TeGViz. Distributed Tera-scale Graph Generation and Visualization.
  1. I. OVERVIEW
  2. How can we generate and visualize tera-scale graphs efficiently? Graph generation and visualization are used in graph mining research with various applications including simulation, sampling/extrapoloation, and graph understanding.
    We proposes TegViz, a distributed Tera-scale graph generation and visualization software. It consists of two modules:

    (1) The graph generation module Teg generates a wide range of graphs including Erdos-R'enyi random graph and realistic graphs including R-MAT and Kronecker directly on distributed systems.

    (2) The visualization module Net-Ray summarizes graphs using spy plot, distribution plot, and correlation plot to find regularities and anomalies effectively and makes it easy to understand generated graphs.
    TeGViz provides:

    > Tera-scale graph generation

    > Fast generation using distributed systems

    > Visualization of Tera-scale graph

  3. II. CODE
  4. We implemented TeGViz on top of Hadoop.

    Currently, TeGViz supports:

    - Random graph generator

    - R-MAT graph generator

    - Kronecker graph generator

    - Visualization of Tera-scale graph with spy plot, distribution plot, and correlation plot

    The binary of TeGViz is available. You can download it here.
  5. III. PAPER
  6. TeGViz: Distributed Tera-Scale Graph Generation and Visualization.
    ByungSoo Jeon, Inah jeon, U Kang
    15th IEEE International Conference on Data Mining (ICDM) 2015, Atlantic City, USA.
    [PDF] [BIBTEX]
  7. IV. PERFORMANCE
  8. The following figure shows the performance of our propsosed graph generator,TeG.
    Comparison of the running time between TeG and existing graph generators. The label 'O.O.M.' means 'Out Of Memory'. In R-MAT and random graph, existing generators run out of memory when the numbers of edges are beyond 10^6 and 10^7 each, respectively, while TeG continues to run. Also, our Kronecker graph generator runs beyond 10^7. Note that TeG generates up to 16384x larger graphs than existing generators.
  9. V. EXAMPLES OF TEGVIZ RESULTS
  10. VI. PEOPLE
  11. ByungSoo Jeon (Department of Computer Science and Engineering, Seoul National University)
    U Kang (Department of Computer Science and Engineering, Seoul National University)

Last Update: Oct 26, 2015