cluster
Latest
Ditching RAM may lead to low-cost supercomputers
Many servers, supercomputers and other monster systems thrive on high-speed RAM to keep things running smoothly, but this memory is wildly expensive -- and that limits not just the number of nodes in these clusters, but who can use them. MIT researchers may have a much more affordable approach in the future, though. They've built a server network (not shown here) that drops RAM in favor of cheaper and slower flash storage, yet performs just about as well. The key was to get the flash drives themselves (or specifically, their controllers) to pre-process some of the data, instead of making the CPUs do all the hard work. That doesn't completely close the speed gap, but the differences are virtually negligible. In one test, 20 servers with 20TB of flash were about as fast as 40 servers with 10TB of RAM.
Facebook launches shared albums, officially making Cluster obsolete
Last week I wrote an article about how pointless the recently released photo-sharing app Cluster -- which raised a ridiculous US$1.6 million in funding -- is, given that the vast majority of its features are already mimicked on more popular services, including Facebook. Today, in an interview with Mashable, Facebook's Bob Baldwin underlined my sentiment in the best way he knew how: by announcing that the social network has just launched its own shared photo album feature. Facebook's shared albums work much like the standard albums that you've likely been using for many years now, but with a few key tweaks. The album creator can invite as many as 50 friends to contribute content to the album, which is a large enough number that nobody should realistically be left out. Varying levels of privacy can then be set based on your needs, allowing the photos to be browsed publicly, by friends or by just the contributors. As far as mobile functionality goes, users will be able to contribute to shared albums automatically, but the ability to actually create these social image portals will be added in the near future. For now, the shared album feature has been rolled out to a limited number of English-speaking Facebook users, which will be expanded to all English-speaking users before spreading worldwide.
Do we really need yet another photo sharing app? No, we do not
Yesterday, photo-sharing app Cluster announced it had scored US$1.6 million in seed funding from a variety of parties, including Instagram investor Steve Anderson. Cluster v1.0 also hit the App Store yesterday, and as it swims in a sea of "me too" photography apps, you have to ask yourself: What's the point? Cluster claims to specialize in one very specific area: Helping people collect photos from events they may or may not have attended. The app's website presents scenarios like "Brenden, Rizwan, & Taylor collected 1,500 photos from their trip to Europe," and "7 friends created an album from their 250 music festival photos," suggesting that Cluster is the easiest way to collect the visual evidence of each event. You upload photos to a "cluster," invite friends either via email, Facebook or by handing out a special invitation code, and let the rest of the group populate the cluster with their own snapshots. You can comment on photos, tag favorites and share your favorites with Twitter, etc. In short, it's the photo section of your Facebook account, with fewer features. Let me say, for the record, that Cluster is a beautifully designed app and nails the aesthetics of what a semi-social photo app should be. That said, there's virtually no reason why this app should even exist. It's a classic case of a product in search of a problem to solve, and launching exclusively on the iPhone (web and Android versions are on the way) makes its existence even more puzzling. First, let's tackle the Facebook argument. Facebook is already the most widely used way to share photos with friends, and it's still the easiest. You can share photos in a very basic fashion or organize them into albums and then tag the individuals that the pictures will interest. So, say you take a hike with a pair of friends and want to share the photos. Would it be easier to grab each of those friends in the Cluster app, create a cluster, force everyone to download the images separately and then re-share them with the social circle of your choosing, or upload the photos to Facebook, tag your friends in the album and -- oh, wait, that's it. Go make yourself a drink. But maybe you don't want to share the photos on Facebook. Maybe you just want to have the photos yourself. You want to collect all of them to hold close to your heart forever. Fair enough. You know what's great for that? Almost every other photo-sharing app that has ever been released. But even if the iPhone didn't already have Bump, Shutterfly, Path, Flickr and a litany of other photo-sharing options available, there's Photo Stream. This is where the whole iPhone-exclusive launch part makes me scratch my head. Using Photo Stream -- which is built into both iOS 6 and Mountain Lion -- you can set up a cloud-based album where you can upload photos, invite others to view them and notify them when photos have been added to the stream. You can favorite, comment and share photos there as well. Sound familiar? It should, because it's exactly what Cluster does, only it's already been included in Apple's mobile and desktop operating systems. Oh, and if you happen to have already filled up your iCloud storage space, it's fine, because shared Photo Streams don't count against your storage limit. Nifty, huh? [Note: iOS 7 updates Photo Stream to allow multiple users to push content to a single album, including video, which will only further trivialize the need for third-party apps such as this.] This isn't just a problem with Cluster, of course, and it seems like a new half-baked social network or photo-sharing app comes out every week, but at some point developers are going to have to stop seeing opportunities where none exist.
IBM Roadrunner retires from the supercomputer race
For all the money and effort poured into supercomputers, their lifespans can be brutally short. See IBM's Roadrunner as a textbook example: the 116,640-core cluster was smashing records just five years ago, and yet it's already considered so behind the times that Los Alamos National Laboratory is taking it out of action today. Don't mourn too much for the one-time legend, however. The blend of Opteron and Cell processors proved instrumental to understanding energy flow in weapons while also advancing the studies of HIV, nanowires and the known universe. Roadrunner should even be useful in its last gasps, as researchers will have a month to experiment with the system's data routing and OS memory compression before it's dismantled in earnest. It's true that the supercomputer has been eclipsed by cheaper, faster or greener competitors, including its reborn Cray arch-nemesis -- but there's no question that we'll have learned from Roadrunner's brief moment in the spotlight.
Custom enclosure designs shove 160 Mac minis into a single rackmount tower
When the Xserve joined the great server farm in the sky, Mac-minded datacenters everywhere were left without a true rackmountable computer; even the current Mac mini wasn't designed for those kinds of tight spaces. Enter the purposefully anonymous Steve, who just filled the gap with one of the cleverer solutions we've seen yet. He and vendors have developed custom 1U shelving, cooling from car radiators and four-in-one power cables that, combined, fit 160 Mac minis (and a managing Xserve) into one enclosure without cooking the machines to death. With each Mac mini carrying a quad Core i7 and an SSD, Steve now has twice as many cores (640) as an equivalent Xserve cluster despite lower power consumption and a 45-second, network-controlled reboot -- all big helps to his unnamed employer's software development, even with the lack of built-in redundancy for Apple's tiny desktop. As many gritty details as Steve can share are available at the source.
Sandia Labs' MegaDroid project simulates 300,000 Android phones to fight wireless catastrophes (video)
We've seen some large-scale simulations, including some that couldn't get larger. Simulated cellular networks are still a rare breed, however, which makes Sandia National Laboratories' MegaDroid project all the more important. The project's cluster of off-the-shelf PCs emulates a town of 300,000 Android phones down to their cellular and GPS behavior, all with the aim of tracing the wider effects of natural disasters, hacking attempts and even simple software bugs. Researchers imagine the eventually public tool set being useful not just for app developers, but for the military and mesh network developers -- the kind who'd need to know how their on-the-field networks are running even when local authorities try to shut them down. MegaDroid is still very much an in-progress effort, although Sandia Labs isn't limiting its scope to Android and can see its work as relevant to iOS or any other platform where a ripple in the network can lead to a tidal wave of problems.
Supercomputer built from Raspberry Pi and Lego, managed by humans rather than Minifigs
If you're a computational engineer, there's no question about what you do with the Raspberry Pi: you make a supercomputer cluster. Researchers at the University of Southampton have followed their instincts and built Iridis-Pi, a tiny 64-node cluster based on the Raspberry Pi's usual Debian Wheezy distribution and linked through Ethernet. While no one would mistake any one Raspberry Pi for a powerhouse, the sheer number of networked devices gives the design both some computing grunt and 1TB worth of storage in SD cards. Going so small also leads to some truly uncommon rackmounting -- team lead Simon Cox and his son James grouped the entire array in two towers of Lego, which likely makes it the most adorable compute cluster you'll ever see. There's instructions to help build your own Iridis-Pi at the source link, and the best part is that it won't require a university-level budget to run. Crafting the exact system you see here costs under £2,500 ($4,026), or less than a grown-up supercomputer's energy bill.
IBM cluster powers Murchison Widefield Array's radio telescope, answers mysteries of the universe faster than ever (video)
Radio telescope operators have as much of a problem coping with the avalanche of data as getting that information in the first place. The Victoria University of Wellington is all too aware and is leaning on IBM for a powerful (if very tongue-tying) iDataPlex dx360 M3 compute cluster to sift through the deluge at the upcoming Murchison Widefield Array. Combined, the 4,096 array antennas probing deep space and solar atmospherics will have the Xeon-based cluster tackling signal data to the tune of 8GB per second, and about 50TB per day -- that's a Nexus 7's worth of astronomy faster than you can sneeze, folks. A 10Gbps network connection will feed the results to Perth to save scientists a roughly 435-mile trek. Construction is still in mid-stride, but the $51 million Australian ($52.2 million US) being spent on the Murchison array may be worthwhile if it helps solve the riddles of star formation and solar flares.
48 PandaBoards chained together in solar-powered ARM cluster
Michael Larabel already had a 12-core PandaBoard-based mini-cluster under his belt. Clearly, the only way to outdo that is to go bigger, better and greener. The Phoronix founder took 48 of the OMAP 4460-powered boards, got them up and running on Ubuntu 12.04 and chained them together in a massive ARM cluster of Linux goodness. Even with 96 cores chugging along at 1.2GHz the cabinet of tiny computers used only 200 watts -- a threshold Larabel was able to meet with a solar panel strapped to a handtruck. Sadly we don't have any performance figures yet, but MIT, where the little ARM experiment was conducted, should be releasing benchmarks and video soon enough. In the meantime, hit up the source link for some more details and photos of this 96-core, solar-powered wonder.
The Road to Mordor: Thoughts on Update 5
Update 5: Saruman's Obviously Not Compensating for Anything With That Tower will undoubtedly be the last hurrah of Lord of the Rings Online -- this year, that is. It's an interesting update that looks to patch in the rest of Rise of Isengard's content (namely, the instance cluster) and provide a little something-something for players who have already reached the end of the epic storyline and are looking for more. Earlier this week I took a dev tour through some of the main parts of the update, and while I wasn't able to see all of it (such as the non-raid instances), what I did preview certainly filled my head with opinions, analysis and further questions. Generally I came away pleased with what we're going to experience next month, although the comments section of the tour showed split feelings on what's in store for the patch. Is this a case of too little, too late for those dissatisfied with Rise of Isengard's release? Will it breathe new life into dungeon-running, especially for the many soloers out there? Is it what this game needs right here and now? Hit the jump and I'll give you my honest thoughts on it all, both the good and the bad.
Mongoose Studio's dozen iPod cluster display is an expensive way to watch Tron (video)
Most of us would be happy to get a single video to play on our PMP without having to run it through some sort of transcoder first, but the folks at Mongoose Studio needed something more complex. They've released a video of a project that's been in the works for some time, clustering 12 iPod touches into a sort of bezel-riffic widescreen display. All are controlled by a master, 13th iPod that can cause them to display a clock in the interest of verifying synchronization (which is far from perfect, as you'll see in the embed below), or to trigger the playing of a movie. We're guessing that the footage must be manually split into appropriate files for each device, and we're also guessing that horrible things would happen should someone come along and re-shuffle them. But, if you have a lot of friends with iPhones and a lot of free time to prep the film, this could make for a rather interesting movie night -- until someone gets a call, that is.
EVE Evolved: EVE Online's server model
Almost any time a discussion about EVE Online comes up, one way or another we end up talking about the server. EVE Online is unique among today's most popular MMOs for its single-server approach. While most MMOs deal with large number of users by starting up large numbers of separate servers with identical game universes, EVE maintains only a single copy of its game universe on a massive cluster of servers. CCP's decision to go with a server model that doesn't use any sharding or instancing whatsoever has had a major impact on in-game activities and how the game has developed.Server woes:Unfortunately for CCP, maintaining their vision of a single game universe has proven a lot more difficult and costly than anyone anticipated. Working with IBM, the EVE server cluster is maintained in London and is currently the largest supercomputer employed in the gaming industry. Even with this massive power behind the EVE universe, there are still problems as CCP tries to keep the server upgraded ahead of its ever-expanding playerbase.In this article, I discuss the unique gameplay that is possible thanks to EVE's server model, the problems the server currently faces and what CCP is planning to do about it.
Microsoft and Cray deliver "mainstream" CX1 supercomputer: starts at $25k
C'mon, who here doesn't want their very own supercomputer to do, um, whatever they want with? In an effort to make sure every man, woman and child has an absurdly powerful number cruncher in their home (let's go with OSPP, or One Supercomputer Per Person), Microsoft has tag-teamed with the fabled Cray in order to "drive high productivity computing into the mainstream." The Cray CX1 Supercomputer comes loaded with Windows HPC Server 2008 and incorporates up to 8 nodes and 16 Intel Xeon CPUs (dual- or quad-core); additionally, it boasts up to 4TB of internal storage, 64GB of memory per node and interoperates nicely with Linux. The CX1 is said to be the most affordable supercomputer offered by Cray (not to mention the "world's highest-performing computer that uses standard office power"), but it'll still run you anywhere between $25,000 to well over $60,000. Chump change, right?[Via NetworkWorld]
Fastest-ever Windows HPC cluster nets 68.5 teraflops
In the grand scheme of supercomputers, a homegrown cluster constructed by Microsoft is just a boy among men. However, said rig has set at least one record by becoming the "fastest-ever Windows HPC cluster." At the International Supercomputing conference in Dresden, Germany, it was announced that this beast ranked 23rd in the world "with a problem-solving performance of 68.5 teraflops." The National Center for Supercomputing Applications utilized a beta version of Windows HPC Server 2008 to hit the aforesaid mark, and if you're curious as to what it took to get there, try 9,472 cores of processing power. There's more where this came from for the hardcore nerds in attendance right in the read link.[Via Slashdot]
Linux cluster stuffed in an Ikea filing cabinet
People have been stuffing PCs in all kinds of things they don't belong in for a while now, but this Linux cluster creatively packed into an Ikea Helmer filing cabinet might be the first time we've seen furniture actually modded into a useful case. Sure, it looks like an ordinary filing cabinet, but it's packing six machines with Intel Core 2 Quad processors on Gigabyte S-series mobos with 8GB of RAM each, allowing it pump out 186 Gflops -- enough to complete a render job that takes a 2.66Ghz quad-core Mac Pro nine hours in just 64 minutes. Yeah, that's quite a filing cabinet. Hit the read links for tech specs, instructions, and updates on Helmer II.[Via Make]
SiCortex intros SC072 Catapult -- 72 processor cluster for $15000
Off hand, we can't think of how we'd truly utilize the horsepower generated by a 72-processor cluster shoved into a "whisper-quiet, low-power deskside cabinet," but we'd happily draw up a plan if forced. SiCortex -- the same folks who delivered the bicycle-powered supercomputer -- has introduced its new SO072 Catapult, which features a standard Linux environment, 48GB of RAM and a trio of (optional) PCIExpress slots. This aptly categorized high performance computer (HPC) sucks down less than 200-watts of power, sports a pair of gigabit Ethernet ports and has room for six internal hard drives. Reportedly, each of the 12 SC072 nodes is a multi-core chip with six CPU cores, and while $15,000 may seem steep for your average tower, we'd say this is a pretty good value considering the hardware.[Via Gadgetopia]
Sony erects massive PS3 server cluster for Warhawk mayhem
C'mon, we all knew PS3 clusters were good for more than academia, right? Thankfully for those who are itching to jump right into a worldwide dogfight when Warhawk lands, it looks like Sony has you covered. Granted, the game will allow for PS3 owners to host and play on their own matches, but the Ranked-Dedicated servers that you may also opt for shouldn't be lacking in terms of sheer power. Constructed by the SCEA IT team, this ginormous PS3 cluster will soon be used to connect Warhawk gamers everywhere, and while we're never told precisely how many PlayStation 3s were scrounged up in order to make this happen, feel free to click on through for another shot and start countin'.
NC State engineer crafts academic cluster with PlayStation 3s
While universities have been cranking out supercomputers and research clusters for some time, an associate professor at NC State is utilizing IBM's highly-touted Cell processor in a slightly different form to craft his own farm. Similar to the Xbox Linux cluster from years past, this concoction consists of eight PlayStation 3 consoles networked together and powered via Linux in order to handle ridiculous amounts of number crunching. Dubbed the "world's first" PS3-based academic cluster, the creation boasts the ability to utilize "64 logical processors," and is set to be used to handle various research tasks when sly CSC students aren't firing up a round of Ridge Racer 7 after hours. Nevertheless, Dr. Frank Mueller noted that the biggest limitation in its current state is the "512MB RAM constraint," but did insinuate that he might try retrofitting additional memory if future tasks deemed it necessary. Still, we can't help but wonder how many spots the Pack could jump in the RIAA's Most Wanted list if this thing became a dedicated torrent server.[Via TWW]
Rig of the Day: Me and the cluster
It's just another day at the office. That is, an office that houses two racks bulging with Xserves. Here you'll find Flickr user el frijole standing next to 42 Xserve G4s, 5 G5 Xserve cluster nodes and 1 Xserve G5. We'd call that a pretty decent rig, wouldn't you?"Me and The Cluster" posted by el frijoleIf you'd like to see your own rig featured here, simply upload photos into our group Flickr pool. We'll select an image every day to highlight.
Rig of the Day: iPod and Goliath
"This iPod is linked to another iTunes Music Library massive Xserve cluster. Do you want to change the link to this iTunes Music Library massive Xserve cluster?" We realize that this isn't your typical Mac rig, but we just had to point it out. Flickr user XaOS posts this shot of his iPod getting together with its (much) bigger brother. That's one heck of a music collection you've got there."cluster and ipod" posted by XaOS.If you'd like to see your own rig featured here, simply upload photos into our group Flickr pool. We'll select an image every day to highlight.
