Posts

Joining Hands

Clusters and multiprocessing systems are both very interesting topic, and makes one a person think of ways he can advance his home system with low costs. A cluster is a system where more than one computer is linked together and thought of one computer. There is a number of types of clusters.

The first type of cluster I’m going to talk about is tight coupling. Tight coupling is where multiple CPU’s share BUSes and input/output devices. An example of this type of cluster, even though we do not tend to think of it as a cluster is a personal computer with a dual core. Intel has provided the name with multiple cores. When there is more than one processor they may either be symmetric or follow the Master-slave model. As the name suggests in the symmetric model each CPU is in control, no other CPU is in charge. In the master-slave model there is a master CPU in which responsible in organizing the load.

In loose coupled systems each computer is thought to be able to stand alone; they do however share the tasks. In this system there are two models that can be used the share nothing model or the shared disk model. Their names explain how they work. With the shared disk model there must be a disk on some type of network or BUS that all computers are able to access.

Beowulf Clusters are very interesting. Despite their cool name, they are considered cost efficient. If someone is able to create a Commodity-Off-the-shelf Beowulf cluster a very powerful computer may be created. Google uses a Beowulf cluster, that said imagine the type of Beowulf cluster one may create at home for simple uses as a gaming computer.

Grid computer is also an interesting topic. I will not elaborate on this topic; however I will only include some information about how grid computing was used in research that the University of California at Berkley. An application was installed to computers, and when idle (screen saver was running) the installed software would use the computers CPU to analyze data, and return results to a huge database.