I at first wondered if the scene the other day at the Energy Department’s Oak Ridge National Laboratory was something like the scene in “National Lampoon’s Christmas Vacation,” when Chevy Chase caused the local power plant to tap into emergency nuclear power when his Christmas lights came on.
That’s because the Oak Ridge crew flipped the switch on Titan, the world’s newest and likely fastest supercomputer. Titan is capable of an amazing 20,000 trillion calculations each second. That’s 20 petaflops if you are up with the lingo.
But it turns out the strain on the local power plant wasn’t anything out of the ordinary, because Oak Ridge designed Titan for relatively low power consumption.
One secret to Titan’s success is that it uses graphical processor units in conjunction with standard processors. The GPUs are familiar to gamers, because they are needed to be able to play bleeding edge titles like Call of Duty. Titan contains 18,688 nodes, with each holding a 16-core AMD Opteron 6274 processor and an NVIDIA Tesla K20 graphics processing unit GPU accelerator. And the 700 terabytes of memory doesn’t hurt matters any.
Combining GPUs and CPUs not only gives Titan its amazing petaflop power, but is also more economical in terms of power use. "One challenge in supercomputers today is power consumption," said Jeff Nichols, associate laboratory director for computing and computational sciences. "Combining GPUs and CPUs in a single system requires less power than CPUs alone and is a responsible move toward lowering our carbon footprint. Titan will provide unprecedented computing power for research in energy, climate change, materials and other disciplines to enable scientific leadership."
Titan, an upgrade of Oak Ridge’s Jaguar supercomputer, marks a new architecture design path for energy-efficient supercomputers. Titan is not only 10 times more powerful, but it takes up the same space Jaguar, while only using a tiny bit more power.
Jaguar had been the fastest U.S. supercomputer on the Top500 ranking, but at a clocked speed of 1.75 petaflops before the upgrades it trailed supercomputers in Japan (the K Computer, which clocked 10.51 petaflops) and China, as well as the newest champ, Energy’s Sequoia supercomputer, an IBM BlueGene/Q system launched earlier this year at Lawrence Livermore National Laboratory with a sustained speed of 16.32 petaflops. Sequoia also uses a more energy-efficient design.
I realize that Titan will be invaluable for climate research and all the fields of physical sciences Oak Ridge will use it for. But I have to say, what I really want to do is play some cool shooters with it. After all, we don’t want all those GPUs to go to waste, do we?
Posted on Oct 30, 2012 at 10:01 AM3 comments
Most public-sector employees in the Mid-Atlantic and Northeasternm states are likely going to be at home for a few days with the government closed and Hurricane Sandy bearing down like a 900-mile wide missile getting ready to strike. As I write this blog, the first insidious bands have already started to snake their way into the D.C. area, and the worst is yet to come. If you have a smart phone or Twitter device, there are ways you can get information to help you get through the seemingly massive weather coming our way.
In previous storms like 2010’s “Snowmageddon”( all the really bad ones get funny names; this one, of course, goes by Frankenstorm), I learned that cutting-edge technology can sometimes fail at the worst possible times, which means falling back on the least common denominator to keep yourself safe and informed. I even wrote a book about how that exact thing happening on a massive scale during Hurricane Katrina in New Orleans.
While I would always recommend an emergency radio and plenty of fresh batteries, Twitter is becoming a great source of news too for folks hunkered down and trying to get information. So long as you have remembered to charge up your smart phone, there are lots of helpful Twitter feeds you can get, even without power. For Frankenstorm in particular, NOAA’s National Hurricane Center is pumping out the latest news, minute by minute. A quick glance shows that Advisory 29A has just been issued, stating that the storm is turning faster and harder than originally expected, and zooming towards landfall in Southern New Jersey.
The FEMA twitter feed is also on fire today, with posts coming every few minutes. They have very helpful advice on how to avoid things like carbon monoxide poisoning when using a portable generator. It’s required reading for smart survivors.
And the governors of the various states that are affected by the storm are also tweeting helpful advice, depending on the official. New Jersey Gov. Chris Christie just sent: “Don’t be Stupid. Get out.” But there is also a lot of other information like what bridges and tunnels are closed for those trying to follow his blunt advice.
CNN has a pretty detailed list of helpful Twitter feeds that are covering the storm. A lot of them are filled with striking pictures and non-mission critical feeds, and several are outside of Twitter, but their list also has many of the best ones to use.
I think with Twitter being so helpful, that including a device capable of getting the feeds, plus extra batteries, is a good addition to any survival kit. And unlike a radio, it also lets you communicate back to the outside world, letting everyone know your plight if you so choose.
I’ll try to tweet what’s happening locally to our GCN Lab Guys feed in case you want to see how we are handling this one. Above all else, stay calm. We’ll all get through this together.
Posted on Oct 29, 2012 at 8:52 AM0 comments
I rarely advocate people breaking the law, or even the rules, but this is a special case. Last year GCN contributor Carlos Soto penned an instructive piece about how government employees could use Facebook and Twitter without getting fired.
In it he advocated giving Facebook some specifically incorrect information, such as an incorrect date of birth, which one person who commented pointed out was against Facebook policy. If you do that, you are basically agreeing to a terms of service with Facebook, and then breaking it out of the gate.
But if you are government employee, I think it’s one of only two choices that you have now. It’s either do that, or don’t use the service at all.
Wired recently reported that Facebook is close to agreeing to a settlement with users who were part of its experimental “Sponsored Stories” program. In that program, each time a user clicked on “Like,” Facebook could use that user’s picture and profile in an ad for whatever was clicked on -- without permission and without compensation. Facebook could say “Joe Smith, a computer programmer with the Department of Commerce, likes Happy O’s Cereal!” and show his profile picture (assuming Happy O’s paid Facebook to do it, of course). Now if Happy O’s happens to be made in China by slave laborers, Joe might be in trouble with his bosses. For that matter, regardless of where Happy O’s are made, government employees aren’t supposed to endorse products.
People sued Facebook over Sponsored Stories and won. Each person used in the ads could get up to $10, depending on how many people fill out the class-action-suit forms or how much the lawyers siphon from the till.
But get this: Facebook isn’t actually planning on changing its way of doing business. Instead, the company plans to make every user agree that it can do what it wants with its users Facebook information. Facebook isn’t saying what it did was wrong, which it clearly was, but instead just getting its legal ducks in a row so it can exploit users for cash after the trial.
Facebook’s proposed new agreement states: “You give us permission to use your name, profile picture, content, and information in connection with commercial, sponsored, or related content (such as a brand you like) served or enhanced by us. This means, for example, that you permit a business or other entity to pay us to display your name and/or profile picture with your content or information. If you have selected a specific audience for your content or information, we will respect your choice when we use it.”
In other words, if you “Like” something and happen to be pseudo-famous, well known in your community or if you work for a respected organization like government (and have this in your profile) Facebook can exploit you each time you click a “Like” button. And there is no way to opt-out of the program, other than not using the service.
Soto’s suggestion to give the company inaccurate information could help protect you in some cases, though it also defeats the purpose of using Facebook. And if Facebook found you out, they could eliminate your account.
Facebook’s seemingly desperate attempt to start making money from its users way really rubs me the wrong way. If the company persists, it seems it might be best if government employees didn’t use Facebook at all.
On Twitter, I follow many feds through the @GCNLabGuys account I share with Greg Crowe. Every one of them either says that the feed is their personal viewpoint or an official posting, and almost all add that if they follow someone, it doesn’t count as an endorsement.
Unless Facebook allows feds to do something similar, we could see a mass exodus from the service for government employees more worried about losing their jobs than which of their friend’s children are learning to climb the monkey bars. And I don’t blame them one bit.
Posted on Oct 25, 2012 at 3:10 PM1 comments
Java, especially for government, is the devil you know. Although many people, including GCN security expert William Jackson, questions whether the problems with Java, including recent high-profile zero-day security exploits make it more trouble than it’s worth, it still runs on everything from televisions to PCs to cell phones, and is a part of every modern browser.
Of course, if anyone is willing to take on the devil, it’s Google. The search giant may do so by building its own demonic host, in this case a programming language developed in secret, called Dart. Google has added a lot into Dart in an attempt to lure developers away from Java. The Dart project includes a modern scalable language, libraries and tools to help developers build large complex Web applications.
The part about the Web applications is really the driver for Google. The company lives and dies on its ability to use and manipulate the Web, and even Java’s strongest supporters acknowledge that large-scale applications were not originally considered in its development. Those of us who have learned how to code Java know that, compared with something like C++, it’s a pretty simple language, which is why it can reside comfortably on your phone or in your car’s dashboard. But for big projects, it needs to be jury-rigged. Google’s main problem with Java is that it makes Google Docs run slower than Google thinks it should.
Microsoft reportedly isn’t happy with Dart’s appearance. Redmond officials argue that, should Dart catch on, a whole new language would need to be added to browsers, which would open up new security holes. They say that flavors of Java exist that could be used for large-scale applications without totally tossing the nearly ubiquitous language out the door. Groovy, for example, could prove to expand Java’s capabilities while also making it easier on programmers.
Will Dart hit a bull’s-eye with programmers in government? I don’t know, but Google has taken aim toward that target. The Dart software development kit is now available for download. With many public-sector agencies moving to Google Apps for Government, a language that makes Google Docs run better might catch on.
My advice would be to get it, try it out and see if what Google promises will work for your organization. The language itself is pretty similar to Java, so you should be able to make an informed decision in no time.
Posted on Oct 23, 2012 at 9:57 AM4 comments
As GCN’s watcher of all things in the emerging technology field, one surprising fact I discovered pretty quickly is how many advanced robots exist in our world, or soon will. Robotics is a field on the bleeding edge of technology, with new discoveries and advances made every day. And government is on the forefront, or at least has a hand in driving, most of these innovations.
Just as a quick recap, we have the Defense Advanced Research Projects Agency working on the next generation AlphaDog, which is a sort of glorified pack mule, but one that could make a soldier’s life a lot easier once all the bugs (fleas?) are worked out.
Then we have the National Oceanic and Atmospheric Administration and the Navy working with a company called Liquid Robotics on swimming robots to map the Arctic Ocean, a project I’ll write more about soon.
Then we have the 501 Packbot, which is used by the military as well as by state and local governments for bomb disposal and patrols. The company that makes it, iRobot, is giving it a step toward actual intelligence. Interestingly enough, iRobot is the same company that makes those little disks that zoom around the floor vacuuming up debris on their own. I’m not sure I like the idea of those things thinking for themselves, especially when mine nuzzles up near my toes. Then again, some extra smarts might give it a fighting chance against my cat, which is always plotting its demise.
This week I learned about a robot being developed jointly by NASA and the Florida Institute for Human and Machine Cognition that nobody, not even my cat, could dislike. The X1 is one of the new exoskeleton type of robots that fit around the human body. Literally meaning an external skeleton, the exoskeleton’s secret is that it not only can move along its various joints, but can also restrict movement.
That’s really important because it’s the formula humans use to walk and to stand. We have to be able to move our legs to go anywhere, but we also have to hold our shape to prevent falling over once we get there. The X1, once perfected, could help wounded soldiers or those who have lost their mobility due to disease or accident, to walk again. It could also be used in therapy to help people rehabilitating from an accident.
NASA is interested in the X1 because it can make astronauts stronger in space. Remember when Ripley beat up the alien queen in the movie “Aliens”? That could never have happened without her trusty exoskeleton. A robotic exoskeleton also is what Iron Man uses to fly around in.
In reality, there is already a robot in space, the Robonaut 2, which resides on the International Space Station. not an exoskeleton, but a fully functional human-looking robot (though not to the extent of Data from “Star Trek”) that is controlled remotely and runs on 38 PowerPC chips and 350 sensors. The X1 owes a lot of its technology to Robonaut, though in a lot of ways it’s simpler because a human is actually there with the robot running things, not in a remote location. The operator can use his or her own senses to keep the X1 from walking into walls.
NASA's video showing a man in a wheelchair using the X1 to stand up and walk across a room is pretty compelling to watch. Right now, the person manipulating the X1 still must use crutches for balance, but that could change in future versions of the robot.
The X1 currently has joints for leg movement, and plans are in the works to add powered and controllable joints for the ankles and around the hips. That should provide much greater range of movement, and walking without crutches or any other aid. As the technology gets smaller, it’s even possible that someday people could wear these suits under their clothes and walk normally without anyone even knowing about their disability.
NASA seems pretty happy with the progress so far. Michael Gazarik, director of the agency’s Space Technology Program, issued a statement saying, "It's exciting to see a NASA-developed technology that might one day help people with serious ambulatory needs begin to walk again, or even walk for the first time. That's the sort of return on investment NASA is proud to give back to America and the world."
I’ve talked to robotics experts and almost all of them have said that walking naturally is one of the most difficult things for a robot to accomplish, which is why most robots have wheels or tracks instead. That’s because there’s a lot more going on when we step across a room than we realize: thousands of “readings” per second are made with respect to balance and position. Some muscles have to contract and move, while others need to hold in place to prevent falling down.
If NASA’s X1 can accomplish all that, developing other tasks such as arm movement or turning a head will be like child’s play by comparison. I for one wish the designers all the luck in the world, and look forward to reporting on their continued success with this most noble of projects.
Posted on Oct 16, 2012 at 9:22 AM1 comments
I used to love working in computer security. Suiting up in my glowing armor, loading up my weapons, and heading out on patrol in cyberspace, ready to do battle with all types of hackers and malware was all in a day’s work.
I know what you’re thinking, that it sounds like a fantasy based on the movie “Tron.” Real IT security involves watching streams of completely uninteresting data and reports or, more likely, listening to some user complaining about his computer acting funny.
But it doesn’t have to be like that. The folks over at Lincoln Laboratories at the Massachusetts Institute of Technology think that it’s possible to make computer security more like a video game, or more like the “Tron.” And they say that doing so would actually make security more efficient, bringing the human equation into the mix.
The researchers lay out their plans in the NewScientist. Basically, the MIT team took all that data and put it into a graphical interface, like you might find inside a shooter video game. Maps of an actual physical infrastructure were combined with network topology data and fed into a gaming engine called Unity, which rendered it all in 3D.
Then engineers were assigned to different parts of the mapped network. When suspicious activity was detected, the security personnel could warp, or simply run, over to the troubled node using either a keyboard or a video game controller. Messages can be sent to other guardians as well, in case backup is needed, or to check out nearby systems for signs of the infection.
Once the virtual/real IT folks got on the scene, the protectors of the network can use tools to eliminate the threat. Eventually, the MIT team wants to incorporate anti-hacking programs into the game interface, so that running an AV scan might be like drawing a pistol or a sword. The idea is to make sure that those tasked with defending a network stay focused.
Plus, with a staff of real IT humans on the job, hackers won’t be able to block the defenders, since they are essentially real people interacting with the virtual environment. You can’t trigger a stack overflow against a real person, after all. The system has been tested on a 5,000-node computer network, and the results were encouraging, according to the creators.
I can’t help but think that public sector agencies could benefit from using this type of system, and actual human resources, to combat computer-based threats. For one thing, their systems — from military and federal civilian to state and education — are constantly under electronic siege in one form or another.
For another, forecasters expect a shortage in the IT security workforce in the years to come. Public/private groups are always looking for ways to attract young talent into the growing list of cybersecurity programs at universities. Making the defense of a network into something of a game might just get a few people interested.
Plus, I really want to load up and gun down some malware, gangster style.
Posted on Oct 12, 2012 at 12:55 PM3 comments