Smart to get computers to work together
Collaborative project DSZOOM developing efficient cluster techniques
Anyone who has ever used a search engine on the Internet has very probably made use of the services of a computer cluster. The primary advantage of such clusters is that they can solve problems faster while being cheap to construct. Certain types of programs cannot be run efficiently using today’s clusters, however, new research is bringing online clusters that can meet the challenge.
A computer cluster is two or more computers connected in a network that utilizes cluster software to allow them to work together as a unit. A good example of a computer cluster is the world’s largest search engine, Google, which consists of more than 10,000 PCs. Computer clusters are not a new concept; they have been in use since the 1970s, when they were primarily used to improve the reliability of computer systems.
Today’s largest servers have around 100 processors in a cabinet and can cost millions of dollars. There are several physical constraints on building bigger servers, including space and generation of heat.
Certain types of programs have nonetheless required large and expensive servers, that is, it has not been possible to run them on cheap clusters. A team of researchers at the Department of Information Technology is now at work creating new cluster software. This work is being done in a project called DSZOOM, which is a collaborative effort involving Uppsala University, Sun Microsystems, and an institute that supports cooperation between the academic world and industry, called PSCI (Parallel Scientific Computing Institute).
The aim of the research is to enable the same program to be run on a DSZOOM cluster as on a larger and much more expensive server. This will make it possible to purchase several smaller computer systems and connect them in a cluster, thus bringing down the total cost of the system considerably.
Foto: © Martin Cejie
”Certain types of programs can’t be run efficiently using today’s clusters. The aim of the research is to enable the same program to be run on a DSZOOM cluster as on a larger and much more expensive server.”