Cluster Computing

From Redbrick Wiki
Revision as of 02:23, 20 April 2007 by Receive (talk | contribs)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)

In the summer of 2001, Redbrick members built a Beowulf Linux Cluster for the School of Computing, DCU. We used 23 old lab computers that would have been thrown out, were granted access to a lab (LG28) in which to put it together for the month of July, experimented with a few different setups, and finally gave the school a system that was used for research in the "Modelling & Scientific Computing" Research Group.

CA4 projects using the cluster included analysis of digital video (cutting shots & scenes), and parallelisation of the QCL (Quantum Computing Language). Postgraduate research was conducted mostly on the analysis of biological data (DNA & protein sequences), but also on models of biological systems (immune system response, drug dissolution).

Further Developments:

With the loss of a postgrad admin for the cluster (Karl Podesta, who left DCU to join Securelinx Ltd), and the arrival of a commercial cluster in the School of Computing (bought through a tender process for the National Institute of Cellular Biotechnology or NICB, 16 dual-CPU nodes Dell Poweredge running Red Hat Linux), the old Redbrick cluster has been since decommissioned (roughly 2004). It provided a platform for 2 years of research within the "ModSci" research group while it lasted. The new NICB cluster is now only available for use via special request by students & staff of NICB, or of the Masters in Bioinformatics, and not open to all students. A new commodity cluster specifically for the School of Computing has been set up by Neil Costigan working in cryptography on elliptic curves (DCU Crypto Cluster).

Cluster Computing in General

Almost all the major supercomputers in the world currently (2006) are clusters. Clusters are just a bunch of computers connected together in an ordinary network, but which use special software on top of the O/S (such as "MPI" programs) for a dedicated cause (such as running a simulation of something sciencey). Clusters are cheap to put together, so they are now (in the last 5 years) commonly used in scientific labs or universities all over the world. Another initiative, called "Grid Computing", aims to connect all of these clusters together into one big blob of computing power (sort of like the electricity power grid, which you just plug into to get power on demand!). Ireland has a Grid itself, called Grid Ireland. National grids are currently developing, and are also being connected accross countries (there is an EU Grid, and also a TeraGrid in the United States).

Related Links: