distributed - Can Giraph reuse the graph in memory from previous job for subsequent jobs? -
i've around million nodes & billion edges, giraph keep graph in memory such can used again subsequent jobs or loads & stores complete graph before & after every job respectively?
what mean subsequent jobs?? when launch graph algorithm (shortest path example), giraph loads underlying graph in-memory , keeps until computation completed (all super-steps done). if check-pointing enabled giraph stores current state of vertices @ end of super-steps (depending on configurations) fault-tolerance purposes. after computation completed results stored in hdfs (or else hbase) , java processes terminated. there not rational reasons giraph keeps graph in-memory subsequent graph algorithms.
Comments
Post a Comment