Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 



Forgot your password?
typodupeerror
×
Network Games Linux Technology

Technical Details Behind the LAN-Party Optimized House 123

New submitter Temporal writes "Yesterday, Slashdot reported on my LAN-party optimized house. But, lacking from the internet at that time were key technical details: How do I boot 12 machines off a single shared disk? What software do I use? What does my network infrastructure look like? Why do I have such terrible furniture? Is that Gabe Newell on the couch? The answer is a combination of Linux, PXE boot, gPXE, NBD/iSCSI, and LVM snapshots running on generic hardware over generic gigabit ethernet. I have even had several successful LAN parties with a pure-Linux setup, using WINE."
This discussion has been archived. No new comments can be posted.

Technical Details Behind the LAN-Party Optimized House

Comments Filter:
  • by Fallingcow ( 213461 ) on Friday December 16, 2011 @05:51PM (#38404540) Homepage

    ...plenty of wine would make a Linux-centric LAN party tolerable, and perhaps even enjoyable.

  • No, the burning question was what kind of chuckle-head would spend all that money and stick his friends on fixed height desks all (it looks from the pictures) at writing height, not typing height.

    He's going to give all his friends who are not very tall shoulder and wrist problems.

    Look at the pictures. They're all (except for one very tall guy) very badly positioned for work.

    He's going to need a to dole out ibuprofen by the pound to his guests.

    • by Hatta ( 162192 )

      Adjustable chairs are pretty easy to come by.

    • Re:Desk height (Score:5, Interesting)

      by Temporal ( 96070 ) on Friday December 16, 2011 @06:00PM (#38404594) Journal

      It's a fair criticism. I wanted to make the desks lower but I also wanted the monitors to sit higher when folded up, and the desks were getting stupidly deep, so I had to compromise. In practice, though, people aren't typing at these desks, they're gaming, which in my experience (as someone with some RSI problems) is not as sensitive to desk height.

      • I can think of a couple of ways around it, but each with their own disadvantages, and what you did seems simplest and clean. That said, I'd definitely prefer height adjustable rolling chairs (floor scratches and storage space be damned).

        Incidentally, where does the sound come from on the stations? And how loud is it in that room, as I don't see many sound absorbing surfaces.

        • You could use USB headsets, but that is a good question as it does not appear to be answered anywhere in the article.

  • Great follow up to the previous article.

    Kudos to you.
  • Crashman?! (Score:2, Funny)

    by Anonymous Coward

    Why would you ever name a device Crashman? This just seems ominous!

    "Which machine do you want to use for the LAN party?"
    "Oh, Crashman looks really stable."

  • Does anyone know if this setup would work with Windows 7 images? Not the installer, PXE boot the whole OS. I've often wondered.
    • by Temporal ( 96070 ) on Friday December 16, 2011 @06:11PM (#38404684) Journal

      Yes, that's exactly what I'm doing -- PXE-booting the whole OS over iSCSI. I'll edit to make that clearer.

      • From the article: Once Windows was installed to the iSCSI target, gPXE could then boot directly into it, without any need for a local disk at all. Yes, this means you can PXE-boot Windows 7 itself, not just the installer. Now I fully intend on doing this, just not to your extent. I read up on gPXE and where it can be installed to, the web site says: "We, the Etherboot Project, create network booting code that allows computers to load their operating system from a network. Our code can be stored in a numb
      • Yes, that's exactly what I'm doing -- PXE-booting the whole OS over iSCSI. I'll edit to make that clearer.

        I've looked at some of the pricey solutions that do various versions of this, and what you've done for free is very impressive. I do want to confirm that I'm reading it right though, and that you're basically booting the same image, so the machine names and such will all be identical within Windows, is that correct? Have you seen that cause any issues for any games? Have you tried any LAN games like this (preferably more recent than Quake)?

        Clearly having the same machine name and SSID makes it unusable f

        • by Temporal ( 96070 )

          Yeah they all end up with the same machine name. I disabled automatic network discovery so they won't broadcast themselves, since it would just lead to confusion. File sharing can be done by copying files to and from a share on the server machine. I don't see why games would care since they operate at the TCP/IP level, not using computer names. I have only tested a couple things so far, though, like UT2k4... will be trying more today (LAN party!).

    • by Junta ( 36770 )

      All you can 'PXE' boot is WinPE, but most anyone wouldn't care about the distinction between that and PXE-bootstrapped iSCSI. PXE boot iPXE, then iPXE interprets root-path to start Windows with their native software iSCSI support. ipxe.org is a place you can do it yourself. xCAT automates everything from stgt iSCSI creation, to unattended windows install from DVD into iSCSI root, but may be a bit much for a home setup when you can just learn how to do it yourself manually at not much more effort.

  • I can't imagine a single machine serving out over iSCSI to have performance acceptable to play any modern, intensive game. How's it all work?

    • by Temporal ( 96070 ) on Friday December 16, 2011 @06:15PM (#38404708) Journal

      I can't imagine a single machine serving out over iSCSI to have performance acceptable to play any modern, intensive game. How's it all work?

      I couldn't imagine it either, but it turns out it works fine. Obviously the load times aren't blazingly fast but no one has ever complained about them being slow either.

      Note that most games load all data upfront. Once they've done that, the game runs without doing much I/O.

      Also note that an iSCSI image can be fully cached client-side, so if you load the same game twice, it's probably going to load directly from RAM the second time. (Most games are 32-bit so there's a good 4GB of RAM in the machines doing not much other than disk cache.)

      • by man_ls ( 248470 )

        Very cool. Thanks for posting that write-up.

      • by jd ( 1658 )

        PXE, et al, use TFTP, if I remember rightly. In principle, there's nothing to stop the files being delivered by multicast FTP (yes there are at least three, they use Scalable Reliable Multicast, FLUTE or NACK-Oriented Reliable Multicast respectively). Since OS images and the games themselves don't differ between machines, if you have N machines you get file delivery about N times as fast. (About because lost packets are resent, so it's not truly linear improvement.)

        • by Temporal ( 96070 )

          PXE, et al, use TFTP, if I remember rightly. In principle, there's nothing to stop the files being delivered by multicast FTP (yes there are at least three, they use Scalable Reliable Multicast, FLUTE or NACK-Oriented Reliable Multicast respectively). Since OS images and the games themselves don't differ between machines, if you have N machines you get file delivery about N times as fast. (About because lost packets are resent, so it's not truly linear improvement.)

          Indeed, as mentioned in the blog, I was at one point trying to develop a UDP-based blog device protocol that would broadcast blocks back, on the assumption that all the machines would be loading the same data at the same time. However, when I got the system up and running without that, and didn't see any performance problems, I decided to abandon that idea and focus efforts elsewhere.

          • by jd ( 1658 )

            That makes sense. The bottleneck defines the upper limit of performance. If you do go back to working with UDP, I do suggest grabbing either the NRL's NORM (snapshot version since they don't produce official releases any more for some reason) or FLUTE, since those give you the libraries you want for no effort. Both are listed on Freshme....sorry, Freecode. (That new site name STILL sounds too much like Freescale, the Motorola chip spinoff.)

        • by Junta ( 36770 )

          Well with gPXE (pretty much dead now as the meat of the project forked to iPXE), most setups use tftp to transfer ~64 kilobytes and then other protocols take over (like http for a tftp workalike loading into ram or iSCSI for block storage loading content on-demand). Now with games, level load times may vary, but generally everything is in memory (or at *least* cache) when it comes down to actual gameplay.

          Stateless ram-based OSes delivered can work too, but the multicast generally isn't worth it. A system

      • by wierd_w ( 1375923 ) on Friday December 16, 2011 @06:55PM (#38405022)

        If all the image needs to do is hold the basic OS and a single game deployment, why not pxe the whole image to a dedicated hardware ram drive, that can make full use of the sata controller?

        Acard has a number of such devices, and while pricey, would absolutely floor disk io performance in a game rig. [acard.com.tw]

        There are quite a few other devices of this type on the market as well.

        Using these in the systems, you could still netload the system images to the game rigs with pxe, but when the image has finished being pushed, just reboot them and you have a bitchin fast row of locally booting systems. Power them off when they need a new configuration pushed.

        • by adolf ( 21054 )

          If the machine serving as an iSCSI host has sufficient RAM, it already almost works that way:

          1. Client A boots up. Loads OS from solid-state disk. Subsequent clients are likely to load directly from the server's RAM cache, which just got populated automatically.
          2. Client A runs a game. Loads game from disk. Subsequent clients are likely to load directly from the server's RAM cache...

          So the first load is at disk (SSD) speed, and the subsequent loads are at RAM speed (although both instances are limited

        • Or just stick 16gb+ ram in your machine, setup a soft-ramdisk...
          Best would be if things were ECC compatible, but well-tested regular ram would probably do fine.

          • Agreed, but most "massive amounts of ram" capable desktop boards are multiprocessor cuda designs.

            This means a cpu hit and cpu bus io bottleneck for disk io.
            This bottleneck will also be having to service exchanges between the videocards and system ram, and program execution requirements. (Most games are uniprocessor, even today.) The ram is present, but to get to it you have to talk to another processor/core.

            The issue with disk io vs cpu involvement is important. It is the single most important feature of a

            • I'm still thinking that with memory bandwidth in the many-GB/sec, the bottlenecks won't be apparent - if your entire game image is a GB or two, it'd still be loaded in a fraction of a second.

              And I know for a fact we can have single-processor(multicore, though), four-slot(or 6-slot Intel) memory motherboards, and with 4GB DDR3 cards...

      • by mcrbids ( 148650 )

        We found that SATA SSDs weren't just faster than 15k SCSI drives, but dramatically so in database servers. We saw at least 10x performance boost despite being "only" SATA drives.

        I wonder if you've considered this route?

        • by Temporal ( 96070 )

          I think you might be confusing iSCSI (a network protocol) with SCSI (a disk hardware interface). My physical disks are all SATA, and are a mix of SSD and HDD. The bottleneck is network bandwidth, so using SSD everywhere isn't going to improve performance much.

    • Games do not spend that much time doing disk IO after they've started. And most modern drives deliver nowhere near a gigabit of streaming performance anyway.

      • And most modern drives deliver nowhere near a gigabit of streaming performance anyway.

        Well, almost all SSDs do [anandtech.com]. (1Gb/s is about 125 MB/s).

        • An SSD would not be my first choice when trying to improve the performance of a gaming machine. But yes, you're right, almost all SSDs do.

    • Why not?

      Modern games aren't bottlenecked at the disk. Even if they were, the bottleneck then becomes the network, since it's been read from disk into RAM at this point. But try it -- next time you fire up a modern game, when it's attempting to load a level, watch the hard drive light. Usually it flickers from time to time. Then fire up Defrag or something, which you *know* will generate a lot of disk activity, and compare.

      And that's just loading. Once it's loaded, it's in RAM. The disk is just a nice big pl

  • Haha ... (Score:5, Interesting)

    by vivek7006 ( 585218 ) on Friday December 16, 2011 @06:05PM (#38404640) Homepage

    "I purchased 12 copies of Windows 7 Ultimate OEM System Builder edition, in 3-packs. However, it turns out that because the hardware is identical, Windows does not even realize that it is moving between machines."

  • God dammit (Score:5, Funny)

    by Anonymous Coward on Friday December 16, 2011 @06:13PM (#38404694)

    Every computer using geek has an Asian girlfriend except me.

    How they fuck do they do it?

    You have to be white, nerdy, and socially awkward. I've got all three! Come hither, ladies!

    • by dbIII ( 701233 )

      How they fuck do they do it?

      Meet them, talk to them and convince them that you are a good person. Fixing their computer and paying for their internet access helps too.

    • might be a trend there, three out of the other six IT techs in my employer's company have asian wives as I do. there is a traditional cultural thing where asian women are not treated well by the men. in the USA they see that things can be different, other men have proper respect for women. if you are kind, dependable, genuinely interested in them, have a job you can be their Mr. Right.
      • might be a trend there, three out of the other six IT techs in my employer's company have asian wives as I do. there is a traditional cultural thing where asian women are not treated well by the men. in the USA they see that things can be different, other men have proper respect for women. if you are kind, dependable, genuinely interested in them, have a job you can be their Mr. Right.

        I have also seen that there are many Asian women dating or married to software engineers, but I think it would be a mistake to believe that they all a) were born and raised in Asia, b) have a past history of experiencing patriarchal abuse or discrimination, or c) are involved with their partners only or primarily because of the kindness, dependability, and genuine interest that their partners show in them as contrasted with their previous partners (or other role models they've seen via the media, relatives

  • It would be neat to see how a setup like this would hold out for gaming with Windows 8, Hyper-V and remotefx (assuming gaming cards play nice with it), then you could consolidate expenses to a machine or two, and keep your existing "client" hardware for years, if I understand how it will work...
  • So he does not seem to get the full 6 mbps that http://www.sonic.net/ [sonic.net] offers on DSL. Must be to far from the CO or RT to get the full speed.

    he bases Comcast but they have faster speeds and better upload. But the Sonic.net directv bundle is better for TV.

  • Fake!!!! (Score:4, Funny)

    by Anonymous Coward on Friday December 16, 2011 @06:49PM (#38404962)

    This is a made up story by the liberal media.

    Blizzard clearly stated that no one is interested in LAN parties and the whole concept is dead. Please move on. There is nothing here to see.

  • The question isn't "How", its "Why".. money doesn't seem to be the big issue here, so why not spring for Server 2008R2 and manage all the boxes from there? it does all this updating/registering/etc your hacking together, and for around $800, versus your hourly rate x hours hacking, seems less expensive and the result is a heck of a lot more manageable. I'm all for the do-it-yourself type, but managing disk images? Yeah I can spend my time better elsewhere

    • by Temporal ( 96070 )

      The question isn't "How", its "Why".. money doesn't seem to be the big issue here, so why not spring for Server 2008R2 and manage all the boxes from there? it does all this updating/registering/etc your hacking together, and for around $800, versus your hourly rate x hours hacking, seems less expensive and the result is a heck of a lot more manageable. I'm all for the do-it-yourself type, but managing disk images? Yeah I can spend my time better elsewhere

      No, it really does not do anything like what I'm doing... and anyway, setting this all up was lots of fun.

    • Why bother spending $800 bucks for an OS you may not particularly like the style of anyway? For a task like this, I really don't see a particular advantage that Windows would hold unless all you know is Windows. The tools to do this sort of stuff are trivial to work in Linux (I personally think easier than Microsoft tools, but that may be a preference).

      I would give MS the benefit of the doubt on a setup suggesting AD account management, though I haven't tried 389 which may have a nice integrated feel. St

  • I'm curious as to what sort of latency he gets by extending USB through repeaters and if he uses hubs on his keyboard/mice at the other end? I know HDMI can be run quite a long distance (relatively speaking), but USB isn't made to be run long distances at all. If I remember correctly, there is a maximum distance USB can run before you need to add a hub/repeater. Adding to that, hubs in my experience interfere with peripherals, especially gaming peripherals (fast and high dpi mice). He doesn't really go into
    • by Temporal ( 96070 )

      Monoprice sells 32' USB extensions that have a repeater on the end. I'm not sure how they get away with just one repeater instead of two as seems mandated by the USB spec, but they do and it works.

      • Does this result in noticeable input latency that you can feel through gaming, similar to the sort of input drifting feeling you get with wireless peripherals?

        Also, do you use hubs on the end for the keyboard/mouse or do you run a separate cable for each one? If you use hubs, have you run into issues with using them? Do you have gaming quality peripherals on the end as well (high dpi and polling mice and high poll rate keyboards)?

        I ask as gaming peripherals tend to put more strain on USB connections then no
        • by Temporal ( 96070 )

          Does this result in noticeable input latency that you can feel through gaming, similar to the sort of input drifting feeling you get with wireless peripherals?

          No. According to the USB spec, maximum allowed round-trip latency to any peripheral (regardless of the number of hubs) is 1.5 microseconds, many orders of magnitude below what a human being could sense. Electricity will propagate over a 32' cable in around 50ns, another couple orders of magnitude smaller still.

          Also, do you use hubs on the end for the keyboard/mouse or do you run a separate cable for each one? If you use hubs, have you run into issues with using them? Do you have gaming quality peripherals on the end as well (high dpi and polling mice and high poll rate keyboards)?

          I have a hub attached to each one, with a cheap keyboard and a Logitech MX518 mouse. Guests are welcome to bring their own peripherals if they so choose.

          I ask as gaming peripherals tend to put more strain on USB connections then normal peripherals and can crap them out when they're operating on the edge (like the 32' repeater you mentioned). They draw more power and operate close to limitations of the spec in terms of latency.

          The hubs are powered. They are USB 2.0, so

          • I wouldn't ask unless it is something I'm actually sensitive to. I've used wireless peripherals and they do have a drifting feeling to them. While increasing the length of a wire itself may not add latency, a hub, which is essentially a repeater adds it's own latency. Just the same way bad switches can mess up a network or make lag. Not everything is equal even if there are standards they adhere to.

            Buy a G500 and do a side by side comparison with your MX518. It's not snake oil.

            I was overall just curious if
  • If you need more than a damned good switch, internet connection, and wireless AP to optimize your LAN parties (real LANners have their own boxen, TYVM) you're doing it wrong.

Software production is assumed to be a line function, but it is run like a staff function. -- Paul Licker

Working...