Slashdot Mirror
Startup Building Floating Data Centers
1sockchuck writes "A Bay Area startup is planning to build data centers on cargo container ships, which would be docked at piers in major Internet markets. The company, known as IDS (International Data Security) says it plans to use biodiesel to power its generators and use heat from equipment to manage temperature on board the ships, reducing their reliance on grid power. IDS is telling prospects that it hopes to eventually have more than 20 floating data centers docked at ports around the U.S."
wouldnt this leave them far more open to forms of terrorism? i.e. if these floating data centers hosted say, all the websites that godaddy.com host (which is alot), and someone "cut the cable" which would be alot easier to find on a ship, since it has to come out of the ship somewhere... all these websites would instantly go online, where in a building, the cable would come in, underground, directly into the rooms the data center occupies. ships are easier to sink that buildings are to destroy.
if the ships use wireless rather than wired, there would need to be a large antenna on the ship, which would:
1. be a target for everyone
2. allow people to intercept any connection.
portfolio
Why bother with biodiesel? Cargo ships use bunker oil, which is 1 step up from crude. They'll already have massive generators and massive fuel capacity, with readily available fuel.
If they really wanted to be green they'd deploy some sort of thermal gradient generator, sinking piping down below the thermocline of the ocean.
I want to delete my account but Slashdot doesn't allow it.
The Navy is not exactly hurting for money, and they justify the expense since the electronics are located near its users. This venture is needlessly placing the data center on water, when the data users are mostly land based.
You will have more options on land. First of all, why place the containers on a ship when a container yard will do? Need to move the data centers to another location... Hire a truck!
You are looking at least a 48 to 72 hour downtime (if you are lucky). Being on a large container vessel (TFA is talking about decommissioned container ships), you will need to sail far enough away from the hurricane. Keep in mind the current state of hurricane predictions, the time it takes to disconnect from shore, scheduling a bar pilot, tow, bunkering, and sailing to destination. Once you reach the destination, waiting for bar pilot to board, tow, mooring, and making data connections to shore...
You could have co-located your data center in another region and switched to them during your emergency... Save the expense of vessel movement and the additional risks involved in ocean transportation. Better yet, use a container and truck your data center to another location further inland... Container based data centers are a neat idea, Container shipped based data center is an idea that went too far.
These comments are my own and do not necessarily reflect the views or opinions of my employer or colleagues...
I agree with you to a certain extent, but I think it comes back to economies of scale. When you factor in maintenance costs for ocean-based vessels, on top of the fact that many land-based data centers are now being built in areas with many cost advantages (being built near large quantities of dark fiber, being built on cheap land, built near energy sources or near areas where renewable energy sources are/will be available, being built to minimize maintenance costs on the infrastructure itself, etc.) I'm not sure that a floating data center is going to be cost-effective.
Not only that, but the vulnerabilities on a floating data center are going to be, at a minimum, the exact same if not higher. The data connections are going to be exposed, no matter what you do. At least in a land-based data center the fiber is buried, and less obvious. Not only that, but physical security can be made (and is made) relatively difficult in a land based data center, at least directly surrounding it. A floating entity would have a far higher risk of approach due to the traffic that occurs on the water near a port.
As far as exposure to attack on fiber, well, I can tell you the exact man-hole in the city where I live that you can toss an IED and take down virtually every carrier in the market. Those don't exist simply at data centers, but everywhere, due to decisions decades ago that have caused choke points in fiber distibution.
I'd also question the ready availability of dark fiber at a port from multiple carriers. While I don't know as I've never looked at it, it seems to me that ports, being quasi-government facilities, probably weren't wired with fiber with multiple carriers, let alone all the big carriers, as most major data centers are. You might have 2 or 3, but I'd guess that's under contract and the number of directly available carriers is still low. This is a disadvantage for a vendor-neutral data-center that would likely want/need connectivity to all the big boys to entice customers. This might not be true if the port is a launch point for inter-continental fiber, but certainly that's not the case at most ports.
Bill
Japanese electronics companies used to have cargo ships with mainframes which performed data processing tasks in international waters near San Francisco. (Back when Japanese labor costs were low, and computers were rare enough that there was benefit to having a mobile computer.)
San Francisco was nearly completely leveled a couple of times in the 20th century alone by earthquakes.
Wha? The '89 Loma Prieta earthquake caused some serious damage but "nearly completely leveled" is a bit of a stretch. And the 1906 disaster was caused by lack of modern building codes and fire protection as much as anything else. Other cities of that era suffered similar disasters without an earthquake as the root cause (Chicago, for instance.)
There is no reason a data center built from the ground up to survive an earthquake would be less safe on the ground in SF than it would be on a ship in the Bay. You'd be more worried about damage to external things the center relies upon, such as the power grid.
The obvious reason to build on a ship is that the cost of the real estate needed to build a data center the size of a cargo ship in downtown SF would be astronomical.
It weighs about 175 million pounds. Take it out into the open seas where there are 3-foot waves, or actually big enough waves to lift and drop the ship by three feet say every ten seconds. By my Excel calcs, if you use that lift to heave up on a big anchor half the weight of the ship, that's about 30 megawatts of electricity. Plenty enough to power tens of thousands of servers.
The front boiler and engine room spaces of the QM were cleared out long ago, leaving a huge open space for lots of server racks. All you have to worry about is shipwrecks and hurricanes and the effects of humid, salty and diesely air.
I went to school at the University of San Francisco. I worked in the Registrar's office. Every Friday we would make back up tapes of all the school records and drive them up to the a storage facility near Tahoe so that they would survive in case of an earthquake.
This was only 6 to 10 years ago, so obviously there are still some people who think that "modern building codes" don't cut it for earthquakes and are willing (or legally required, like we were) to take some pretty expensive countermeasures.
The real estate cost thing is a good point, though.
I think the real cost savings would be in heating, though... I mean, your entire data center is floating in a pool of coolant, all you have to do is pump it through radiators. It would not be glacial, but good enough. The summary seems to chalk this up to environmentalism, but cooling data centers costs big bucks...
weirdest thing I ever saw: scientology advertising on slashdot.
But those processors mostly keep burning more and more electricity even though they take less space, so these days the problem is electricity, including per-square-foot density problems and total power demand including cooling, which is why companies like Google have been looking at locating data centers in places where power's cheap.
It's amusing how we keep recycling technical problems. Virtualization has been one of the main buzzwords of the past few years, but it's really just a way to re-invent timesharing, using 2000s microprocessors instead of 1970s minicomputers. A decade and a half ago, if you wanted to locate computers and datacomm gear in a telco office to reduce your communication costs or make your services more reliable, you went through a big detailed study on how much power and cooling you needed, how many square feet, how many phone and data lines, etc. Within a few years the hosting market had evolved enough that we knew that a standardized customer network looked like 19" rack-mounted PCs and Cisco routers, and the power and cooling needs per square foot were pretty much the same for everybody, and it changed a bit with 1U servers but we could still usually stretch the available power. But now? We're back to servers that are increasingly customized and non-standard (disks vs. routers vs. blade servers are much different power densities, etc.)
Bill Stewart
New Fast-Compression-only CPR http://preview.tinyurl.com/dy575ks