Slashdot Mirror
Hitachi Announces 400GB Hard Drive
jkcity writes "Hitachi Global Storage Technologies has announced their new 400GB 3.5-inch ATA hard drive, which they claim makes them the new capacity king. Specs on the drive are also available."
← Back to Stories (view on slashdot.org)
huh?
1) hitachi has always made the fastest drives
2) the old IBM HDD division has an excellent
track record. They just f'd up with the
hungary drives.
No, ata 133 is a scam.
A standard 7200 RPM drive generally maxes at a little over 66MB/s (ATA100s just barly needed) (and cause its parallel, it can't share bandwidth).
Note that WD and seagate don't use it.
The hype about SATA is not 150MB/s, but that its serial and doesn't ahve any master/slave nonsense
AFAIK the acronym RAID stands for Redundant Array of Inexpensive Disks. And I guess at the moment such a drive is not what I would call "inexpensive". YMMV.
There are 5 80gb platters in this harddisk. They're just putting more of what makes a normal harddisk into it. I don't think that's a good idea: The result is probably heavier and more mechanically fragile than most harddisks. In my experience, disks with more platters fail sooner than disks with only one or two platters.
The "I" in RAID stands for "inexpensive". Part of the idea behind RAID is you can create a 400GB "drive" using 4 100 GB drives, which should work out cheaper. (ignoring the cost of the RAID controller...)
Nowadays many hardware vendors seem to agree that RAID really stands for "Redundant Array of Independent Disks". It dosen't seem quite as crappy with independent disks as with inexpensive disks it seems...
What you describe is RAID 0 (stripping), if i'm not mistaken. You have different levels of RAID, and the kind you describe does not provide data security; you have to mirror to protect your data from the failure of a drive. With 4x100GB, you could do RAID 0+1, for example, that is stripping+mirroring (2x100GB x2, you'll have 200GB space available and data security).
See http://www.acnc.com/04_01_00.html
blah
Because of the "deathstar problem" they are outsourcing inspection and final testing of the drives to a different company now.
Take the cheese to sickbay, the doctor should see it as soon as possible - B'Elanna Torres, "Learning Curve"
But it *can* transfer from the buffer at that speed, so if the buffer is reasonably large (say 8MB) then ATA-133 should really be faster in some circumstances.
When I am king, you will be first against the wall.
If you create a 400GB array using 4x100GB drives, you end up with one-quarter the reliability of a single 100GB drive. This is RAID-0, which doesn't really fulfill the "R" bit of RAID.
You need something like 3x200GB drives to make a RAID-4 or RAID-5 array of 400GB, which can withstand a single drive failure.
Ydco co
Usually a lot more data is in the memory cache (usually around 400MB for me). So the odds that data would be in the HD cache and not in the memory cache are pretty low, which makes big HD caches useless. /Erik
Erik Dalén
>>The "I" in RAID stands for "inexpensive".
>Umm, no, it doesn't. It stands for "Independent".
I believe you are BOTH right. As I recall, the "I" in RAID *originally* stood for "inexpensive" back in the days when the rapidly dropping price of 5.25" and 3.5" drives were making them very attractive "inexpensive" replacements for larger, *very* expensive mass storage systems. But time passed and the success of RAID arrays made them the primary method for providing high performance data storage and retrival as well as data redundancy. They became the new standard for comparison, so the term "inexpensive" was no longer relevant and was replaced with the word "independent," a term that better describes them. As I was typing this I found this link that seems to agree with my recollection.
If you read the article, you'd notice it says: 45 hours of HDTV broadcast
In case you still don't get it, that's 45 hours of HIGH QUALITY PORN. I mean, that's almost as good as the real thing, right?
In times like these, it is helpful to remember that there have always been times like these. - Paul Harvey
>With 4x100GB, you could do RAID 0+1, for example, that is stripping+mirroring (2x100GB x2, you'll have 200GB space available and data security).
OR you could do RAID 5, have striping and rotating parity, have 300GB of available space and be protected against a single drive failure. Of course, always match your RAID configuration to your specific data requirements, as each RAID configuration offers different trade-offs between usable storage space, read/write performance, data security and cost. YMMV.
Ok, let's do the numbers. A big 8 meg cache will take 8/100, or .08 seconds with ATA-100. It will take 8/133 or about .06 seconds with ATA-133. That's a difference of a whopping .02 seconds. After that huge savings of .02 seconds you're back to the disk bottleneck of 45 MB/sec.
Yah, ATA-100 is just so much faster than ATA-133. In fact if you're only using one drive you can get away with only ATA-66 and interface still isn't a bottleneck. A big on-drive cache is to maximize the throughput of the drive so it can stream data as fast as possible to system memory. It's not really a very good disk cache, as system memory is a better, and larger place for that.
AccountKiller
In case you haven't noticed the drives on your laptop are typically 2.5in wide and much much thinner than desktop drives. Why is this important? Well the limiting factor in HD size is the aureal density of the platter (bits / area). This is currently limited to around 60 billion bits / sq in. So if you want smaller drives to fit in your small laptop then you'll have to live with lower capacities. The platters being used in both 3.5 in desktop drives and 2.5in laptop drives have the *same* aureal density so I'm not quite sure what technology you're waiting for to trickle down. I think what you're really waiting for is a Chiropractor's dream 10 lb laptop not for technology to trickle down. Of course bigger disks tends to mean more momentum and thus high power dissapation so the troubles of stuffing bigger drives into laptops just continues to mount.
And btw you can get 60 GB drives for laptops, that's a considerable amount of space. If you want more get an external USB storage cage or something similar.
The specs say that the MTBF is 1 in 10E14. I suspect that the reason why it suggests low I/O applications is that the amount of data stored is huge per disk, and if you need to shift 400GB regularly, a RAID of smaller drives would work better.
I appear to have a blog. Odd.
What guys are you doing with so huge hard drivers?
Me? Two words : digital video.
The mini-dv cartridges I use hold 1 hour of high quality footage and only cost $10 or so. That is unfortunately 13GB when imported into my laptop. When I finish a project, I dump to an external Maxtor 250GB drive I bought six months ago for about $300.
These capacities make home movies more affordable than ever, it's great.
The whole point of RAID is that the disks are closly dependednt on each other
Really? What about RAID 1? Mirrored disks are in no way dependent on each other. You can do a little learnin' here.
I want to drag this out as long as possible. Bring me my protractor.
Also, gigabit ether in practice maxes out around 400Mbps, i.e. 50 MB/sec. Thats substantially less than even an ATA/66 7200rpm drive.
Of course, gigabit switches are basically unheard of outside of very large networks, so unless you're using crossover cable you're still limited to 100Mbps, which practice gives you about 10MB/s (due to overhead). And many of us are on wireless networks, which will give you even less throughput.
Networking technology still has a ways to go before the disk will really be the bottleneck in such scenarios.
The extra 600mA is probably for the onboard SATA to ATA bridge chip. Many drives still use these.
There can be some issues with the bigger drives. I just got a 200 Gig hard drive and it turns out that the default Debian installer won't work on it. Apparently kernels before 2.4.19 can't recognize drives bigger than 137 gigs. (Not this drive anyway). I had to install Debian through Knoppix. Even Windows XP won't recognize it unless you've got SP1.
I'm a writer. You can generally count on 2k per page. The current book I'm working on is already over 120k and it's only 67 pages so far. Extending this theory to a 300-500 page book, you can expect 600k to one meg.
Of course, this is still way under their inflated figures, but a book is hardly 100k.
Read: Rabbit Rue - Free serial nove
You can read here to find out about RAID levels 2 thru 4 (they aren't used much because RAID 5 is superior). RAID 10 is a combination of striping (RAID 0) and mirroring (RAID 1). Because of the mirroring, RAID 10 can lose a disk without losing data. You'll also find mentions of RAID 50, 51, and 15. These are combinations of RAID 5 with striping or mirroring. It is left as an exercise to the reader to determine disk independence.
--- Often in error; never in doubt!
to its integrated LCD. Also, all notebook LCDs are 'hampered' by an 18-bit color limitation. Tech tricks and better manufacturing help mask the noticability of the panel's reduced color output, but it is a well-known 'secret' of the industry.
Dell's new XPS notebook offers a DVI output, and methinks Apple has offered DVI out on their powerbooks for a while now. However, neither uses a digital connection, internally, to the notebook's display.
Como? Cuando? Que?
Of course, gigabit switches are basically unheard of outside of very large networks, so unless you're using crossover cable you're still limited to 100Mbps, which practice gives you about 10MB/s (due to overhead). And many of us are on wireless networks, which will give you even less throughput.
Um, check your rearview mirror more often...
8-port, workgroup gigabit switches can be had for $150-$200. I just bought a 3com 8-port OfficeConnect switch this week for $150 from CDW.
The prices have dropped a lot in the past 6 months. Gigabit cards as cheap as $25 (probably 32-bit PCI, which is another bottleneck) and 3com server NICs are only $120 or so. Unmanaged switches are down to $1400 for a 24-port.
We're in the process of putting all of our servers onto a central 24-port gigabit switch. The older 24-port 10/100 switches will be star-topologied off of that to connect up the employee's computers. Back when gigabit was thousands / tens-of-thousands of bucks for the cards plus the switch, it wasn't affordable.
Wolde you bothe eate your cake, and have your cake?
One of the most interesting feature of the disk, is the Auto-Spin disabler jumper. When using proper IDE RAID controllers (namely 3ware), the Auto-Spin disabler can be used to slowly spin up a large array of disks without blowing up your PSU.
just remember, Hitachi bought IBM's failed hard drive division, and subsequentially, new Hitachi drives are based on the designs and technologies acquired from IBM. Unfortunately, I'm not crazy or have the guts to play russian roulette with 6 live rounds in a sixshooter (as oppose to the customary single bullet) with my data. I've lost alot already. All 9 IBM 75GXP's I've purchased have died and several 120GXP's that my friends got, against my strongest opposition, have dead also.
What ticks me off the most was that IBM's tech support denied and denied and I got stuck with dead drives that were at the time under warranty.
Although, I would like to see some hardware review site put the Hitachi drives under MASSIVE long-term stress tests (not just one drive but several 10s of 'em or so).
For Hitachi, it's a major uphill battle. They'll have to somehow prove their worthiness again. For one, maybe they shouldn't use the name "Deskstar" as it is synonymous to "Deathstar." Distancing themselves from IBM's flaws would be best for them. It's like how auto-makers make a sub-brand of themselves to distant themselves from the typical stereotypes and so they can sell for more and look classy too (Lexus, Acura, Infiniti, etc.).
400GB will store over 33 hours of DV quality footage. This is a good thing. Time spent on managing and planning disk space is time not spent on editing or various time wasters.
However on hte low end, my home, this won't happen for awhile. What I want to know is when and if these drives will force the price of 200GB drives down? I mroe space, dangit! Truly high end systems won't touch this drive, but a lot of work gets done on less than first line equipment. This could be useful in a few years to uss low class FCP flunkies.
The thing is, old RAID cards will be useless with these monsters. The cheapest card, for a G4 anyway, that will handle simple RAID for these large disks costs about $100. Not a small cost to be neglected.
Why do I have this? I don't smoke.
Rather than re-tooling all their custom ICs to handle serial ATA, tha vast majority of new SATA hard drives are simply placing SATA to PATA converter chips on board. These chips account for the additional power consumption.
-Adam
Some suspected a design flaw.
The fluid bearings would eventually leak (oil), which would make it's way across the disk platter thanks the centrifugal force. Disks spin fast, heads hover just over the disk (extremely close, as in, much closer than the thickness of a human hair) due to the airflow created by the spinning, a droplet of oil on the disk impacts into a head that's not designed to take direct impact of that magnitude. Especially not a huge impact like that from oil attached to a very fast spinning disk, with lots of inertia. BANG! Something that is at the mercy of extremely microscopically tight tolerances gets belted right where those tolerances matter the most! Your data might still be on the disk, but one or more of the heads are now useless.
Loosing oil out of your fluid bearings can't be great either, since it is the oil that is the actual bearing itself.
PS, I worked in gyro compass/stabilizers in a military role during the 80's. I heard that the F-16's gyro bearings were actually individual air molecules! The sleeve and shaft were built to such incredibly high tolerances that there was just enough space between them to use air as the bearings! I thought this was incredible, until they were replaced with fully solid state gyros based on lasers (measuring slight changes in 3 laser beams comprising 3 axis as the aircraft would move around)!
Then IBM issued a firmware upgrade; some suspected the upgrade kept heads moving during idle time to keep them from colliding into each other. Who knows?
Heads coliding into each other? Highly unlikely. Ever pulled an old broken HDD apart? They are practically fused together on an offset arm that allows them to "clamp" one or more platters. One arm moves them all. There might be some drives with more than one set of heads/arms, but I don't know of one yet and if it did exist, shirley they would not be able to hit each other. Be great to reduce latency and access times. Especially if they each only serviced a half of the disk each. SCSI TCQ would love that.
--
There's no reason to become alarmed, and we hope you enjoy the rest of your flight. By the way, is there anyone on board who knows how to fly a plane?
War crimes, torture, lies, illegal spying... Would someone give Bush a blowjob, already, so he can be impeached?