Usage ideas for a cluster at a university

Question

I currently am a student worker at a medium sized university. i work for the college of science (includes math, stat, comp sci, physics, chem, and biology) and as a "technician"; which basically is a catch all for basic trouble shooting and fixing faculty computer, setting up computer labs, writing custom programs, and administrating web and database servers.

We currently have 3 groups of fairly new computers set up for clusters, however they are not in use and abandoned. I would like to do some kind of cluster project, but my biggest hurdle is what to do on/with the clusters. If i knew what applications or projects could utilize a cluster, I could go from there. I have read about clusters being used for video rendering, generating rainbow tables/password cracking, and high availability web servers. But I was looking for something more academic; something that could be presented to the professors and deans as something of actual academic usability.Not something that will never be used.

I don't really want to do video rendering because there is non need to render video. Password cracking would not have much use to the college. High availability web cluster has no use because we don't host but a few static faculty pages and our site (network and the university's websites is another dept). Tools that I could present to the math, physics, or chem department would be ideal. However I'm open for ideas or examples.

Setting up / administering Linux or whatever OS or program is not the problem. It is what to do with the cluster.

tl/dr: what to do with a cluster in a academic setting.

Answer 1

I you don't have a project in mind, folding@home sure could use your clusters!

Answer 2

Check out BOINC, a huge array of projects to choose from.

Answer 3

There are a lot of applications for which a cluster can be used. The best would be to ask the departments.

But just to name some examples:

• Biology: Sequence analysis
• Physics: Cosmology simulations
• Engineering: CFD
• Sociology: Simulating epidemics
• Economics: Derivate pricing

Answer 4

Do a machine learning project! Here are lots of standard data sets, here is an implementation of many machine learning algorithms (see also MALLET).

Or do natural language processing, e.g. machine translation of natural language! Training translation models can be very time-consuming, so usually many machines are used.

Answer 5

Set up a Hadoop cluster, with all the bells and whistles (Pig, Oozie, HBase, etc), find a source of a large quantity of data (Netflix dataset can be downloaded, Spinn3r I hear supplies academics no charge), and the students or professors will be able to find suitable projects with that.

Setting up all the framework and getting the data is a hassle. Building some interesting analysis of data on top of Map Reduce is a feasible project for part of a single semester project.