Internet Archive's new data center in a box

March 25, 2009 3:52 PM

Internet Archive - probably the single largest depository of Open Source content (and the Wayback Machine) - has transitioned its data center from racks of Linux machines to a Sun MD, basically a 3 petabyte data center housed in a liquid cooled shipping container, currently sitting in Sun's Santa Clara campus court yard. Sun and IA have put together an interesting interactive tour of how it works and what it looks like.

The new design uses 50% the power of a comparable traditional system.. which is a good thing for the Internet in general because “If we do not start looking closely at our data centers now, 70% of the world’s data centers will have tangible disruptions by 2011 and the systems will experience world-wide brownouts over the course of the next five years.”¹

posted by stbalbach (37 comments total)

And may I just say, as a cyber-librarian, that I love the archive, and I love this post. A liquid cooled shipping container: this is greatness.
posted by No Robots at 3:58 PM on March 25, 2009 [3 favorites]

That "just crane it into place" part makes me nervous.
posted by R. Mutt at 3:59 PM on March 25, 2009

Step one: put a Sun in a box...
posted by crapmatic at 4:10 PM on March 25, 2009 [2 favorites]

single largest depository of Open Source content

Actually I have no idea if that's true because for example 99% of the books are technically public domain, as are many of the movies and sound, and the wayback content is copyright. Anyway, it's a very large archive of freely available online content.
posted by stbalbach at 4:12 PM on March 25, 2009

So cool.
posted by fuq at 4:26 PM on March 25, 2009

mmmmm, geek porn. Preciousssssss.

if there was an HTML "drool" tag I would have bracketed this comment with it.
posted by yiftach at 4:27 PM on March 25, 2009

ooooooooh yeah, I found a recording of a band playing DJ Shadow's Endtroducing live. Part 2. Just wanted to share that. There is awesome stuff on archive.org.
posted by fuq at 4:34 PM on March 25, 2009

Brewster Kahle, archivist and idealist, from the March 5th Economist.
posted by fantabulous timewaster at 4:48 PM on March 25, 2009

fuq, you need to see this video.
posted by you at 5:00 PM on March 25, 2009 [1 favorite]

"We put a datacenter in a box".

"Why is that a good thing?"

"Well Now you can bring a datacenter in by crane. Or put it on a roof."

"What if I don't want to bring it in by crane?"

"But it's in a box"
posted by kiltedtaco at 5:02 PM on March 25, 2009 [1 favorite]

If y'all missed it, there's a better way to browse the live music archive, but that's just a fraction of the site. There are quite a few netlabels offering free music via Archive.org. Most of what I stumble across is electronica of sorts, but there could be a lot more I'm just not seeing.
posted by filthy light thief at 5:07 PM on March 25, 2009 [1 favorite]

And one night, Archive.org disappeared. The next day, someone was trying to sell "Teh Hole Internet" on sfbay.craigslist.org

(Luckily, the whole project is mirrored in Bibliotheca Alexandrina, so they just needed another sun-in-a-box).
posted by filthy light thief at 5:11 PM on March 25, 2009 [1 favorite]

Oh shit, this is really cool. I get so much music from the Archive it isn't even funny. Thanks, this rocks.
posted by baphomet at 5:13 PM on March 25, 2009

Internet-in-a-Box
posted by Xoebe at 5:35 PM on March 25, 2009 [1 favorite]

I recommend the "success story" subvideo on sun's announcement page.

Sun credited the Archive with coming up with the idea of a datacenter in a shipping container and was presented by Bruce Baumgart at a hacker's conference.

Interesting to think of a shipping container full of computers as a single computer, but that is what it is to the Internet Archive: 3000 disks, 63 computers, 252 cores, lots of ram...

Great to see these things actually get built.
posted by brewsterkahle at 5:49 PM on March 25, 2009 [5 favorites]

... and I still can't get the Wayback Machine to actually work 80% of the time.
posted by markkraft at 5:54 PM on March 25, 2009

When I was a kid all computers were this size and took freaking punchcards and ran on diesel fuel! Now I have an iPhone that's more powerful.

Wait 30 years and I'll be using this thing to post to metafilter...from the beach!
posted by cjorgensen at 5:59 PM on March 25, 2009

Heh, big deal. I carry like three of these things around in my nerdpurse.

What? You never know when you want to make a few copies of the whole internet.

i accidentally the whole internet
posted by loquacious at 6:03 PM on March 25, 2009

brewsterkahle: "Sun credited the Archive with coming up with the idea of a datacenter in a shipping container and was presented by Bruce Baumgart at a hacker's conference."

Interesting. That paper actually credits you with the original idea for a shipping container data center. Very cool. When i visited your Presidio office in 95 or 96(?) and saw the first rack I knew it was going to be something big eventually just figured it would be a lot bigger than a shipping container.

Great to see these things actually get built.

Yeah and the environmental savings from reduced electricity and building footprints etc.. maybe we need one of these in a cave in the Arctic sort of a data version of the Svalbard Global Seed Vault.
posted by stbalbach at 7:04 PM on March 25, 2009

F'ing awesome.
posted by fourcheesemac at 7:33 PM on March 25, 2009

I hope those crane operators are really really careful. How much does one of these cost, anyway?
posted by echo target at 7:49 PM on March 25, 2009

If you don't like it in the courtyard, you can throw it in the ocean.
posted by headless at 7:51 PM on March 25, 2009

When I was a kid all computers were this size and took freaking punchcards and ran on diesel fuel! Now I have an iPhone that's more powerful.

I think about this a lot. I accumulate personal data at about the same rate that hard drives double. When i filled up my 20 gig in my old Cube, 60 gigs was standard. When I replaced the Cube, I got a 160 gig with the new machine. 2 years later, added a 250 gig in the 2nd slot. Just replaced it last month with a 500 gig. That's 25x storage increase in 7 years, and I haven't had to increase my storage footprint by one centimeter.

When I was a kid, we went on a field trip all the way to Berkley, because they had a Computer! This was a big deal -- probably '69 or '70. It was this monolith about 6 feet around (It was either cylindrical or octagonal) and floor-to-ceiling. In the middle of the thing was a 8 or 10 inch B & W display -- the fact that it wasn't monochrome was probably rare & exciting -- they showed us a computer-generated likeness of Abe Lincoln that was made out of about 50 or 100 pixels, in varying (perhaps 6?) shades of grey. It was a very important advancement in technology, we were told. I wish I could track that image down.

Also, I remember carrying my punch-cards (contained grades & pertinent student info, I guess) from one school to the next, when we moved one time.

Anyway, I have room for one of these storage-thingies in the back yard. Y'all swing on by with the crane any ol' time.
posted by Devils Rancher at 7:58 PM on March 25, 2009

In a few years, you will be able to buy a 3 petabyte flash card for your pocket tomograph for $25.
posted by b1tr0t at 10:07 PM on March 25, 2009 [1 favorite]

Devil, was it this image?
posted by Pope Guilty at 10:41 PM on March 25, 2009

Yo dawg, I heard you like marketing gimmicks, so our salesmen put some sick decals on the side of our latest ultra-low-volume digg-bait, and craned it in to our own parking lot!
posted by blasdelf at 12:14 AM on March 26, 2009

The video tour was pretty neat, though one part surprised me. Referring to the image of the shipping container cum data center: "there's nobody inside." Because I expected these things each required one if not two gnome-like sysadmins to sit there hunched on a stool in the dark to monitor the blinkenlights.
posted by exogenous at 6:03 AM on March 26, 2009 [1 favorite]

Costwise, the MD starts at US $560,000, not including the actual servers and software, or shipping insurance necessary for getting the MD to its destination. As 20' cargo containers like the MD are designed to be carried by trucks (previously), their tare (empty) weight is (4850 lbs.), making them 900lbs. lighter than the base version of the Cadillac Escalade. Even with the sensors and racks inside, the MD could be hoisted without too much difficulty, and it would be a given that Sun-authorized personnel would be on hand to supervise the installastion.
posted by Smart Dalek at 6:32 AM on March 26, 2009

Apparently it needs an external water chiller, not included. Slightly disappointing as it takes away from the all-in-one-box idea, but I guess it could be mounted on the roof. And white is not the only color option (but recommended for solar heat reasons).
posted by stbalbach at 7:11 AM on March 26, 2009

Devil, was it this image?

Yes! I've been looking for that for years! Anybody have any history on that?
posted by Devils Rancher at 7:44 AM on March 26, 2009

Oops, meant to italicize the quote, there, Pope Guilty.
posted by Devils Rancher at 7:48 AM on March 26, 2009

How does one do disaster recovery and backup for something on this scale? I mean, a shipping container is pretty fragile, when you look at the freak accidents that happen with traditional brick and mortar data centers. Floods, lightening, drunken truck drivers and oceans eleven style rapscallionism all come to mind as potential catastrophes waiting to happen. Not to mention the giant box that says "The Internet" sitting in a parking lot begging drunk morons to mess with it. All in all a pretty fantastic concept, did I heard google was planning on doing something similar?
posted by crunchywelch at 8:05 AM on March 26, 2009

This flash presentation appears to have some background. Short version: it was created by an early computer artist named Leon Harmon for an early presentation on facial recognition. Harmon was also one of the pioneers of photomosaics.
posted by Pope Guilty at 8:35 AM on March 26, 2009

did I heard google was planning on doing something similar?

Google received a patent for using shipping container datacenters operating in sync with a cluster, but not for individual containers serving as a network. If necessary, the system could have a structure being constructed around it, like a steel-sided building; since the essentials - the servers themselves, the other rack architecture, and the cooling system can be updated without any drastic reworking of the container itself, the center could plausibly be nestled on the lower floor of a building during construction, while the foundation and exterior would be built around it.
posted by Smart Dalek at 9:19 AM on March 26, 2009

Many data centers are built in underground parking garages. Wall it off and it's invisible, but has access to the building resources (power, water, telcom).
posted by stbalbach at 11:54 AM on March 26, 2009

Video from Sun - walk through of Project Blackbox
posted by stbalbach at 5:05 PM on March 26, 2009

Neat! Is the old cluster going to be used for DR?
posted by cj_ at 7:51 PM on March 26, 2009