Hacker News new | comments | show | ask | jobs | submit login
Is Google building a hulking floating data center in SF Bay? (cnet.com)
341 points by chasef 1308 days ago | hide | past | web | 242 comments | favorite

Great article. It's nice to see journalists do actual investigative work, even if it is for something like this. Today's "journalism" seems to be all link-bait opinion pieces mixed in with copy-pastes from twitter.

Agreed I'm just as much impressed with the level of investigative work done here as with the actual data center. Well done CNET.

True, although it certainly seems like Google was half-assing it with the security. If you don't want something to be tied back to your company, don't have company employees work on it.

Sure worked for the NSA, didn't it?

They shoulda searched craigslist: http://sfbay.craigslist.org/sfc/off/4148978964.html

Those stools...where do you put your knees??

That's really interesting. The article claimed the Google structure wasn't likely to be office-space, because there was little-to-no exterior-facing lighting, though. This project seems to have a lot of windows. Not the same thing?

I was thinking exactly the converse: why so much effort for gossip? It's pure tech celebrity gossip in my opinion and brings nothing to the party. Why not spend some more effort on the link between google and NSA? We still don't know how they steal our data.

I've seen hundreds of report regarding the link between Google & NSA. Isn't that enough?

This is just not gossip. It's not about Sergey Brin's love affair.

This is about a piece of technology that is possibly built by Google.

> We still don't know how they steal our data.

Actually it feels to me like they're stealing it anyway they can... so if there is a legal recourse, I think it's safe to assume they've taken it, along with many illegal ones.

Don't you think that one of the interesting issues raised by a floating data center might be a jurisdictional one? Is it possible that google is actually considering putting some of their infrastructure beyond sovereign control? Maybe this isn't just gossip...

this has been answered, it's not possible. It's about making substantial business in the US.

The territorial water is 12 miles wide, it's far. Exclusive economic zone is 200 miles wide, it's really far. Who want to be that far from the coast in the winter without a good reason on the most powerful and dangerous ocean? In the article they say 3 to 7 miles, that's completely territorial and for a good reason: on the first winter storm they'll be plenty happy with sovereign state helping them shelter the thing and save everybody.

Moreover, how do you anchor the thing in international waters? It's very complicated and expensive if the water is deep (which is more or less implied by 200 miles from the shore).

You have a very broad definition of gossip.

I actually think this is a pretty substantial find, not just gossip. They're moving on a really cool concept.

There are actually 4 "vessels". Look at the names -- using binary :)

  BAL0001	BY AND LARGE LLC	*	*	C & C MARINE & REPAIR LLC	2010	Freight Barge	2164.0	249.6
  BAL0010	BY AND LARGE LLC	*	*	C & C MARINE	2011	Passenger Barge (Inspected)	2520.0	260.1
  BAL0011	BY AND LARGE LLC	*	*	C & C MARINE	2011	Passenger Barge (Inspected)	2520.0	260.1
  BAL0100	BY AND LARGE LLC	*	*	C & C MARINE	2012	Freight Barge	2164.0	249.6

FYI, "Buy N Large" was the fictional Costco-like store from Pixar's WALL-E.

I suspect the name may just be to confuse anyone trying to look into it. "By and Large" returns over 250 million results in Google. Anyone looking to find permits, job postings, or random government filings would need to dig through tons of unrelated cruft.

This is done pretty often for shell companies, they will have names like "Twisting River" or just "Investment Vehicle VII, LLC"

If you want to throw people off the beaten path (and you have google's money) it would be trivial to acquire a small company or even create one from scratch that actually does something, anything rather than give a name which is sure to scream "shell company" and get people's sniffers out.

They could have probably found a way to take even the name that was similar to an overseas company that has a connection to shipping, maritime etc. and simply do a variation of that (and make it in a way that doesn't infringe any trademarks).

That way all the pain in the ass bystanders would simply jump quickly to the obvious wrong conclusion and move to something else.

My point being is that if you are truly trying to hide something then at least put some effort into creating a good distraction.

> it would be trivial to acquire a small company or even create one from scratch that actually does something

It's clear they are optimizing for fun.

I kind of doubt that these two names are related except that they both derive from the same common idiom.

Also "by and large" is a nautical term meaning (roughly) "whether sailing against the wind or sailing with the wind".

While we're speculating on the meaning behind the name(s) RE:wind, remember that Google acquired Makani Power a while back. They build wind generated power systems using large kites. What better place to put them into production than at sea where there isn't much worry about crashing to the ground.


unlikely that the Makani planes would be run from a ship. They each have a huge arc, much larger than the diameter of a traditional turbine.

Furthermore, what's the concern with them crashing in to the ground? It's not like they plan to install them directly over populated areas.

From the website: "Opens up large new areas of wind resource, including the vast resources offshore above deep water"[1]

Doesn't sound that far fetched that they would try to anchor this to the barge to try to tap that potential.

[1] - http://www.makanipower.com/why-airborne-wind/

>> unlikely that the Makani planes would be run from a ship.

Something like this, perhaps: http://www.skysails.info/english

Interesting to interpret in that context. Maybe a data center intended to work best exactly when other centers would be experiencing losses in connectivity?

and a common phrase (though a little less used than in previous generations) meaning 'generally'

Confirmed about the generational decline: https://books.google.com/ngrams/graph?content=by+and+large&y...

I wouldn't put it past Google to use a tongue-in-cheek name, especially when a good hunk of WALL-E takes place around a massive starship that serves as a floating ark.

The nautical double-entendre that ZeroGravitas mentions only sweetens the deal.

Noting that the domain byandlarge.com was registered in 1998 by a "domainer".

Interesting. The movie came out in 2008.

But the film concept dates to the mid 90's so most likely that concept was the first use of that name.


It's a relatively common phrase, not too surprising that someone snapped it up.

Except in the movie it is Buy not By...

The picture of what we're guessing is the data center is BAL0010 which is listed as a Passenger vessel above.

Hypothesis found in comments on OP website:

BY AND LARGE = Brin sergeY AND LARry paGE?

They are all attached to one TREQ

Something similar is being outfitted in Portland, ME.


The OP is a much better article (The Press Herald is garbage).

My original guess for the barge-building in Portland was a ocean-based prison facility for the government to use for interrogations. But it's seeming like the floating data center idea is much more likely.

> My original guess for the barge-building in Portland was a ocean-based prison facility for the government to use for interrogations.

Not a bad guess at all, since it really does look like a modern prison ship: http://en.wikipedia.org/wiki/File:Vernon_C_Bain_Correctional...

The layout is startlingly similar, down to the windows and the angled (entrance?) ramp. No stairwell visible on the ME one though. Fascinating.

This is somewhat of a tangent, but how is the tech scene in Portland, ME? It's my hometown and I've had thoughts at various times about returning and setting up shop.

It really depends on what you want to do.

If you want to raise capital and build a startup here, don't. Or, at least, do so but with your "real" headquarters in Boston, NY or CA.

There's a strong effort to try and get things going, but there's a definite lack of capital and the capital you can find in other places will send you a consistent message: "We don't want you in Maine, move here instead."

There are some opportunities around maritime businesses (logistics, mostly) and lots of incentives for trying to build small manufacturing and renewable energy businesses... but the school closest to Portland is looking to close its physics department and the state officials responsible for managing the incentive programs have proven multiple times that they cannot be trusted to develop long-term programs.

If you have or can develop a remote-work type position, it's a great place to live and work from (as long as you can get good, reliable bandwidth).

But if you want to start a software-focused startup ... honestly ... it's an uphill battle.

For reference: I spent the first 20 years of my life in Maine.

>If you want to raise capital and build a startup here, don't. Or, at least, do so but with your "real" headquarters in Boston, NY or CA.

Ah, yeah. I'm in NYC now. I figured that if I started something in Maine, I'd be on the Noreaster on a semi-regular basis. I wouldn't be looking for an angel investor and would be bootstrapping instead.

>but the school closest to Portland is looking to close its physics department

Physics isn't Comp Sci, but I get your point. USM was my first school before transferring, so ouch.

>If you have or can develop a remote-work type position, it's a great place to live and work from (as long as you can get good, reliable bandwidth).

That's sort of what I'm thinking for at least a few years. I'd be looking to buy a place on the West End and then start my own shop when it's reasonably paid down (3-5 years).

>But if you want to start a software-focused startup ... honestly ... it's an uphill battle.

Thank you for the thoughtful response. It definitely helped me put some things back into perspective.

Adding onto justinsteele and numbsafari ... I've been working remotely since 2007 because there aren't a ton of jobs up here except for large-shop Java and .net work, and that really wasn't for me. It's kinda surprising how many people live in Portland and work remotely, and it seems like I keep discovering more all the time. But as far as startups go, I think numbsafari nailed it. I worked for a smallish company back in the mid-2000s, and it's likely that things have changed, but back then the tech-related investors wanted you to move to Boston, and the local investors didn't really grasp how to value software - they were more interested in how many forklifts you had as collateral. :-/

But I do get the sense that things have improved a bit since then.

>It's kinda surprising how many people live in Portland and work remotely, and it seems like I keep discovering more all the time.

That's what I'd be doing at first. It's such a nice city, but business just can't seem to stick.

>I worked for a smallish company back in the mid-2000s, and it's likely that things have changed, but back then the tech-related investors wanted you to move to Boston

Ha! My first job was at GWI.net. I didn't realize how lowly paid Maine tech was until I moved.

The "scene" is cool. There are local meetups, some small, some mid sized. Monktoberfest just drew quite the crowd. As the other commenter said, there ain't much for investment though, and the companies are a mix of web boutiques, large companies (Unum, Idexx, etc), and a startup here or there. I work remotely, which gives me the benefits of being in Portland while making a NYC salary, but even when I was working for a Portland firm I was still making a good wage.

>I work remotely, which gives me the benefits of being in Portland while making a NYC salary, but even when I was working for a Portland firm I was still making a good wage.

I'd be looking to do something similar actually (I'm in NYC too). What kind of work is it predominantly? You mentioned web boutiques and that's sort of what I was expecting.

+1, I've been working remotely from Portland for 11 years now, great place to live but tech jobs are scarce vs. Boston, NYC, SF.

Are you a remote-worker as well? My concern would be that my company in NYC would decide to let me go remote, then change its mind due later about telecommuting and it would put me in a rough spot.

Can someone in Portland check the ship #? The article photo shows that BAL0001 and BAL0010 are in SF. I'm guessing the Portland one is either BAL0011 or BAL0100.

http://www.pressherald.com/news/Myserty_Portland_barge_and_S... is a more detailed Press Herald article, it says it's BAL 0011.

I live in Portland, I'll see if I can spot it tonight...

From the vessel listings above, I'd guess it will be BAL0011.

Does interrogations mean torture? If so, they seem to do just fine with the facilities they have, so why make a floating prison?

Hmm, what if both are right?

It could be a floating prison, filled with a giant data-center! The NSA+Google tie could mean that they both needed an interrogation free zone, one for torture, the other for stealing data. Both would benefit form letting prisoners generate the energy. They could even argue, that it's good for health to breath sea air and do regular "workout". Then they could offer the unused crunching power as an OSaaS – Overseas Sofware as a Service that is similar to AWS and Amazon Mechanical Turk.

Aside from that, isn't it illegal to systematically heat the water of the ocean with hot waste water?? Can't something be done against that?

Why are such prison ships even allowed to be built from tax-money, isn't it illegal?

My first thought: "Finally I'm interested in applying at Google, if it involves using my captain's license!"

My second thought: "Meh, the job interview probably involves writing a breadth-first search algorithm to search for known pirates in the graph of nearby ports."

No, you mapreduce the Pirate search. Send out 100 workers to check on the nearby ports and see which ones come back.

You need to think at Google scale.

Or you can be sneaky. Put out some irresistable, yet covert job offers for pirates in the necessary ports, and see where the applications come from.

Or put a "Report a Pirate" bounty and crowdsource it.

but how do you confirm the validity of the reports

There are two types of reports, fraudulent ones and real ones. The fraudulent ones will come from everywhere, while the real reports come only from pirate infested harbors. Therefore just subtract the mean report density and what remains should be random fluctuations around zero plus the real reports. So if a harbor has more than some threshold of reports, then it is pirate invested.

At this point in the discussion, you could probably correlate it with the number of the letters "R" in the reports. Because we all know that it's the pirates' favorite letter.

Arrrrgh you would think so, but no, it's the C.

Ah, laughed out loud for this string of comments. Great job, all. And not a verge "combo breaker"-comment in sight.

Mechanical turrk?

Just send in 100 journalists and 99 of them will come back.

I think a depth-first search would be more useful for a floating data center :).

The middle of the ocean could be a a very high margin location for IaaS hosting.

Due to speed-of-light delays, some of the ideal locations for hosting high-speed trading arbitrages are in the middle of the ocean. Being half way between two exchanges would allow you to notice a small difference in price of some commodity in two different exchanges and buy and sell microseconds before your competitors on the mainland know that there's an arbitrage opportunity. See http://www.physicscentral.com/explore/action/stocktrade.cfm

If this actually worked, there would be boats sitting in the middle of the Atlantic and Pacific with computers on them. I have no idea if they are sitting there, but I'm pretty sure if it was an idea that worked, it would be done by now.

That was my first thought, too.

On the other hand there are probably cheaper ways to beat your competitors than crewing a boat, paying for fuel/electricity, havign a datacenter crew on board, dealing with the challenges of getting a high bandwidth signal hundred of miles off-shore, etc. Specifically you could invest in ever-faster custom processing in hardware or a lower latency connection will still make you faster than your competitors. You don't need to be as fast as possible, you only need to be a tiny bit faster than everyone else.

On the other hand, if someone can take advantage of economy of scale (by being a neutral provider to many brokers) and provide an advantage over everyone at a lower cost than the above options, then everyone would need to use it to compete effectively.

From talking with a friend who's a retired stat arb quant, the fundamental idea is sound.

My guess for why people aren't doing it yet is that first, there are still untapped locations on land that are easier to get running (example: the Chicago to NY fiber run), and second getting the data to the ocean point isn't trivial.

I think if you want to look for early signaling of someone doing this, look for where telco folks route new fiber runs. If they take a longer path to go near one of the points in the paper I'd bet they're at least hedging for the possibility of using that fiber run for arbitrage.

Edit: a more mundane explanation may be that the arbitrage is so low margin that the payback time isn't practical vs the initial investment.

I don't understand why it's sound. The order still has to go all the way to the exchange. Why is it better to spend 50 ms to get the market data, 1 ms to do the decision, then 50 ms to send the order (middle of ocean to exchange), when you could spend 100 ms to get the market data, 1 ms to do the decision, then 0 ms to send the order (colocated hosting at exchange)?

Your algorithm is making decisions on data that's 50ms older than the data the other guy in the ocean used, I suppose. I'm not into low-latency trading enough to know whether or not that 50ms-ahead data alone could make you money, but based on how much the trading guys go nuts over it I assume it could.

Because sometimes the order you have to place is on the other side of the ocean where the data came from, costing you an extra 100ms.

Just guessing here, but the geographical advantage probably requires a fiber link, which seems prohibitive in open ocean. Putting your ship in the right place probably doesn't improve your latency if you have to bounce signals off of satellites.

It doesn't help if you need to use satellites, but you cloud use microwave. Microwave is line of sight, which would usually limit you to 40miles, but if you were to have a big blimp tethered to a big barge, then you can increase the line-of-sight distance.

I think it's possible to build a platform, like an oil drilling platform, survivable in the middle of the ocean (against storms and rogue waves) and hook fiber into it by dropping fiber in extremely high-strength conduit (like oil drilling pipe) to the ocean floor and then across the ocean to continents.

I don't think the cost/benefit is acceptable considering more serviceable locations that already have fiber, like Hawaii in the Pacific. Operational costs for a floating platform are much higher, and fixing anything major would take days instead of hours.

What about DDoS attacks (N/A if these are private network links) or real pirate attacks?

It's hard to DDOS private networks that aren't accessible from the internet.

It's a fairly common (and silly) dismissal of any new effort that "if it worked, someone else would already be doing it".

Except I didn't dismiss the idea. I'll restate, "Given the urgency and reward behind efforts around gaining advantages of market trades, I can assure you if it's technically possible, they've done it and perhaps we don't know about it."

Notice I did NOT state the corollary, "I don't see trading boats in the Pacific, so it's not possible to do.". I suppose that could be accidently inferred from my statement, but that was not my intent.

It is theoretically sound: TANSTAAFL

It is theoretically sound if you assume that the present state is equivalent to the long-term equilibrium (e.g., ignore friction.)

Mentioning the speed-of-light delays sort of implies that you'd have a straight shot fiber run to both exchanges (no circuitous routes, at least no more circuitous than other exchange participants - can't be behind the rest of the market, right?).

I'm not sure that this is possible without working directly with someone like Hibernia Atlantic or FLAG to get a fiber splice in the middle of the ocean.... which puts this beyond the reach of the average startup and into Goldman-or-JP-morgan territory, pricewise.

For those who want to play at home - where would you drop your offshore arbitrage platform, given the cable spread on this map: http://submarine-cable-map-2013.telegeography.com/ ?

I was Googling to find what a boat-based wave-powered generator would look like, and I found this interesting diagram:


There are some current wave-powered generators (such as http://www.oceanpowertechnologies.com/) - but mainly focused on (networks of) smaller buoys

I don't think wave power is deployed successfully yet[1]. The cooling would not be a leap of technology though.

[1] Its under development http://www.pelamiswave.com/ but I don't think its finished yet.

Yeah, that's why I was trying to figure out how they would power it.

Diesel generators seem... unGoogley.

Plug it in to the mains? I don't think you want tons of servers in the middle of the ocean. How would you get the data out without fibres anyway?

Why would you want it to be in the middle of the ocean? Keep it close the the coast, perhaps hook into a cable landing station (where submarine comm cables come ashore.) Running a submarine power cable also doesn't seem too far-fetched.

With your Google Blimps?

The article's patent diagram reminded me of an Arthur C. Clarke short story (I can't remember the name) about generating power from the temperature differential between the surface of the ocean and cold depths.


Not enough depth in the Bay for OTEC, though. (You need a bigger temperature gradient)

Building it in on an island in the Bay doesn't mean you plan to operate it in the Bay.

Maybe with their windmill kites? https://news.ycombinator.com/item?id=5754619

Diesel might be cleaner than all the alternatives, i.e. powergrid, and overall impact might be reduced if coolings costs are (as they should).

Goodness, that actually looks like something from Road Runner cartoon.

That seems... very inefficient. But I don't know enough to argue it.

Finally, something with better bandwidth and worse latency than Andy Tanenbaum's station wagon.

> Never underestimate the bandwidth of a station wagon full of tapes hurtling down the highway. -- Andrew S. Tanenbaum, Computer Networks, 4th ed., p. 91

  Google was granted a patent in 2009 for a floating data center
Sigh, there is a patent on floating data centers? For fuck's sake...

There's a patent for one specific, detailed design for a floating datacenter. Not for the mere idea of a floating datacenter.

I'd be shocked if claim #1 wasn't "floating data center". They always start out super broad.

How does the data get to/from the platform? Microwaves? Can that work from a slightly moving platform?

My wacky guess would have been off-shore offices for employees that can't get H1Bs.

I doubt it; they would have just bought a cruise ship.

Besides, Tata and Wipro have demonstrated that you can just bring people here to work, coop them 8 to a suite in extended stay hotel rooms, and with sufficient lawyering and a couple of hoops, visas be damned.

Wait, what? I think I missed this one

Google's Underwater Cable Ambitions Expand: http://gigaom.com/2009/12/11/googles-underwater-ambitions-ex...

I'd guess fiber-optic cable.

How does data get from the EU to the US? Or APAC to the US?

Last I checked, all those entities were firmly connected to the Earth's crust.

Maybe something like the Canon Canobeam DT-130 Type SFP: http://www.usa.canon.com/cusa/professional/products/free_spa...

But, you know, more than one... or something.

1000 4G dongles could pull down 1 Tbps and the signals run well over several miles of water.

Wireless communication doesn't work that way. Its bandwidth is not proportional to the number of transceivers.

Very interesting...

Any chance that this will mean the Govt can't subpoena data from this data center if it is floating far enough from the US coast? Or does it not matter because clearly the US will be the closest harbor?

In international waters, it matters more which flag your ship is sailing under than which port is closest.

It has to connect to the U.S. somehow. So they can still subpeona data moving through the switch.

There's going to be a cable connecting this to the mainland, which one could argue means the ship is moored to the U.S. coastline?

Might be a fun legal battle.

> So they can still subpeona data moving through the switch

Well if they weren't an american corporation they would be doing this anyway. Besides, perhaps the most important data wouldn't ever have to be transmitted into the country.

On international waters you follow the laws of the country where the ship is registered.

So, most likely Liberian or Panamanian law...

From the pictures, it looks like the ship is registered in the US (though that could change once it is finished).

It's the other way around: The whole block is the subponea treatment unit of Google. It's offshore so it's not under US jurisdiction.

Don't know how this'd work in actual international waters, but no, this is in a bay, nothing international about it.

Its being built at an island in SF Bay. That doesn't mean its going to be operated there.

Would servers be affected by the instability of the ocean? That is, if they use spinning drives, would the heave and ho of the ocean affect it? Would that affect other electronics and components?

I doubt the heave and ho of the ocean would affect spinning disks like hard drives. We used to put hard drives in iPods that got thrown around by people jogging with few failures.

What is a concern is salt water, wind blowing the surface of of the ocean can add salt into the air that will eventually get into the containers. Salt moisture will be a huge long term problem.

The salt issue is something that's probably already been solved (couldn't find any information on it though). There are underwater science labs, medical rooms at sea, and probably tons of research into how to keep salt out of computers from the US Navy.

For a boat that big you're never going to get accelerations higher than a half G or so, and water is squishy so the jerk is going to be very moderate. I expect that frequent acceleration will actually have some impact on how fast the hard drives wear out, but not a large one.

Protecting everything from salt spray is going to be the bigger concern.

This would be far from the first time electronics were put on a boat.

If you are doing business in the US you will play their rules.

So it appears as though Google is looking towards a future where they are an independent country, or at the very least free from regulation, visa-a-vis Seasteading.


No, this does not appear at all. It appears they want to save on real estate and cooling costs by having floating data centers.

Google has owns so much real estate and immovable property in the US it is silly to suppose they are thinking of leaving the country just because they built a couple of barges.

> It appears they want to save on real estate and cooling costs by having floating data centers.

Is this really that expensive that's still advantageous considering the costs of a barge + electricity + data link to the sea + fuel + maintenance ?

Not only is cold seawater usable for cooling, but the article mentions the feasibility of using the sea for power generation.

And interestingly, Patri Friedman used to work at Google...

Patri only worked on prediction markets at google, so far as I know, and left to spend more time on Seasteading and related projects.

Source: he's a friend of a family member.

Everyone's work on prediction markets at Google was a 20% project, so Patri would have had his main job doing something else.

I hope they fit it with sub-surface portholes and webcams.

The article keeps saying "huge", but really that's tiny for a ship. Google could buy an old cruise liner and fit several of those inside. Or maybe Google is going to buy more little ships and plonk one of these in each?

I hope that they have adequate security, because all that gold and copper is worth stealing for some people.

I think they will use a RAIS [0] design.

[0] Redundant Array of Inexpensive Ships

I guess they'd better make the first one not too ambitious and ramp up later.

It's still in beta.

So it'll get deployed, a bunch of people and startups will come to rely on it, then the guy who's driving this as his "20% project" will get moved to another group, and it'll sit un-maintained for several years slowly rotting, then sink without a trace…

I see. we're slashdot now.

Well, actually* it was Stack Exchange that first identified the poison that is metadiscussion.

* tongue firmly in cheek as I knowigly contribute to the problem in the hopes of educating everyone away from it

I wonder if they're taking this approach because it's modular. You can add more barges and shipping containers a lot easier than another cruise liner.

Somebody told them about docker and this is what they did....

Maybe Google will put pressure on the city to upgrade the island's power grid, which has had serious reliability issues in recent years: http://www.sfexaminer.com/sanfrancisco/treasure-island-endur...

This will get lost in the rest of the comments here, and the article was comprehensive, but:

A friend of mine just got back from Treasure Island (she works as a stewardess on millionaires boats), and hung out with a lot of the guys working on this. This is definitely Google's doing, and she was told that it is indeed a data-centre, but that's all the info she got.

another cool thing about floating data centers in california is they're pretty much earthquake proof... built in shock absorbers (aka water).

You can probably charge a premium for easily accessible highly secure data I would imagine.

... assuming megalodon isn't still around

An earthquake can cause a tsunami. Depending on conditions, the tsunami wave could be hundreds of feet high. So, not earthquake "proof."

Tsunamis are not giant walls of water like what you see in disaster movies. They're more like an unending tide: the water just keeps rising until it floods the area. Anything floating loose on the water at the time is at risk of being carried inland, but is unlikely to be completely destroyed.

A floating data center barge anchored a mile offshore is as close as you can get to tsunami-proof. At worst, its connection to the mainland would be severed if the tsunami destroyed the local land-based telecom infrastructure. Any data located on the barge would still be intact until communications could be restored.

Hence why they're also sometimes called "Tidal waves"

No more Hollywood movies for you. The only "tsunamis" with "waves" hundreds of feet high are created by cliffs or glaciers collapsing. And even then, the wave height is really the run-up against another cliff or mountain. Such waves can be dangerous to boats in confined waters, such as a bay. In the open ocean, tsunami waves are mostly harmless and even barely noticeable to boats. And all this is moot because the slip-strike motion of the San Andreas fault doesn't produce tsunamis that I've heard of.

Such as the San Francisco Bay, where this currently is?

Also, the Japanese Earthquake killed a guy in California:


It doesn't need to be a close earthquake to kill people.

The wave of a tsunami remains under water, until it hits the coast and the water isn't as deep as the wave is high. If you are far enough away from the coast, a tsunami has no effect nor will you notice it unless you have the equipment necessary to detect it.

The worst thing a tsunami could do to a floating data center is break the internet connection, assuming a cable is used.

just to slightly build on the other comments, I believe this barge is designed to be out from shore quite a way, if it was right up against the shore I could see this being an issue.

Reminds me [1]. Gmail like AI breaks loose and uses fully automated floating data centers. Highly recommend.

[1] book named "Avogadro Corp: The Singularity Is Closer Than It Appears:

That sounds like the game "Endgame:Singularity": http://www.emhsoft.com/singularity/

I hope it is a floating data center. If not, it is a fun detective type article.

If it is built by Google and it's not a floating data center, I wonder what it could be.

It's the opt-out village :).

Not many units there!

It's Google Island!


Never underestimate the bandwidth of a barge filled with hard drives.

Today we went sailing. I took ~62 higher res jpg pictures of this area.

You can see these here: http://tinyurl.com/floatingcontainers

Awesome Initiative!!!


Maybe this is just so their data centers can be movable between cities so they can negotiate better tax agreements and power contracts.

Or to cover peaks like say the Olympics just anchor it near to where it is needed.

Actual original reporting, wow, sadly unusual these days for any publication. My respect for CNET might be resurrected.

Actually it just struck me this might be part of Googles DR plan for California getting hit by the big one

I can see it already.

In a dystopian future, when the world as we know it is covered in water, one mythical ship traverses the oceans, powered by stratosphere kites or nuclear power. Nobody knows where it's from.

All we know is that many years ago, it became sentient.

And its name is Google.

Anyone consider what the real estate & taxes savings will be with this? For high demand areas where you'd like to set up a data center, something offshore might save a fortune in land costs and property taxes alone.

It likely can't be justified on power and real estate costs alone. There are counties in Oregon that give away land and provide very cheap hydro power for data centers.

I was referring to areas where real estate is extremely expensive, and data demands warrant close proximity thereto - such as San Francisco. Floating a data center on "free" water surface may be cheaper yet still have thick short data pipes to intended customers.

With movable data centers, Google can easily move them across oceans, as popularity and usage of different Google services inevitably shifts across countries over time, in order to be closer to wherever their users are to reduce network latency.

For example an "Orkut data center" would have been very useful to Google in 2008, when they announced they were moving its operations to Brazil because somewhat unpredictably this is where their core user base developed the most: http://en.wikipedia.org/wiki/Orkut

They would also be able to move operations to places where energy is cheaper or sea water temperature is cooler. In general, it could be used to minimize costs looking at a variety of factors. I'm still curious how they will connect to the mainland. A fibre connection would hamper the flexibility of having a boat. Employees would rather live on the mainland than on a boat. So perhaps they the boat is stationary and will enjoy the ample amount of cooling water and potential energy the waves create.

Or you could just shift the processes to a newly commissioned, and therefore cheaper per cpu/ram/storage per dollar!

A completely new data center takes years to build and deploy.

Sure, but so will finding an appropriate cable landing site, getting a mini-substation built near that landing site, etc.

Cable landing is a hard problem.

A massive floating data-centre out in international waters would be a pretty powerful protection against governmental interference in your business. Well, except for the military kind.

Not really.

It's a similar idea but comparing Google putting it's resources behind a similar idea to something like Sealand is not really a fair comparison.

The outcome could be vastly different.

Yeah that last part of your thought is the only important one. Nobody, _nobody_ has better undersea cabling tapping abilities than the NSA.

Ahem, (Unfortunately as a Brit) I think Tempora/GCHQ might be up there with the NSA.

Seems pretty telling to me, with the patent and coffee shop cashier testimonial.

I wonder how the hardware will hold up to the environmental harshnesses of the open water? Or even just sea spray...

I thought that was the best part: corp spends tens of thousands in lawyer time and craftiness to be stealthy, but forgets that it gives corp-branded charge cards to its employees, who won't spend $3.50 on a latte while on company time (and forget to ditch their badges when offsite).

Reminds me of the fake Apple company names and back stories when they were building the iPhone and had to meet with cellular companies. (I guess Steve Jobs was less recognizable back then?)

Just enough leaks to conclude it's a Google-owned project. Although I can't offer an alternative explanation, I'm a bit skeptical in this post-Snowden world.

A datacenter in/near SF doesn't make much sense. Electricity there is just too expensive. And, no, the waves in the bay aren't a solution.

I suppose you might build a datacenter there and then ship it to some third-world port where you don't want to do construction.

While I love the idea of a gmail datacenter in international waters with automatic guns to deter boarders, that really wouldn't fit Google's current political strategy.

I believe this was literally Chapter 1 (ok, maybe chapter 2 or 3) of "Avogadro Corp: The Singularity Is Closer Than It Appears" [1]. Life imitates art, I guess? Or maybe that author is more prescient than I gave him credit for.

[1] http://www.amazon.com/Avogadro-Corp-Singularity-Appears-eboo...

Can somebody explain the point behind the secrecy? What does Google have to gain by taking all these steps to hide they're the ones building it?

So that the reveal can be done on their own timing, only when things are ready, for maximum desired effect.

Trickling details out about something still taking shape leads to expectation problems, and perhaps less favorable coverage (in magnitude and tone) at intended launch time.

> Can somebody explain the point behind the secrecy? What does Google have to gain by taking all these steps to hide they're the ones building it?

It's too big to hide the fact that someone is doing something, so by making obvious efforts to keep it secret, they prompt people to look into it, find clues that its tied back to Google, and give more press to the effort than it would have gotten if Google had issued a press release about building floating data centers.

The same reason anyone hides trade secrets. For competitive advantage, etc...

Perhaps also for legal reasons. For example larger companies use shell companies for purposes of reduced liability.

Perhaps corporate espionage/sabotage? Or maybe to keep things relatively low-key around the site.

Or perhaps the tech/approach is untested, and they want to do some preliminary tests/dry runs before putting their branding all over it.

These security measures are pretty tame in the grand scheme of things though. I think if google did want to completely mask their involvement, they would have at least issued different badges and credit cards to their employees on site.

I think these will probably get floated somewhere within the bay, and they will probably ruin the bay views of some people. They have probably found a way to build them without going to public hearings and want to finish the job before the locals get wind of this and try to pass new laws to prevent their installation.

google is always incredibly secretive about their datacenter technology. it's a big advantage to be able to run a datacenter more cheaply and efficiently than your competitors.

Just wondering what the effect of all the waste heat will be on (what's left of) the Bay's ecosystem.

Google uses Treasure Island for holiday parties, perhaps that's what explains the presence of employees?

This floating data centers could be used for more privacy/security. Very interesting article thanks.

I am having great mental images of pirate ships full of hackers attacking Google floating data centers in international waters. The floating data centers will of course be stoutly defended by a combination of machine gun wielding robots and Google engineers with laser beam interfaces for their Google Glass.

Could this be a routing center for CC payments? That would be a very interesting tax evasion method...

It's likely that this is a way to get around the visa limits the US is having. If you build a giant cruise ship and park it on the bay next to google then you can have engineers working near HQ without legally requiring them to have visas!

The idea of a bunch of non-citizen engineers working (and living?) in a windowless floating cargo container doesn't seem to me to be something that would be a great idea, PR-wise.

Sounds kinda like the fictional Arsenal Gear ship: http://en.wikipedia.org/wiki/Metal_Gear_%28weapon%29#Arsenal...

Could this be an attempt to skirt the rules and laws imposed by the US on our (Google's) data? Would Google have to provide data under a warrant if said data is physically located in international waters?

if the ship is under US flag, yes

International waters, immune to NSA?

Also, does wave power really yield sufficient energy?

Much more likely it belongs to the NSA.

the patent mentioned in the article says 3-7 miles offshore. that's well within governmental jurisdiction.

I wouldn't be surprised if it was a GoogleX project of some sort.

How would a data centre on a barge cope with constant motion (hurtful for cheap spinning discs?) and data link latency? Presumably cable doesn't work well if the barge moves?

> putting data centers inside shipping containers is already a well-established practice.

Is this true? Does anyone know why? The article seems to assume it without saying why.

I'll try to find the link in a moment, but Microsoft was doing something like this 5+ years back. The premise was similar to Google using commodity hardware in their datacenters. Throw together "compute blocks", set them all up and when performance degrades sufficiently you remove the container and install a new one in its place.


Also, I'm surprised I got the time right, for some reason 2005-2012 is one big blur to me and I usually pick the wrong end of that range when guessing when things happened. It also seems that none of these articles mention what I recalled reading in articles years ago about just replacing a box once its performance degrades. I'm either misremembering or that ended up being a non-feature once fielded.


A more recent article, this one is MS in VA: http://www.wired.com/wiredenterprise/2013/02/boydton/

A critique of the approach: http://www.computerworld.com/s/article/9080738/6_reasons_why...

The newest Azure data center in Iowa(?) was going to be open air. Basically a pile of these containers in a field, with a big fence over it and a roof similar to a big car-port (no walls). They figured it was going to be better for cooling.

Iowa? Doesn't sound like a great idea, they have tornadoes there, which can easily throw shipping containers around.

Yep, outside Des Moines. I don't have a reference for the rest of it, that was from conversations with MS employees.


joezydeco: Your reply seems to be dead, I'm not sure why.

I think that's the same one the datacenterknowledge.com article I linked refers to though.

In addition to the links below, Raytheon has deployed many 'Electronic Modular Enclosures' which house power/cooling/servers enclosed in a TEU'ish container. They've been in use since 2005'ish in the DoD and brought up in the recent 'Navy ship uses Linux' story from last week.

A previous employer used a TEU contained power/cooling/servers for mobile disaster recovery circa 2003. More recently some have used fuel cells in place of diesel generators.

They're really easy to add, expand, replace. Google's been using them for many years.

http://en.wikipedia.org/wiki/Modular_data_center http://en.wikipedia.org/wiki/Google_Modular_Data_center

Yes it is common.

eg. https://www.youtube.com/watch?v=zRwPSFpLX8I

A shipping container is a standardized container. It's like designing a PC case. Standard sizes makes things very convenient.

It's like Legos and interchangeable computing parts. But bigger.



(Don't downvote me. I'm obviously joking.)

A place for those with no visa to the USA to work!

I can just see google hq now.

"Boy that was quite a storm last night...oh wait...gmail just floated away and is halfway to hawai'i"

It might be a NON-HULKING floating data center.

This could be part of Google's Project Loon. Instead of building a data center in Africa why not ship one over.

It's being built at the same hangar where they shot the Battlebots TV shows.

Bruce Sterling will love this.

Well looks like we're going to need to Trevor to do some investigation.

Maybe they're just trying to lay some underwater cable... to Japan

How do you get a patent for a "floating data center"..

Submit it to USPTO and other national or trans-national (e.g. EU) patent registries like any other patent.

This one's included in the article: https://www.google.com/patents/US7525207

Satire aside. My point was the ridiculousness of it. What's next, patents covering putting data centers in space, under the sea or under stairs etc..

There's a Chinese patent on putting a data center in space [0], and a US patent for a partially submerged data center [1], but I can't find anything for putting a data center under the stairs... you should patent it!

[0] https://www.google.com/patents/CN103186179A?cl=en

[1] https://www.google.com/patents/US20130044426

Looks like the patent disclosure refers extensively to challenges and opportunities of the maritime environment - power generation, heating and cooling, etc.

This is precisely how Feynman got the patents on the nuclear submarine and the nuclear airplane.

Who has jurisdiction?

That or a metal gear. My hopes are on the latter.

nice information regarding google building hulking floating data center.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | DMCA | Apply to YC | Contact