Forgot your password?
typodupeerror
Hardware Software Linux Technology

Flash Mob Supercomputer? 259

Posted by timothy
from the all-the-cool-kids-are-doing-it dept.
dan of the north writes "The NY Times (free reg yyy bbb) is running an article on flash mob computing. More info on the first event in SF on April 3, 2004. The goal is to run Linpack and "build a home-brew computer powerful enough to be added to a list of the world's 500 fastest computers." Minimum requirements are 1.3 GHZ Pentium III/AMD equivalent or better with 256MB of RAM, a 100 Base-T network connection and a CD-ROM - laptops preferred. "After taking a shot at a speed record, the computer will be reorganized to serve as the host of a giant multiplayer video game tournament." Cool... a 2fer!"
This discussion has been archived. No new comments can be posted.

Flash Mob Supercomputer?

Comments Filter:
  • Wicked. (Score:3, Interesting)

    by torpor (458) <ibisum@gmail. c o m> on Monday February 23, 2004 @04:20AM (#8360736) Homepage Journal
    Won't be long before sporting events and rock concerts will be able to host such supercomputers, too ...

    Imagine, iPod2 has WLAN ... good enough node spec for me! ;)
  • by baneblackblade (682424) on Monday February 23, 2004 @04:30AM (#8360763) Homepage
    This doesn't sound like such a bad idea. seems like a lot of things could be done this way much faster and more efficiently. if this works we should look into other applications for the Flash Mob, like a rocket-building day for the X-Prize, or a random code-swap where a bunch of us get together and hand eachother a blank disc with the source code to something nifty on it to play with.
  • by HermesHuang (606596) on Monday February 23, 2004 @04:45AM (#8360801)
    Next thing you know they'll have a room at science conferences where people leave their laptops when they're not presenting so that protein folding or whatnot gets worked out on-the-spot.
  • OK, here is the deal (Score:5, Interesting)

    by heironymouscoward (683461) <heironymouscoward&yahoo,com> on Monday February 23, 2004 @04:48AM (#8360814) Journal
    I have an idea as how to make Wifi hotspots economical. Imagine the Matrix meets Slashdot - in our wifi hotspot, imagine a network that sucks the living cycles from a beowulf cluster of you!

    Here is the deal... to use a hotspot you have to download a package that connects your computer to the local "grid". In exchange for network access the grid gets your spare CPU cycles. The best hotspots could leverage the power of hundreds of notebooks, and then resell this on the market as a computing resource commodity, for multiplayer games, data crunching, whatever.

    Though... I'm running a high fever and this is perhaps the fruit of a deranged mind.

  • Dorm Clusters (Score:5, Interesting)

    by HFShadow (530449) on Monday February 23, 2004 @04:55AM (#8360832)
    The next logical step I think would be for a university to provide free internet in dorm rooms, as long as you leave your system on and run a distributed computing client for them. The student saves $$ without any noticable problems on their side, the university gets free computing time, seems like a win win situation.
  • Re:I'm there! (Score:1, Interesting)

    by Anonymous Coward on Monday February 23, 2004 @05:14AM (#8360881)
    This event is more about community-building and fun than about science. Most of the people attending probably aren't billing by the hour, but still: Who's going to leave their computer unattended? Is the computational power of a single CPU worth binding its owner's time for the duration of the event? A bare-bones system which meets the requirements can be built for less than $200 (ATX-case, board, AthlonXP, 512MB, no CD, no HD, PXE boot). How many flashmob supercomputers are you going to join before it occurs to you that you could have contributed that kind of node to a community-controlled permanent supercomputer instead? More cycles, less investment.
  • Tough one to call... (Score:5, Interesting)

    by Anonymous Coward on Monday February 23, 2004 @05:15AM (#8360885)
    I really think 1200 is overkill for this. Take a look at the Top500 list and see how badly all the Gigabit Ethernet systems scale: most of them have worse than 50% efficiency and that's with only 1/4 the number of nodes. Now cut the interconnect bandwidth by a factor of TEN, cause Apple is pretty much the only company putting Gig-E standard in their (pro) computers, and it seems to me that hundreds of people are going to be sent packing because adding them to the cluster would actually make it SLOWER.
  • by kb (43460) on Monday February 23, 2004 @05:18AM (#8360895) Homepage Journal
    If "non-destructive" excludes the buggy 2.0.36 Adaptec SCSI drivers which fried a handful of AHA2940 controllers... yep :)
  • by billstewart (78916) on Monday February 23, 2004 @05:45AM (#8360962) Journal
    SETI@HOME isn't on the Top500 list, because it's not running Linpack, but according to its stats page, it's been running at about 63 TeraFLOPS today, which is comfortably #1 on the list. So this should be fun...
  • Re:Dorm Clusters (Score:5, Interesting)

    by joib (70841) on Monday February 23, 2004 @05:59AM (#8360988)
    I don't think it will work. Reliability, bandwidth and latency tend to be the biggest problems with purpose built clusters. I imagine a setup like the one which you describe is going to be much worse in these aspects. Not to mention that most simulation software is written for homogeneous clusters, i.e. all the computers in the clusters should have an identical software environment, and the slowest node limits the speed of the entire calculation.

    A similar but slightly less pie-in-the-sky thing would be to use the lab computers for batch processing during the night. Lab computers tend to be centrally administered and perhaps identical computers could be grouped together so that CPU cycles aren't wasted to the extent that they would be in a homogeneous environment.
  • by Anonymous Coward on Monday February 23, 2004 @05:59AM (#8360990)
    All Slashdot postings linking to NYTimes articles should always also link to this url:

    http://www.majcher.com/nytview.html [majcher.com]

    A javascriptlet there will allow you to generate a totally random login for viewing the article. Every Slashdotter which accesses the article should create a new random login in turn, filling their database with useless random login id's that are only used once and then forgotten about.

  • What, no macs? (Score:4, Interesting)

    by morganjharvey (638479) on Monday February 23, 2004 @06:16AM (#8361016)
    Minimum requirements are 1.3 GHZ Pentium III/AMD equivalent or better with 256MB of RAM
    Perhaps I'm missing some fundamental requirement of cluster comptuing, but why wouldn't macs work? I'm sure a 1.25 ghz G4 could hold its own with the above mentioned. It can also run linux. College campuses seem to be a hotbed of mac users, so it seems that they would want to tap this. Does clustering require that all nodes be of the same architecture?
  • Re:Wicked. (Score:5, Interesting)

    by torpor (458) <ibisum@gmail. c o m> on Monday February 23, 2004 @06:21AM (#8361022) Homepage Journal
    You walk into the concert arena, your iPod2 uses zeroconf (rendezvous) to identify itself to the concert host system.

    anyone whos' got the 'share my compute cycles' checkbox set on their ipod2 automatically shares their ipod2's cpu cycles, again announced by rendezvous, with the rest of the system.

    for the duration of that gig, the machines are all bound together, beowulf-style, to distribute the live recording of the event that is being produced for prosperity during the concert. some 'secrets and surprises' are thrown into the tracks too, just for grins.

    at the end of the gig, everyone walks away with their own digital recording of the event, custom, unique, 'branded to the event'.

    the whole thing was included in the price of admission, and open to anyone who walks into the concert arena with their boxes turned on ...

    I can totally see this happening. In fact, if I had the resources, I'd start a company that does just this service for concerts and gigs and such ... now would be round about the right time to get into this market, since its infancy-stages ...
  • by archilocus (715776) on Monday February 23, 2004 @06:26AM (#8361035) Homepage

    Interesting term "flash mob". The first time I encountered it was in the writing of sci-fi author, Larry Niven. In some of his series personal teleportation becomes ubiquitous giving rise to the 'flash mob'.

    When a news broadcast reports a certain kind of story (riot, fire, etc) people start to teleport into watch the fun. The news reports the growing mob and before long it reaches critical mass and turns into a real riot as people take advantage by teleporting in and doing a quick bit of looting.

    I'm not sure if Larry originated the term though ? Anyone know an ealier source ? Is it a 'real' phenomenon ?

  • by the3mcsand1dj (755599) on Monday February 23, 2004 @06:53AM (#8361089) Homepage
    Well, not exactly... but kinda.

    Since I was a small boy, I have been facinated with computers... I researched paralell processor computers (by this I mean that I asked my Computer Engineer uncle about them) and quickly thought about alternate interfaces for "linking" processors, other than building custom mother boards capable of distributing computing power amongst multiple processors... and was quickly discouraged by the lack of comunication speed inherent in the options available at the time (namely: Serial, Paralell, ISA ... etc.).

    And then, what's interesting... is that I was JUST thinking about this the other day .. and told my friend Kevin about my idea that I had when I was a kid... and said "I want to design a software or maybe a type of "add-on" card that would allow you to connect multiple computers together to form a type of super-computer".

    Then I log in today, and see this headline! Blink a few times and read it again, then read the article and don't quite understand it and then realize that, although neat, it's not that cool the way they are doing it...

    First, even 10/100 LAN is a bit slow to properly transfer data at rates that would be condusive to properly handling large amounts of data. I would liken this method to a Muliti-Processor Xeon motherboad with a 66mhz front side bus. Yes, the processors are super-fast, but what's the point if you can't send them the data or receive it fast enough to use them?

    Second, if all they are doing is connecting computers, what do they plan to acheive? From the article I see no plans for organizing the computers into "Pre-Processing", "Processing" and "Clean-up" groups in the true form of a Parallel Processor computer...

    And Third, I thought of it first... haha

    Anyway, I still wish I could be there to see what happens and how it goes... I hope the benchmark tests will be properly designed to include both small and large datasets and sufficiently complex procedures...
  • by Lost Race (681080) on Monday February 23, 2004 @07:05AM (#8361122)
    I thought of a couple more problems...

    1. Heat: If they expect to have 1200 people sitting in the bleachers with 1300+ MHz laptops running at 100% for the duration of a real benchmark, they'll have to expect a lot of burned laps.

    2. Theft: If they have a designated hookup area for the computers (much more logistically feasible, ethernet-wise at least) they'll have a hard time getting all 1200 computers successfully reunited with 1200 owners. It'd be very easy to grab the wrong one "accidentally". Given that this is San Francisco I'd expect vultures looking for any opportunity in such a huge computer thief buffet.

  • Re:Article Text (Score:1, Interesting)

    by Anonymous Coward on Monday February 23, 2004 @07:12AM (#8361140)
    "It coluld be that electrical power will be an issue"

    Perhaps people should bring a good UPS along. I don't know how long the machines have to be up, but it may be possible to run enough machines on UPS long enough to achieve their goals.
  • Re:licenses (Score:2, Interesting)

    by marksie531 (152393) on Monday February 23, 2004 @07:44AM (#8361246)
    You have given me an idea... How cool would it be to have a massive war (e.g. 500+ people). There hardware is there, all we need is a game which could handle this sort of scale.
  • by carlmenezes (204187) on Monday February 23, 2004 @12:56PM (#8363751) Homepage
    ...like a KNOPPIX ISO image that is specifically meant for setting up a distributed computer on a local LAN. How about it? I can think of a lot of places that could use the power of distributed computing that don't necessarily have all the knowhow of how to set it up - let's take the most obvious : public schools. Something like a Flash Mob Computing KNOPPIX ISO downloadable would be like a gift from God.

If the facts don't fit the theory, change the facts. -- Albert Einstein

Working...