Slashdot is powered by your submissions, so send in your scoop

 


Forgot your password?
Close
typodupeerror
×
BSD Operating Systems

NetBSD Runs a Marathon 37

Posted by michael from the boy-are-its-inodes-tired dept.
hubertf writes: "Last weekend, R-KOM and the University of Applied Science (Fachhochschule, FH) Regensburg, Germany, took their share of the Regensburg city marathon by putting a video and image of each runner reaching the goal on the Internet. A cluster of 45 machines from the Fachhochschule Regensburg, each running the NetBSD operating system computed over five thousand films. The cluster machines were operating on a common NFS storage, performing two steps. First, the video stream provided in MPEG format was split into single pictures, then a six seconds long movie was assembled for each runner reaching the goal, showing his personal run through the goal. Overall computing time was about 20h in which the five hours of video material was split into 670.000 images, that were then re-assembled into 5500 MPEG streams of about 1 megabyte each. A short english-language description of the Marathon Cluster is available, and there's also a german language version which has many details on the setup and operation of the cluster."
This discussion has been archived. No new comments can be posted.

NetBSD Runs a Marathon More

NetBSD Runs a Marathon

Comments Filter:

  • A study in contradiction (Score:1)

    by Anonymous Coward
    It's kind of funny to contrast the two disparate groups involved in this event.

    First we have the lean, healthy, finely tuned and wirey marathon athletes, competing with all their might to excel in the grueling 26 mile competition.

    Second, we have the morbidly obese athleticly challenged NetBSD nerds sitting around on their broad behinds, nursing their computers, eating donuts, and taking webcam pictures of the first group - the athletes.

    As they say, it takes all kinds to make a world.

  • Heh. (Score:3)

    by Wakko Warner ( 324 ) on Wednesday May 30, 2001 @12:46PM (#188433) Homepage Journal
    The teaser for this article fails to mention that the NFS server ran Solaris 2.6 [feyrer.de].

    - A.P.

    --
    Forget Napster. Why not really break the law?

  • If they had SDL they could run this Marathon [bungie.org]... :P

  • Whether you agree with the troll or not, it was interesting/funny to see the English translation say that NetBSD was a Unix/Linux-like operating system. Now if that doesn't get a *BSD person a little ruffled, I don't know what would!

  • If they had SDL they could run this Marathon... :P

    Uh, we do [netbsd.org] and, incidentally, I am. (Too lazy to bother getting a uname and an Aleph One shot on the screen at the same time, though.)

    Matter of fact, seeing as we have full binary Linux compatibility on the i386 architecture, I could have just installed all of this by RPM.

    Trust me, you ain't got anything we ain't also got.
  • Heh - when they did this for people entering and leaving the Superbowl [slashdot.org], it was pure evil in the eyes of many Slashdotters...

    Still, it's a neat trick.
    --
    the telephone rings / problem between screen and chair / thoughts of homocide
  • those finishline cams have been available for most (if not all) races of the Ironman series for quite some time now. the site that's covering them (live.ironmanlive.com [ironmanlive.com]) also provides updates during the race, including up-to-date swim/bike/run split times. (for my performance at Ironman California this year, look for athlete #213 :-)
    I tried to find more technical info on the sportvision [sportvision.com] homepage, but a quick glance didn't reveal any specifics on their approach.
  • 1: The more informative text was in german (which I cant read...) yet the illustrations were in english?

    2: I dont know about the rest of you - but any mpeg encoding I've done that has taken chopping, decreased size, added logos and taken any stills takes quite a while on one machine - this cluster did A LOT in a short amount of time! Nonetheless they added the runner name and time on the pictures (and in the movie?) using, i assume, the transponders they equip at races now.

    This is neat. Good usage of technologies to automatically do some stuff that would take humans forever to do by hand!
  • Amusement parks don't make mpeg movies for people, they make still shots. If they only wanted to make still shots available on the web, they probably could have gotten by with a couple of machines. But they were making individual movies for people. Picking out the correct frames and then mpeg encoding them is computationally expensive. Hence the cluster of 45 machines.

    --
    I hope we shall crush in its birth the aristocracy of our monied corporations ...
  • I ran my first marathon last year, and I thought chip timing was really cutting edge! =)

    Seriously, I don't think that I would have wanted my picture taken when I finished; I was crying with relief!
  • The hard part, identifing the runner's and their time was done with ankle radio trackers - I'm assuming. So really all this did was given a list of times and 5 hours of frames, mpeg encode 180 frames per runner. Doesn't seem like that difficult of a task that 45 machines would be needed. If they used re-used information then the encoding could probably be done in less then 12hrs on 1 server... but it sounds like it was just a brute force shell script that brute forced the videos as seperate operations. Whatever works!
  • Albeit a very nice demonstrtatioon of parrall distributed processing, th ssi is just a missuse of a rare comodity, parrelle computing environments. Most folks cannotg afford a cluster before any type of big iron, but still an expensive club to join.

    This type of system sounds more usefull as an example to demonstraigh a system like echelon, or a futur cousin, that can take random snap shots of folks walking down the street, and use optical recognition software to id your facial characteristics, aka biometric recognition.

    I belive a few conspiracy theorist have talked about this before, the ability to do somehting like say in England, where the goverment has abundant cameras in public places and, for example, use an AI to track you every move in public places where the ccd camera's connect to the central system.

    This computer is nothign more than a primitive example of the same type of myth, at least until now.
  • I would have expected a lot to happen in my lifetime, but my former school on slashdot, this is simply too much. I think I'll never say never again.
  • That LinuxNetBSD Cluster Rocks. I need one of those to help me fork linux 7.2 kernel for big iron.

  • But (Score:2)

    by zpengo ( 99887 )
    It'll run a Marathon, but will it run Tribes 2?

    (Oh, wait...it will...)

  • Maybe the lower number of BSD posts is because BSD users spend their time actually coding and stuff instead of posting flamebait trashing other people's operating systems?
  • The error encountered is:

    Not a valid referer.

    --
  • Good point - if we get to do this again, I'll ask Sun for a fully-stuffed E10k to do the job. :)
    (AFTER we got the software running on Solaris, which flaked out on us on Solarix 8/x86...)

    Oh, and try to get some decent video material. ;/

    - Hubert
  • Probably not. By what I understand, Beowulf provides some special ways for communication between nodes.

    What we used was just plain services available on every Unix system - rsh and NFS.

    - Hubert

  • Re:A Marathon? (Score:3)

    by hubertf ( 124995 ) on Wednesday May 30, 2001 @12:15PM (#188451) Homepage Journal
    Actually, we had SDL.

    The software used in the first step of the cluster - splitting up the MPEG stream into single pictures. We used a customized version of dumpmpeg, which uses smpeg and SDL.

    For the second part, we used mpeg_encode to create the 5500 MPEGs from the 660.000 JPEGs.

    - Hubert
  • Some of us care about ourselves and our friends acheiveing goals other than top fragger in Quake III.
  • Man, I hope my employer implements one of these soon... we'll have a lot less slackers shuffling in around 10:30 am and putting 8:00 down on their timecard! (Oops! I almost used an emoticon! I should kill myself now.)
  • The Ironman crew set up a camera at the finish line, and then just streamed it using RealPlayer [ironmanlive.com] ... seems like that would take much less computational resources than what these guys did, although it wouldn't be as personal.
  • Just to get this straight for the records: That was a school in Germany. Members of the species homo nerdus germanus are not known to feed excessively on donuts, not are they usually fat, hence they're only very loosely related to the home nerdus americanus.
  • *BSD on the frontpage twice in twenty-four hours? Where are the real /. editors, and who the hell are you?
  • I have to say that this is a very cool idea. I don't think I want to see my own crossing of the finish line of the last 1/2 marathon I ran a few weeks ago (my nipples were bleeding which made for a pretty bad day).

    Simpler solution(?): An MPEG encoder (and a pair of Betacam machines running time of day TC for backup), a RIAD array, and the ChampionChip system would work just fine given a 24 hour turnaround time.

    The time on the chip is translated to a frame number which is Mediacleaned out and mailed via a bot.

    I don't see why it should take 45 machines...

  • Last year I finished, probably about 3000th, but heard there were photos of each runner available at the finish. Reality was a bunch of photographers taking wide angle shots and offering to crop out a picture of each runner. With my luck it would be one of those "horror" shots like they tabloids like to use along side bad-news-for-this-person headlines. Run 10K, feel fine, cross the line in a sprint and all I'd have to show for it would be some expression like I'm trying to inhale a booger.

    I hope this stuff gets around, it would be nice in a couple years to be able to put on a website and email URL to friends and family.

    It may look like I'm trying to hoork a loogie, but this was my best finishing time, yet!

    --
    All your .sig are belong to us!

  • or you could start coming in at 10:30 like the rest of us

    (and you call yourself a programmer *sniff*)
  • If the cluster had used optical recognition to grab the runners number or face and made the film from that, then I'm impressed!

    If they just used the cluster for some automated editing based upon the runners finishing time, then so what?

    Ask yourself would this have made it to /. if it had been a cluster of Win2K machines?
  • Mandatory karma whoring using Bablefish [altavista.com]

    Translated German Version [altavista.com]
  • Well, at least I'm bad at karma whoring. :)

    The direct link to the translation is here [altavista.com]

    For the attentive, the first link required you to click on the link to the german version to get the german to english translation. Whoops, my bad.
  • Comment removed based on user account deletion
  • Comment removed based on user account deletion
  • but,

    Imagine a Beowulf Cluster those!
    __________________

  • So it took 45 computers to do what amusement parks routinely do with $15,000 worth of custom video equipment? Efficent!
  • I work for Sportvision [sportvision.com] and I do the live internet production of our Ironman coverage. We encode the entire finishline stream in Real Video - about 8 hours @ 350 megs. We use a script to generate files based on the timing data that reference the correct minute or so of video around the finisher. Go to live.ironmanlive.com [ironmanlive.com] to check it out. Go to any athlete that finished any race and you can watch them finish!
  • The Solaris NFS Server needed 2 Reboots. -- Juergen

Slashdot Top Deals

Trap full -- please empty.

Close