Indexed on: 29 Apr '16Published on: 29 Apr '16Published in: Computer Science - Data Structures and Algorithms
The LFR benchmark is a popular benchmark graph model used to evaluate community detection algorithms. We present the first external memory algorithm that is able to generate massive complex networks following the LFR model. Its most expensive component is the generation of random graphs with prescribed degree sequences which can be divided in two steps: They are first materialized as deterministic graphs using the Havel-Hakimi algorithm and then randomized. Our main contribution are HP-HH and ES-TFP, two I/O-efficient external memory algorithms for these two steps. In an experimental evaluation we demonstrate their performance: our implementation is able to generate graphs with more than 10 billion edges on a single machine, is competitive with a parallel massively distributed algorithm and on smaller instances faster than a state-of-the-art internal memory implementation.