4

I have graph which contains 20 millions nodes and 18 millions relationships. I had tried a lot of products for visualization such as Sigma.js, D3.js (in Neo4j), igraph (in R), graphvis (dot doesn't generate output file), gepthi/cytoscape/pajek (I do not have enough RAM but I have only 8Gb RAM).

Maybe I am doing something wrong? May I use CUDA to render my graph? Maybe it would be better if I write a program for render?

Dmitry Shvedov
  • 3,169
  • 4
  • 39
  • 51
Andrew Polovinkin
  • 41
  • 2
  • Are you trying to show all the individual nodes and relationships? I feel like that would be a bit excessive. It all depends on what you want to visualize through the data, for my university project I simply pictured one node and all it's relationships (as one nodes also) simply to display how they are connected, I had way too many nodes and relationships to show them all. – Martin91 Apr 07 '16 at 22:55
  • Yes, I wanted show all nodes and relationships but now i think to show node and dependent nodes with relationships. its good idea! – Andrew Polovinkin Apr 07 '16 at 23:08
  • 1
    I answered a similar question, but on uxstackexchange --> http://ux.stackexchange.com/questions/89227/graph-exploration-and-data-overload/ . One of the suggestions was for a hive plot if you want to see properties of the graph rather than every node laid out. Beyond that it's aggregation / filtering, and with your size of graph that'll need to be done before you get to the visualisation stage. – mgraham Apr 08 '16 at 09:23

0 Answers0