"Efficient processing of large graphs is challenging. Graph algorithms often exhibit poor locality of memory access, very little work per vertex, and a changing

"Efficient processing of large graphs is challenging. Graph algorithms often exhibit poor locality of memory access, very little work per vertex, and a changing degree of parallelism over the course of execution [31, 39]. Distribution over many machines exacerbates the locality issue, and increases the probability that a machine will fail during computation." https://kowshik.github.io/JPregel/pregel_paper.pdf https://kowshik.github.io/JPregel/pregel_paper.pdf ^ (Pregel is Google's batch graph processing framework) FWIW most graph computations can fit on a single powerful machine these days. There really aren't that many that require distributed processing.