Fujitsu Announces World's Largest Capacity Storage

Adam Eliason writes to tell us that Fujitsu has announced the world's largest capacity storage array. From the article: "the ETERNUS 8000 and ETERNUS 4000 storage arrays. Weighing in at 1.36 petabytes, or 1.36 million gigabytes, the ETERNUS file storage arrays push the envelope for enterprise data storage systems. Fujitsu uses 2,760 nearline fibre-channel 500GB disk drives in its flagship ETERNUS server (model 2100) and can be configured with up to 256GB of cache."

Comments

Boring

[Devistater]

This is boring, set up an array of 14 petaboxes and you have the same thing. http://www.archive.org/web/petabox.php Nothing new here, except maybe using 500gig drives to do so. WOW what an idea!

Re:Boring

[shmlco]

I'm tired of hearing about super-large full-sized drives and RAID arrays. Where's my 250GB 2.5" notebook drive?

Re:Boring

[hector_uk]

you could probably fit all the pr0n ever made on it.....

Re:Boring

[Devistater]

Looks like 160gig (the new perpendicular recording) is as high as laptop hdds go at the moment:
http://www.newegg.com/Product/Product.asp?Item=N82 E16822148073

But you could just get a normal 3.5" drive and shove it in a USB case.

For that matter, you could have one 160 gig laptop drive inside, and one outside in a USB case, sometimes laptop hdds dont need externally powered USB cases, they can be powered by the USB port. And they are pretty tiny and easy to carry. Using two 160 gig drives would be over your 250 gig desire :)

Re:Boring

[networkBoy]

but then you have both an extra drives worth of power draw and it's external, which is prone to more damage.
If anything put the drive in the option bay in place of the DVD.
-nB

Re:Boring

[ar32h]

Not even close. With a PetaBox you get a rack filled back-to-back with really low end PC hardware (VIA C3) and a boatload of IDE disk.

With the Eternus you get a disk array that you attach to your SAN.

They are two completely different products for totally different jobs.
You'd run you financial database on a Sun attached to a Eternus.
You'd run your Google clone off of a PetaBox.

Re:Boring

[ivan256]

It's not as easy as you may think to get good performance out of very large arrays. What would you use to cluster those petaboxes and provide a unified frontend to the cluster while maintaining a high IOPS and throughput?

You're right that there's not that much new here, but the first half of your comment indicates you have no idea the level of complexity involved in building an array this big.

Re:Boring

[shmlco]

All I know is that if I'm going to be running a half-dozen different OS'es on my notebook I'm going to need a LOT more space.

Re:Boring

[Devistater]

Its probably also 500 times more expensive :)

Re:Boring

[Devistater]

Yep you are right, I dont.

Re:Boring

[Blackhalo]

I think you'd die from exhaustion before you got through even half of it.

You kids

[mboverload]

Back in my day we only have 3 terabytes for pr0n.

Re:You kids

[dohzer]

Does the "have" imply that you are somehow trapped in your time?

My power bill is crying...

[javaDragon]

... If by any chance my appartment electricity counter does not explode, it's going to cost me more in electricity just to keep the damn' ting running than I can afford.

So, what would be the highest AFFORDABLE capacity storage ?

(I'm currently using a Buffalo TeraStation, a bit slow but not full yet)

Re:My power bill is crying...

[monsted]

They would be much cheaper to keep running than the thousands of 73G and 146G drives many of us are using now.

One of them could easily replace the 30ish EMC systems current running here - with 75% space left over - and save a fortune on service and power. 256G of cache isn't something terribly impressive, though, as both EMC and HDS has had that option for many years.

Re:My power bill is crying...

[odourpreventer]

Can't they have some sort of Power-up-on-demand? Running only those needed at the moment shouldn't be to difficult or time-consuming I believe.

Re:My power bill is crying...

[Jeff DeMaagd]

I would imagine that those drives are being continually accessed. Besides, any time a drive is powered down, you get a very slight chance that it won't start back up again. That is why drives have a rated number of start-stop cycles. In a RAID, it's not that big of a deal, but it is still more expensive than the minor cost savings you might get from powering the drive down only to start it up again in a few minutes. For home use, I can see it being worthwhile as the drive might not be used except a few hours a day, though say, torrenting directly to the drive is going to prevent a drive sleep.

1.36 Petabytes? Or 1.36 million gigs?

[iogan]

"Weighing in at 1.36 petabytes, or 1.36 million gigabytes"

I'm not mathematician, but isn't one gig 1024 MB? And one MB 1024 kB? And one tera 1024 GB, etc.. you get the picture...

Re:1.36 Petabytes? Or 1.36 million gigs?

[mwvdlee]

No, you're confused with Kibi, Mebi, Gibi, Tebi and Pebi.
But they've only been around since 1998, so you're forgiven ;)
http://en.wikipedia.org/wiki/SI_prefix.

Re:1.36 Petabytes? Or 1.36 million gigs?

[my $anity 0]

You're a company, trying to outdo other companies. Would you rather have a gigabyte be 1,000,000,000 bytes, or 1,073,741,824 bytes? Let's take a 60 gig drive, which now has 60,000,000,000 bytes. You could call it 60 gigabytes, or 55.9 gigabytes (binary). Which would you do? Now, at the level of 1.36 PB, it would make the difference between 1.36GB and 1.21GiB. So, it's a marketing thing. You want to sound like you have as much space as your competitors, and nerds may know that 60GB=55.9GiB, but people who think they are savvy, who are most of the people who buy new drives, would thing 60>>55.9, and they're the same price! Even though most consumers don't buy 1.36PB arrays, this carries over anyway.

Re:1.36 Petabytes? Or 1.36 million gigs?

[4D6963]

Yeah, but I think that "implicitly" in the storage domain, a kilo=1024 (as in the networking domain a kilo=1000)

Re:1.36 Petabytes? Or 1.36 million gigs?

[way2trivial]

I think IMPLICITLY in the storage domain, a kilo=1000 and a giga=100000

have you EVERY purchased a hard drive? have you EVER seen the disclaimer on the side of the box?

here is seagates take
http://www.seagate.com/products/discselect/glossar y/index.html#cap
Capacity:
Capacity is the amount of data that the drive can store, after formatting. Most disc drive companies, including Seagate, calculate disc capacity based on the assumption that 1 megabyte = 1000 kilobytes and 1 gigabyte=1000 megabytes.

Re:1.36 Petabytes? Or 1.36 million gigs?

[4D6963]

OK, didn't wanna do the karma-wikipedia-linking-whoring, but you didn't leaveme the choice this time ;-)

Gigabyte

1,073,741,824 bytes, equal to 10243, or 230 bytes. This is the definition used for computer memory sizes, and most often used in computer engineering, computer science, and most aspects of computer operating systems.

Well alright, it only half proves my point, mostly that it says right above that that "1,000,000,000 bytes is the decimal definition used in telecommunications (such as network speeds) and some computer storage manufacturers", but see, this is *only* used by computer storage manufacturers because they *can* cheat (just like ATM speeds advertised instead of IP speeds for ISPs).

Now, when in this article we turn petabytes into millions of gigabytes, we are not being those nasty hard disk manufacturing cheaters, we'll rather use GiBs and PiBs for GBs and PBs because that's what we always use, right? (just look at how your OS counts your hard disks bytes)

Re:1.36 Petabytes? Or 1.36 million gigs?

[PGC]

The folks that came up with those `standards' should be put against the wall...when is that bloody revolution ?

Re:1.36 Petabytes? Or 1.36 million gigs?

[KingJoshi]

Computer memory is not HD space. Hard-drives are the exception to the rule. So a 100GB hard drive will not be reported as 100GB by the OS. You can hate it, as I do, but it's now pretty much implicit regarding hard-drives.

Re:1.36 Petabytes? Or 1.36 million gigs?

[Kjella]

Or maybe using the standard most of the computer world, including but not limited to operating systems, applications, RAM, GFX card and so on use for MB/GB/TB/PB? And if you buy a 1Mbit DSL line here you will get 1024kbit.

Face it, the hard disk industry pulled a real cheap shot by shipping 1,000,000,000 bytes where the user expected 1,073,741,824. You can quote me SI standards up and down but I assure you that they didn't give a flying fuck about that, except as a convienient excuse.

Yes, I'm in favor of using MiB, GiB etc to end this confusion, but don't pretend that the HDD industry "were right all along". They created confusion where there was none purely for the sake of marketing and profit. If they hadn't done it, we would happily consider "MB" as a separate and distinct unit from 10^6 "B".

Re:1.36 Petabytes? Or 1.36 million gigs?

[aderusha]

I'm not sure where you are buying DSL, but the datacoms world has been using 1k == 1000 for decades now, long before storage vendors caught onto it.

Re:1.36 Petabytes? Or 1.36 million gigs?

[swillden]

And if you buy a 1Mbit DSL line here you will get 1024kbit.

Not on any DSL I've ever seen. Can you provide evidence for this claim? In general, data communications speeds have always been measured with SI prefixes (powers of ten, not powers of two).

Face it, the hard disk industry pulled a real cheap shot by shipping 1,000,000,000 bytes where the user expected 1,073,741,824.

Umm, if the user expected that, it's because the user didn't know anything about hard drives. The first hard drive ever, in 1953, had a capacity measured in millions of characters. I haven't been able to find any references on Seagate's 5.25" 5MB hard drives (the first PC drives) so I'm not sure if that was closer to 5*10^6 or 5*2^20 (I'm pretty sure it wasn't exactly either), but by the time drives got to 40MB or so all of the manufacturers had settled on SI prefixes, rather than binary prefixes. I'm not actually certain there was ever any confusion... it would make sense that drive space was always measured in SI units, because the PC storage industry was started by people who came from the mainframe storage world, which had always used SI units. Wikipedia's article on binary prefixes explains it this way:

Hard disk drive manufacturers state capacity in decimal units. Since most computer operating systems report drive usage and capacity in binary units, the difference causes an apparent loss between the advertised capacity and the formatted, usable capacity. This usage has a long engineering tradition, predating consumer complaints about the apparent discrepancy, which began to surface in the mid-1990s. The decimal-based capacity in hard disk drives follows the method used for serially accessed storage media which predated direct access storage media like hard disk drives. Paper punch cards could only be used in a serial fashion, like the magnetic tapes that followed. When a stream of data is stored, it's more logical to indicate how many thousands, millions, or billions of bytes have been stored versus how many multiples of 1024, 1,048,576, or 1,073,741,824 bytes have been. When the first hard disk drives were being developed, the decimal measurement was only natural since the hard disk drive served essentially the same function as punch cards and tapes. Thus, today, any device that is addressed or seen as "storage" uses the decimal system to identify capacity.

In any case, by the time drives got to 1GB (your example), the practice had been firmly established for decades. Those of us who started with floppies and moved to hard drives sort of expected binary prefixes, but I'm not sure why, because even floppies didn't really use them. 360KB diskettes actually held 60*1024 bytes, but by the time we got to the 1.2MB and 1.44MB diskettes, the prefixing had gotten really confused. For example, a 1.44MB diskette holds 1440*1024 bytes. "MB" in that case is an odd mixture of binary and SI prefixes.

I suppose what would have met user expectations with that first hard drive was a device that stored 5000 KiB, since DOS didn't report free space in MB, but in binary KB). I'm really glad the storage manufacturers didn't continue measuring in units based on power-of-ten multiples of KiB, though. That would have been really confusing.

Re:1.36 Petabytes? Or 1.36 million gigs?

[swillden]

So, it's a marketing thing.

Actually, it's a historical thing. The first hard drives were measured with base 10 units, because the storage technologies that had preceded them -- punched cards and magnetic tapes -- had also been measured with base 10 units. Those devices had storage measured in base 10 units because it was the most convenient and logical way to count. Early in the history of computing, memory was also measured in base 10 units, too, but that changed because it was significantly easier to manage in binary units, since it was addressed with power of two pointers. No such compelling reason ever emerged for storage (though it did make sense for a time to measure storage in power of 10 multiples of 1024).

Re:1.36 Petabytes? Or 1.36 million gigs?

[iminplaya]

10^24, 10^24, 10^24

Re: 1.36 Petabytes? Or 1.36 million gigs?

[HTH NE1]

Yes, I'm in favor of using MiB, GiB etc to end this confusion

So am I, as long as it doesn't confuse the steel industry and their units of kips (kilopounds of force (1000 lbs)), and the computer industry and their MIPS (millions of instructions per second) and KIPS (thousands of...).

Meanwhile, don't confuse the musicians. (1.36 million gigs? How many sets is that? And how many songs per set?)

Finally, to Stephen Colbert: Megamerican? Keep kicking it up and what do you get? Gigamerican, Teramerican, Petamerican, Examerican.

Re:1.36 Petabytes? Or 1.36 million gigs?

[Anonymous Coward]

Amen.

Re:1.36 Petabytes? Or 1.36 million gigs?

[Glonoinha]

Sounds like we agree. Let me know if you guys need me (see .sig for details.)

Re:1.36 Petabytes? Or 1.36 million gigs?

[PGC]

May you be able to continue slaying for many years to come ! Go forth, Slayer of MebiBytes.

Re:1.36 Petabytes? Or 1.36 million gigs?

[PGC]

But HD space IS computer memory and definitly not an exception. You can hate it, but that's how it is.

Re:1.36 Petabytes? Or 1.36 million gigs?

[way2trivial]

Whatever the technical definition of gigabyte should be or is, matters not to me or my post.

What does matter to me- and the element I responded too, was that the original poster has COMPLETELY wrong a claim that "hard drive manufacturers call it like it should be at 1024 per K"

anyone who works with Hard drives (the small 3.5" or 2.5" or smaller elements that require power and ribbon cables as opposed to
'Cletus, you know the hard drive of a computah is what doesn't mean a monitor or keyboard" )

  KNOWS that hard drive manufacturers most certainly DO NOT use 1024 maths but 1000's...

Re:1.36 Petabytes? Or 1.36 million gigs?

[BrianRaker]

I think IMPLICITLY in the storage domain, a kilo=1000 and a giga=100000

I think you're short a couple decimal places there on your "giga" value. Where byte=1, a kilobyte would be 1,000, a megabyte would be 1,000,000, and a gigabyte would be 1,000,000,000.

For sake of this argument (and not to be extended outside of this conversation), kilo==kebi, mega==mebi, and giga==gibi.

Re:1.36 Petabytes? Or 1.36 million gigs?

[daenris]

Well, I can't definitely say that DSL uses 1024kbit to mean 1mbit, but I do assume they operate on powers of 2, as my DSL (before it was upgraded) was 384kbps down. And now I have a "1.5Mb/384Kb" line. So, based on the Kb numbers they quote, I'm led to assume that the 1.5Mb is actually 1536Kb. Of course this may not be the case, but it's what the numbers imply.

Re: 1.36 Petabytes? Or 1.36 million gigs?

[BrianRaker]

Based on what I read here on /., I'm sure there are plenty of Examericans and even more wannabe Examericans. :p

Re:1.36 Petabytes? Or 1.36 million gigs?

[swillden]

Well, I can't definitely say that DSL uses 1024kbit to mean 1mbit, but I do assume they operate on powers of 2, as my DSL (before it was upgraded) was 384kbps down.

It would be nice to hear from someone who actually works on ADSL technology, but I've done a lot of data comms work with other technologies, and it's always been base 10. I would expect that 384kbps actually means ~384000 bits per second, including overhead bits for framing, etc. I agree that 384 looks suspiciously power-of-two-ish, though, being 2^7*3.

The reason data comms stuff is (nearly?) always base 10 is because data rates always derive ultimately from frequency ranges, which are measured and allocated in base 10 powers of hertz, and which are actually generated by oscillators whose frequency derives from a crystal or semiconductors acting like one. None of those things have any preference for powers of two, so the world that signals engineers work in is all base 10.

You got me curious, and I just went off to look at what drives DSL data rates, and the result is interesting -- there is a power of 2 involved, though it's kind of artificially imposed and I don't think it survives to affect your final data rate.

ADSL, at least using ITU G.992.1, which is one popular standard for ADSL, has 1104Khz of bandwidth, divided into (here's the power of 2) 256 channels of 4.3125 Khz, called bins. A given DSL data connection doesn't necessarily use them all, though, and they don't necessarily all carry the same amount of data. For example, bin 0 isn't used at all, and bins 1-7 are generally reserved for voice. Some bins are allocated to upstream data while most go to downstream data (but echo cancellation can allow some upstream bins to be used for downstream data as well). Each bin can transmit between 2 and 15 bits per symbol, depending on the amount of noise and the signal attenuation, and there are 4312.5/2 = 2156.25 symbols per second, meaning each bin can send between 4312.5 and 32343.75 bits per second. But as you get farther from the DSLAM, different bins suffer different amounts of noise. In particular, higher frequencies suffer worse. So the DSLAM and your modem will send different numbers of bits per bin, based on the noise level and signal attenuation.

On top of that, some of those (very weird-looking) numbers of bits are allocated to error coding. And if that weren't enough, most DSL providers artificially cap speeds based on your contract -- and *that's* where the powers of two in your final rate probably really come from.

Complicated stuff. After reading all of that I have no freaking idea whether or not the 1.5Mbps you have is base 10, base 2 or what. If I had to bet, I'd bet it's an approximation of a real number isn't necessarily all that close to either 1.5 * 10^6 bps or 1.5 * 2^20 bps.

Re:1.36 Petabytes? Or 1.36 million gigs?

[Anonymous Freak]

My favorite is the inconsistent floppy disk usage.

A 3.5 inch, high-density floppy disk is 1440 binary kilobytes (or kibibytes, for you wierdos who like the term.) Or 1,474,560 bytes. Yet floppy disk manufacturers (and most consumers) refer to them as 1.44 megabytes. That would imply 1.44 * 1024 * 1024 = 1,509,949.4 bytes, which is incorrect. (Or 1.44 * 1000 * 1000 = 1,440,000, which is also incorrect.) So the commonly accepted terminology is a bastard base-10/base-2 multiplication mix.

Then there are the companies that advertise them as '2 MB unformatted'. There is the 'DMF' format that gets 1,720,320 bytes, (1680 binary kilobytes, also inconsistently called 1.68 MB,) but even that doesn't come close to 2 MB (even 2,000,000 bytes.) (Yes, I know the underlying technical reason, but it's still dumb in the same way computer CRT monitors are advertised based on how physically large the tube is, even the parts of the tube that are under the plastic shroud.)

Re:1.36 Petabytes? Or 1.36 million gigs?

[swillden]

My favorite is the inconsistent floppy disk usage.

Yeah, those were the "power of 10 multiples of 1024" I mentioned.

First look..

[Squalid05]

..at hugh hefner's porn server!

Re:First look..

[HTH NE1]

..at hugh hefner's porn server!

Not Larry Flint's?

Nice

[thelonestranger]

I'll take two please.

no clusters

[SolitaryMan]

You don't even want to imagine a beowulf cluster of those...

Consider yourself lucky...

[PinkyDead]

...you could be the guy in Fujitsu quality control that has to count them.

But...

[Anonymous Coward]

DOES IT RUN LINUX?

joke now satisfied!

Google Drive!

[simonjp]

So I guess that GDrive no longer needs to wait as the storage solution for everyone's junk is here...
Perhaps those new Seagate 750GBs would be a nice swap to the standard (small) 500GB drives :P

Now what would be really cool...

[Baldrson]

With that many drives and a bit extra investment you could afford to do some custom drives. Develop some 500G drives with wafer-scale parallel read heads in which you can read the entire contents of a disk once per revolution and have compare circuitry out on the heads looking for matches. With all the drives synchronized to do that in parallel, you get something like 120 searches/sec of the full text of the entire petabyte.

Now a Beowulf cluster of those would be cool.

Overcompensating

[Anne_Nonymous]

Sounds like the Japanese are overcompensating for shortcomings elsewhere.

This is news?

[Gates82]

A large company buys prodcusts from another large company and puts those parts together in one package. I mean really this is not news. It might be nice to have a single point product order for al arge datacenter, but how hard is it really to connect a bunch of HD's with infiniband? And with fibrechannel I'm assuming they are using Seagate drives. I only see SATA and SCSI reaching 300gigs on their website.

--
So who is hotter? Ali of Ali's Sister?

Re:This is news?

[Anonymous Coward]

What or who is Ali of Ali's Sister?

How long to boot up? How much power?

[ivan256]

A quick bit of math tells me that if they spun all those drives up at the same time the ting would draw at least 300 amps at 220V. Since this thing probably plugs into a 30 amp circuit, I wonder how long it takes to complete the staggered spinup... I wonder how long it takes for the power usage costs to exceed the purchase price...

Re:How long to boot up? How much power?

[Anonymous Coward]

The power specification lists a maximum power usage of 82,900W so presumably more is needed than a 30A circuit. But if you can afford the machine you can probably afford the custom electrics to run it.

Re:How long to boot up? How much power?

[HotNeedleOfInquiry]

The old "washing machine" hard drives of the 70's had logic to power up one at a time starting with unit zero. Sequencing power on modern drives is even easier and no doubt they do it.

I don't even want to think of how much audio

[Skapare]

Based on the highest quality of standard definition video on a 4.7 GB DVD (one hour), this thing would store just a bit over 33 YEARS of video. I don't even want to think of how much audio that would hold in MP3 format.

Re:I don't even want to think of how much audio

[Anonymous Coward]

At a VERY rough guestimate, 874.4 years, if you take 1mb = 1min, as is the rough rule of thumb for "industry standard" 128kbit/s mp3 quality.
If, like any self-respecting ear, you prefer your music at something that doesn't make your teeth want to chew cardboard or aluminium foil, it might be a little less.

I've got a good idea ...

[Anonymous Coward]

Why don't we get an existing and well accepted concept (that 1024 bytes is a kilobyte) and completely change it (to 1000 bytes in a kilobyte) so that hard disk manafacturers can get away with giving you less bytes for your kilobyte?!

Stupid corrupt standards organisations.