Project:Zeus

Zeus

Among the server stuff we've got in the space we have several IBM x3590 M2's.

4 of them can be linked together to form a single computer, with the bits we've got this will have 64 cpu's and 512Gb ram.

This page is for listing the uses to which we would put such a machine.

Short term usage

play around with it for a week or so

• Test the BSD's on it and make it available to *BSD developers for kernel testing - NetBSD at least has only recently had support for > 32 cpu's and may not scale well at 64 CPU's giving developers access to a machine of this size will enable them to test scalability.

• Try to get on the steam linux beta and run steam/source on it.

• muck about with it cos we can.

• We'd then sell it...

Long term usage

• Zeus would *not* be left on 24/7. We would use a remotely controllable PDU to power it off when not in use, and could charge for the power used.
• We could set up password protection for the PDU.
• We cannot use large amounts of internet bandwidth with it - the spaces internet connection wouldn't tolerate it.

Render farm

• There is some interest from the graphics hackers (?) for a blender/yaffray render farm
• Could also be used for video rendering.
• Data would have to be transfered from/to Zeus with external HD's or by copying over the hackspace lan.

Very large dataset processing

• There are a few hackspace members interested in space and photogrammetery. We could get Lunar & Martian photographic datasets from NASA/JAXA/ESA and process them into virtual planets and then go on virtual moonwalks and mars walks with Project:OculusRIFT Rifts

Internet Simulator

• Using lightweight process isolation (e.g. FreeBSD Jails + dummynet, usermode Linux, etc..) set up a system that approximates the internet with e.g.:
  • many virtual isp's
  • with end-users with realistic bandwidth
  • an 'Atlantic' and a 'pacific' with appropriate latency

We can then use this to:

• test out the various network topologically aware bittorrent schemes and see how much of a difference (both in terms of bandwidth usage over backbones and in terms of download speed) the different schemes make.
• setup our own isolated versions of some of the proposed distributed p2p social network and darknet systems (e.g. freenet, bitcoin, tahrir), and
  • try to break them:
    • can they be tampered with in transit?
    • can they be persuaded to ddos people?
    • can they be poisoned with false info?
    • and so on.