To help you get started, we have a collection of sparse matrix structures, some of which are amenable to row-wise distribution to mappers in map-reduce programs. If your connectivity program is basically just the transitive closure of the graph, then that would probably suffice (although I would worry about the output getting large). The MST algorithm will probably stress things a bit more.
On Mon, May 31, 2010 at 2:47 PM, Neal Clark <[email protected]> wrote: > I will have to take a closer look at the Mahout data structures before > I can be certain how hard it would be. >
