense -- money that would have gone to taxes anyway.
But you see, 99.999999999999% of Americans aren't big shot sys admins and therefore we don't have million dollar annual technology budgets. Instead, in this increasingly annoying cycle of hardware obsolescence, we are forced to make wise, cost effective decisions and if you take my example of my experience with SCSI in the previous paragraph, you might come to the same conclusion I did: that SCSI is nice, but not worth the money especially when you can get comparable performance at %300 lower cost. I mean come on, there's really not all that much difference between 66 Mb/sec and 80 Mb/sec.
So no, SCSI won't die tomorrow, but it will get more expensive as more and more average consumers turn away from it to more cost effective alternatives like UDMA 66 (the less units you sell = smaller production cycles = higher unit cost).
You completely lost me in your logic of USB... I don't know anyone of connecting phones, speakers, scanners, etc. to SCSI.
Hrm, let's see what peripheral SCSI devices are on sale today at www.pricewatch.com...
1) SCSI Scanners
2) SCSI JAZZ Drives 1 & 2 gig
3) SCSI Magneto Optical Drives
Then here's a comparable list of USB peripherals on Pricewatch:
1) USB keyboards
2) USB Mouse's
3) USB game controllers (joysticks, wheels, etc)
4) USB hubs
5) USB scanners
6) USB Printers
7) USB digital cameras
8) USB modems
9) USB Network cards
Granted there nobody's buying SCSI peripherals anymore, however this isn't due to the lack of products, because before USB the only peripherial connectors offered were SCSI, Serial & Parallel. So why are there so few peripheral SCSI devices today, hmmm? Because everyone's now buying USB peripherials. You know what's really sad? I went to Pinnacle's web site recently, and tried to click on a link to their MO Jukebox section, and it was a dead link. I mean they can't be selling any MO drives anymore if their web site doesn't even link to one of their core product lines. It's sad, but you know what, I trust the decisions of the market more than any individual. The market likes USB, because it's integrated and you don't need IRQ settings, the speed is negligible as most people are used to slow parallel and serial connected peripherals anyhow.
So what's left? SCSI is only useful now, as you said yourself for hard drives. But what is the wholesale price of a Ultra SCSI 3 drive? How bout a Compaq 34.4 gig SCSI3 drive? $1550. Now why in the hell would I buy one of these when I can buy more storage for $3-400 between two 20 gig UDMA 66 drives? The price difference may be small to a corporation, but to the average Joe, that's a big screen TV. Hence, SCSI is pricing itself out of the marketplace, and will therefore eventually wither and die.
IMHO the coup de gras will delivered to SCSI when solid state RAM drive prices drop to the point where they are a cost effective solution.
SCSI is kind of a pain, and very expensive. A high capacity SCSI drive can often be 4x as expensive as a UDMA/66 drive, not to mention the hugely expensive SCSI2 controllers...
Additionally, USB is rapidly replacing SCSI as the peripherial connector of choice, so IMHO SCSI is a dieing technology.
I'd be interested to see when IBM makes this new technology commerically available, and whether or not the higher density will reduce seek times/rpm significantly.