Cray director of high-performance computing solutions Wayne Kugel
says the rising popularity of clustered systems shows the evolution
of the supercomputing arena, where both clustered and massively parallel
systems will complement one another. In bioinformatics, for example,
Kugel says that clustered systems are best used to extract base data
while parallel systems work to integrate the data. Because parallel
systems have better internal communication and less latency, they are
best-suited for enormous individual problems--the top-ranking Earth
Simulator built in Japan by NEC is one example of such a system.
(
http://www.newsfactor.com/perl/story/18440.html)