Loading my own partitioned files already to powergraph (graph analytics engine)

User 1030 | 12/6/2014, 11:00:39 PM


I want to provide my own partitioned files for a certain graph to powergraph. Is it possible? I understand there is a phase where powergraph does "finalization" of the graph (all machines communicate meta deta of the partitions). Will it do it automatically if I let every machine load its own partition from the disk? or Do I need to modify the code explicitly? if it is the later, where can I find the snippet of code that does this?


User 6 | 12/7/2014, 6:36:08 AM

The functionality of hard coded partitioning is not supported in powergraph. You may need to change the code for that. The place to start looking is <a href="https://github.com/graphlab-code/graphlab/blob/master/src/graphlab/graph/distributed_graph.hpp">here</a>.

I suggest also taking a look at <a href="http://ipads.se.sjtu.edu.cn/projects/powerlyra.html">powerlyra</a> where additional partitioning strategies are implemented.