EMC rolls out $4 million petabyte array

Forget about that puny terabyte RAID system you just set up. If you want to be on the cutting edge in storage, you need to start thinking in petabytes. And if you're ready for that level of storage, there's now someplace to get it: EMC has launched its first petabyte array, a version of the company's flagship Symmetrix DMX-3 system that includes nine room-filling cabinets of drives. The Symmetrix DMX-3 system is built around 500GB drives; the petabyte version includes a whopping 2,400 of them and goes for about $4 million. Of course, if your needs are more modest, you can always opt for one of EMC's "entry level" systems; a two-cabinet version that will give you a mere 480TB of storage is available for just $250,000. But if you're into, say, NSA-level data mining, want to challenge the Wayback Machine, or just want bragging rights, a full PB is the only way to go.

[Ben @ Jan 30th 2006 9:16AM]

Geeze I wonder what the redundacy is offered with that many drives. Chances are you will have multiple drive failures at a time.

I would hate to lose a petrabyte of information since backing that up would be a near imposability.

[Chris @ Jan 30th 2006 9:37AM]

0.5TB per disk x 2,400 disks = 1.2PB of storage
(in round figures, ignoring 1000 v.s 1024 etc.)

So there is some redundancy in there (or overhead?).

But whether that means it can survive the loss of 400 drives is another matter!

[Victor @ Jan 30th 2006 9:47AM]

Half a petabyte in two cabinets. A full petabyte in nine cabinets, not four. So it doesn't scale linearly. How is the remaining space used: extra cooling, redundancy?

[Peter @ Jan 30th 2006 10:00AM]

1PB = $4m
480TB = $250,000

Methinks someone moved a decimal point somewhere.

P.S. Engadget's email validator is borked... doesn't like '+' in email addresses, despite the fact that they are perfectly valid.

[Mike Altieri @ Jan 30th 2006 10:10AM]

What we really need is a $4, million-petabyte cabinet.

[Zach @ Jan 30th 2006 10:22AM]

Why wouldn't you buy three 480TB units for $750,000 if you want the same 1200TBs of space that 2,400 500GB units would provide?

[John C. Randolph @ Jan 30th 2006 10:22AM]

Looks like you can get this much storage from Apple for about half the price.

-jcr

[YourTechSupport @ Jan 30th 2006 10:28AM]

When ever I discuss things like storage space or bandwidth connections, I like to utilize a unit of measurement that everyone can instantly understand.
Alot of people don't know how big a Terabyte is. Let alone a Petabyte. Those that do know wonder what the bloody hell you'd even need that much space for, or why you'd need a pair of OC-rating connections for it. Are you spying on the planet? Backing up the entire continent of australia? No! You're storing...

bah, nevermind, I'll just stick with my original comment: That's alot of porn.

[rsteck @ Jan 30th 2006 10:35AM]

Agreed, there has to be a Math error somewhere in the pricing, I don't see anything right off on EMC's site that says anything about what they cost. I think the 480TB is supposed to be 48. We pay about $10,000 for 3TB, if there was a 480TB system for 1.3 million less than what we pay, I am sure we would already have it.

[Benny @ Jan 30th 2006 10:57AM]

You know...this is neat and all but it still won't hold my game collection...Oh well...

[Christian @ Jan 30th 2006 11:05AM]

Hmm... having a million harddrives in half a million RAID1-sets wouldn't be much more insecure than having two harddrives in one RAID1-set. Chances of both disks failing in the same set, is still pretty small...

[Bryan @ Jan 30th 2006 11:05AM]

480TB = 960 x 500GB disks.

Price of 500GB SATA disk on Froogle: $329

960 x $329 = $315840 for the BARE DISKS, with NO RAID

Yup, I'm pretty sure someone left out a decimal point.

[Eric @ Jan 30th 2006 11:14AM]

EMC's systems generally use RAID 5 protection. That's probably the reason for the capacity difference.

[ken @ Jan 30th 2006 11:23AM]

You guys are missing the point. EMC is not just storage, it also makes excellent software to work with that storage, you wont find that with most other vendors.

[George @ Jan 30th 2006 11:39AM]

"grreat"

Somebody has been spending too much time with tony the tiger.

[m @ Jan 30th 2006 12:00PM]

It's not a mistake on the pricing. The $250,000 is much lower end storage. The $4mil is a symmetrix which is top of the line. Everything is redundant. Only banks and hospitals w/ the $$ and need will buy these basically fail-safe machines.

[homer @ Jan 30th 2006 12:44PM]

yeeeh in 10 years this will con $100.

[Andrew @ Jan 30th 2006 1:52PM]

From the article:

"The new, entry-level version of the DMX-3 only requires two cabinets and 96 drives for its 480 TB of capacity, but even so will cost a cool $250,000 (�140,000). This suggests that a fully-packed, nine-cabinet DMX-3 boasting a petabyte of storage could cost up to $4m."

It was probably ZDnet that moved the decimal. The entry level unit, with 96x500GB drives, should be 48TB, not 480TB. For $250,000, that makes much more sense.

[Lee McCaleb @ Jan 30th 2006 3:55PM]

By the way, the DMX-3 also supports 73GB, 146GB, and 300GB drives as well as the 500GB drives. RAID support is 0, 1, 1/0, 5 (4+1) or (7+1). You can mix and match any of the drives as well as RAID sets. Sparing is done on a global basis. The recommendation is 1 spare drive per 30 production drives. You can start with 60 drives and expand hot and online to 2044 drives. The conectivity options are Fibre channel, iSCSI, or FICON. The DMX-3 will also support up to 2048 host systems. Host OS support includes UNIX, LINUX, Windows, Z-OS, and AS400 to name a few.

Just in case you were wondering...

Lee

[Galo @ Jan 30th 2006 4:18PM]

So how long does it take for this thing to defrag? lol :-p

[Merlinmurph @ Jan 30th 2006 4:48PM]

Guys, when you're figuring out the cost of this storage system, the actual disk drives are just one part of the whole enchilada, and probably account for much less than half of the total system. This is not a JBOD (just a bunch of disks)
First, on the frontend there are host adaptors, which connect to the host computers. These go for $25K? 50K? Let's just say "a lot". And a machine like this would have a bunch of these - 30? 40? 50? connected to a potpourri of host systems.
Second, on the backend, there are specialized controllers for the disks to maximize throughput. There's one of these for each bank of disks.
Third, on the middle end is the cache. These are caching controllers, and thus contain a ton of cache, though I have no idea how much cache a 1 PB system would have. Without massive cache, you have to access the disks for an I/O, which is what you want to avoid like the plague.
Then, there's probably the most expensive part, the software to run the whole shebang, and we're not just talking handling reads and writes. There are all sorts of special functions a customer can do to basically create copies of his data.
Add in all the incidentals like power systems, specialized busses, redundancy on top of redundancy, customer service/support, and all sorts of blinkling lights, and you've got a $4 million machine.

[J S K @ Jan 30th 2006 5:41PM]

Also keep in mind, people don't by one Symm. These are generally sold in some sort of "paired" configuration. Generally used in remote data mirroring/redundancy/replication applications.

Cooling is handled by some loud-assed exhaust fans on the top. how do I know? we have two Symm 5s in our lab without the exhaust noise defusers that you can get (optional). Loud as heck! Seriously.

[Joe @ Jan 30th 2006 5:42PM]

FINALLY someone who actually understands large scale storage systems, Merlinmurph. I was about to post the same thing. I used to work for Avid, which has its own storage systems, although completely different philosophy and needs. Their large systems have redundant fail safes to ensure that playback never fails. That's because half their market is now in the broadcast field. So when a news station is playing back video live on TV, it can't have stuttering or wait for data to be read. From a consultant I've talked to that works on their storage systems currently, when one of these babies is playing back video for a live broadcast, it's pulling together data in parallel from a LOT of different drives.

So yes, the actual storage itself is a very small part of the overal system.

[Shang @ Jan 30th 2006 6:27PM]

Finally, someone tells me whats after terabytes....jw, whats after petabyte, unlimitedbyte???

[sk07 @ Jan 30th 2006 6:51PM]

In responce to Shang's question, after petabyte comes exabyte, then zettabyte, then yottabyte, then brontobyte :-)

[G @ Jan 30th 2006 7:37PM]

Ken, m, Merlinmurph, and Joe,

You guys are missing the point. No one was discussing the justification for the $4m price or the $250k price.

The question was, why would 1024TB be $4m and 480TB (1/2 the capacity) be only $250k (1/16 the price).

Bryan calculated that the discs alone (assuming SATA and not SCSI) would cost over $300k. This does not even include the additional hardware and software that comes with a solution like this.

The whole point is that SOMETHING is off here. Either the capacity is off (must be smaller), or the price is off (must be higher). I suspect it's the capacity, since I doubt 480TB would be considered an "entry-level system."

[mike @ Jan 30th 2006 8:55PM]

Didn't you listen? #18 said that the entry-level system is actually 48 TB NOT 480.

[Shang @ Jan 30th 2006 9:29PM]

Thanks, sk07...that'll be useful information next time I am on millionaire :)

[Liam @ Jan 30th 2006 10:57PM]

if you want to know what comes after petabyte...

http://en.wikipedia.org/wiki/Byte

[Fallen Kell @ Jan 30th 2006 11:32PM]

Re: Post #9, rsteck, man you are sure getting ripped off for 3 TB at $10k. You can get 3.5TB from Apple for $8.5k retail (let alone any discount that your sales rep might be able to get your company). Heck, if you have a real good sales rep and you buy a lot of equipment, you can get the 7TB version for about $10.5k-$11k...

[Don Wilson @ Jan 31st 2006 1:22AM]

In 10 years we'll be laughing our asses off at these room-filling storage cabinets, sitting next to our 3" 150PB handheld drives.

[Mukul Kumar @ Jan 31st 2006 2:34AM]

I see a lot of questions on capacity vs. costing. See the following URL for more details:

http://www.emc.com/products/systems/symmetrix/DMX_series/pdf/C1304_Symmetrix_DMX3_SS_ldv.pdf

[Merlinmurph @ Jan 31st 2006 8:02AM]

It's hard to compare prices and raw storage. EMC has their high-end Symmetrix line, and they have their lower entry level (but really hot shit stuff) Clariion line from buying Data General years ago. The prices just don't scale like you'd like them because there's a lot more than raw storage in a box. Unless you're looking at boxes with similar capabilities, you're comparing apples and oranges.

Also, as others have mentioned before, total raw storage is not a good number to use because all the storage is protected somehow, either by RAID schemes, remote mirroring, etc.

Whether the actual $$$ numbers we've seen are correct, I have no idea. Contact your local EMC/NetApp/IBM/Hitachi/HP/etc. rep

[uncledrax @ Jan 31st 2006 9:08AM]

I hate to say it, but if you think this think runs on 7200rpm SATA ][ drives, you're nuts..

The few SAN's I've seen run 15k Fibre Channel drives, we happen to use 146GB ones, and pricewatch lists those around $850ea.. (we actually use HP drives, and they are more in the $2k+ range each IIRC)..

Then I'll echo what some above said.. you still gotta purchase FC switches, cabling, drive enclosures, applicance servers, etc..

It's not always just about the raw drive capacity.

[Jeff @ Jan 31st 2006 9:14AM]

#20: I sure hope that the buyer isn't dumb enough to hook this up to a Windows computer. Pretty much any other OS defragments on the fly.

[Jeff @ Jan 31st 2006 10:11AM]

#34: "I hate to say it, but if you think this think runs on 7200rpm SATA ][ drives, you're nuts..

The few SAN's I've seen run 15k Fibre Channel drives, we happen to use 146GB ones, and pricewatch lists those around $850ea.. (we actually use HP drives, and they are more in the $2k+ range each IIRC).."

Dividing the capacity by the number of drives shows that at least one of the products use 500GB drives. Are there 500GB SCSI or FC drives? They certainly aren't available in 15k speed, or even 10k.

[Cory @ Jan 31st 2006 2:15PM]

Just looking at the article on ZDNET, it boasts that the Symmetrix uses a "new" 500GB drive. I'd be willing to bet that they are FC drives and not SATA. I'd also be willing to bet that they are 10k RPM minimally, if not 15k RPM drives. I'm also wondering if they're going to be 2Gb, 4Gb, or 10Gb fiber...? As for the high-end vs entry-level, I don't know of a Symmetrix system that runs on SATA. According to ZDNET also, the entry-level DMX-3 that uses 2 cabinets is 7TB (96 drives) for $250k which would mean they are 72GB drives, probably 10k SATA drives, I'd wager.

[merlinmurph @ Jan 31st 2006 10:22PM]

#37 From the link above (#32), all the drives are FC

[Rami @ Mar 25th 2006 4:28PM]

man, i want to do a small math in here :D.

let's say, i have a hotel, and everyday i have 30 new customers, every customer will need a space in my db, and let's not complicate things and let's say that space is translated as is that means no extras and no less no compression nothing, just let's do it simple

a file with customer information reservation..etc will be 2kb fair enough?

30 x 2 = 60kbs a day
let's now add some extra to average things out
let's say 100kb a day after we put totals and blah blah blah.

so yearly

100kb x 365 = 36500kbs / 1024 = 35.6 megas a year

add 4.4 megas extras that will be 40 megas a year :D

so a petabyte is 1073741824 megapytes (1024*1024*1024)

1073741824 / 40 = 26843545.6 years :D.

double things up for back up, triple it, still after 20 years maximum the whole system, os, hardware, software in the hotel/hospital whatever will be replaced, i really, don't think these petabytes are good for anyone who is really looking for daybyday storage or even back up, they are only good for few companies such as HUGE radio stations, and HUGE tv stations who use digital media to broadcast, and still one petabyte is more than enough, so wow lol.

sorry for being boring, just was bored haha.

[angrygoatface @ Mar 31st 2006 6:39PM]

1 Petabyte RAID Array: $4,000,000.

1 Petabyte of pr0n: Priceless.