Last week a research team from Stanford has announced the results of running a simulation program on a million-core Supercomputer . Then both the Wired article " Researchers Set Record With Million-Core Calculation " and GigaOm's " While we waste four cores, scientists use a million at a time " did some Hadoop name-dropping when rewriting this announcement. But articles failed to make the difference between parallel data processing and parallel computing. In the space of distributed computing, there are different flavors of architectures: high performance computing systems, grid computing, data grids, etc. The purpose and constraints of these massively parallel computing models are fundamentally different.
via NoSQL databases