Skip to Main content Skip to Navigation
Theses

Edge partitioning of large graphs

Abstract : In this thesis, we mainly focus on a fundamental problem, graph partitioning, in the context of unexpectedly fast growth of data sources, ranging from social networks to internet of things. Particularly, to conquer intractable properties existing in many graphs, e.g. power-law degree distribution, we apply the novel fashion vertex-cut, instead of the traditional edge-cut method, for achieving balanced workload in distributed graph processing. Besides, to reduce the inter-partition communication cost, we present a block-based edge partition method who can efficiently explore the locality underlying graphical structures, to enhance the execution of graph algorithm. With this method, the overhead of both communication and runtime can be decreased greatly, compared to existing approaches. The challenges arising in big graphs also include their high-variety. As we know, most of real life graph applications produce heterogenous datasets, in which the vertices and/or edges are allowed to have different types or labels. A big number of graph mining algorithms are also proposed with much concern for the label attributes. For this reason, our work is extended to multi-layer graphs with taking into account the edges closeness and labels distribution during partitioning process. Its outstanding performance over real-world datasets is demonstrated finally.
Complete list of metadatas

Cited literature [131 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-01956979
Contributor : Abes Star :  Contact
Submitted on : Monday, December 17, 2018 - 1:03:29 AM
Last modification on : Sunday, October 25, 2020 - 2:47:19 PM
Long-term archiving on: : Monday, March 18, 2019 - 1:12:33 PM

File

2017PA066346.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-01956979, version 1

Citation

Yifan Li. Edge partitioning of large graphs. Social and Information Networks [cs.SI]. Université Pierre et Marie Curie - Paris VI, 2017. English. ⟨NNT : 2017PA066346⟩. ⟨tel-01956979⟩

Share

Metrics

Record views

214

Files downloads

770