Douglas Eadline writes “A discussion of dynamic parallel execution (or losing control at the high end) is presented. In a previous column, a discussion about programming large numbers of cluster nodes was given. By large, the author means somewhere around 10,000. At the end of the article, the author had promised to mention some real alternatives to MPI and even suggest some wild ideas. This article expands on the these ideas and develops the solution a bit further.”
Link: clustermonkey.net
