I do a lot of work with very large graphs (100s of millions of nodes), running similar sorts of algorithms, and also have found there's a lot of benefits to goi

I do a lot of work with very large graphs (100s of millions of nodes), running similar sorts of algorithms, and also have found there's a lot of benefits to going with a single machine with lots of memory. I'm happy to read that article, because the conventional wisdom at the moment is very much to go down the road of hadoop, and large scale parallelism; but from experience, there's certain graph tasks it just doesn't work well for.