What to do with a cluster of old computers running Ubuntu?

Question

At our university, our section had a grant of several old P4 computers. We decided on using them to create a 3x3 grid, which would function as a medium for our off-hand processing work. However, as of now, this is all we have done :

This is a video streaming running in 9 parts using vlc/ssh scripts to manage it.

As of now all the computers are connected to a single network, and managed using ssh key based authentication. I would love to hear some innovative ideas. Some hardware details are provided :

1gb RAM, Intel Pentium 4 2.4 GHz, 40GB HDD running Lucid

Particularly, I was thinking about the 40gb Portion being used as a cloud-based storage space for people. Something along the lines of sparkleshare or Owncloud but they don't seem to offer multiple computer based network. If there is someway to harness the computing ability of this grid, I would also like to hear about it.

Answer 1

If the computers have a reasonable network connection, there are a number of services you can run to make the internet a better place. Few of the things I'm about to list are directly parallelizable across a cluster, but many of them work very well when used on a pool of machines with a DNS round-robin to load-balance and reduce the impact if any one of the machines goes down. In addition, you can perform maintenance on one machine at a time, while leaving the cluster and the services unaffected.

• run a network time daemon and add it to the time server pool
• run a GPG keyserver and add it to the keyservers pools
• run a Tor bridge or relay, and mirror the Tor Project website
• run a Mixmaster email mix server
• run a Convergence or Perspectives notary to keep track of SSL irregularities
• run a full instance of the SSL Observatory
• run a Tahoe-LAFS node to participate in a secure, distributed filesystem
• run an I2P or GNUnet node

The great thing is that most of these services aren't very processor-intensive, so they run really well even on older hardware. For these sorts of services, it's less important how much total processing-power you can contribute, and more important that there's a diverse set of machines involved. That's why I'm recommending them rather than contributing to one of the many @home projects:

for an older machine, you get a lot more make the world a better place with these sorts of services, than something which just wants all the CPU it can get.

Of course, with your leftover CPU time, you can still try to cure cancer or search the skies.

Answer 2

Put some sort of distributed computing client on there and donate your unused cycles to science.

Maybe something like http://folding.stanford.edu/

Answer 3

You may try mining some bitcoins. This virtual currency based on peer-to-peer protocol (and which can be easily exchanged for real money, at the moment of writing a bitcoin is worth about $11) has coins, which are generated not by a single institution, but by the network of all its users. The chances of having generated some amount of such money are the greater the more computing power you have. When you will have started, you can calculate if it will pay off.

Answer 4

Yes, also you can check out BOINC as well: https://boinc.berkeley.edu/.

You can add any BOINC supported project as per your wish. The most famous BOINC supported project is SETI@HOME, I think, but there are many other projects in fields like medicine as well. Here is a small list of projects: https://boinc.berkeley.edu/projects.php

Answer 5

A late answer, but why not a beowulf cluster ?

Answer 6

You can partake in the Great Internet Mersenne Prime Search! Install the client on several or all of the computers and let it run! If one of your computers discovers a new type of Mersenne Prime, you could win USD 3000 or even USD 50.000. It's like a fun lottery where the only entry fee is your electricity and computing power! And even if you don't find any prime number, you're contributing to science by helping confirm/reject Mersenne primes found by other people's computers.