The main function of high performance clustering (also called high performance technical clustering, HPTC) is to scale computer power. To do this it is necessary to split the job into many `sub jobs`. The management unit used to coordinate all these sub jobs, and a special result node could be single points of failure. Typically, these components will be protected by an availability cluster.
The crash of one of the computing nodes isn't a disaster because the job could be done by another node. The performance of the complete system would be hampered, but the total perfomance wouldn't be endangered.
One example of a high performance cluster was the CLiC (Chemnitz Linux Cluster) of the University of Chemnitz in Germany till June 30th 2007. There were 528 computing nodes using the LINUX operating system.