Bachelorarbeit
Distributed Graph Partitioning for Largescale Graph Analytics  
Bearbeiter 
Lukas Rieger

Betreuer 
Dr. rer. nat. Christian Mayer
Dr. rer. nat. Muhammad Adnan Tariq 
Prüfer 
Prof. Dr. rer. nat. Dr. h. c. Kurt Rothermel

Ende  2016/12/01 
Beschreibung  
The trend for Big Data analytics continues to attract academia and industry to find feasible solutions for the challenge of processing huge data sets. Large companies such as Google, Facebook and Yahoo are seeking data analysts that are able to extract value out of web, user or financial data. Because of the size and complexity of the data, data analytics are usually performed in the cloud (and hence in large data centers) and distributed across many computers in order to parallelize execution and improve latency perceived by the data analysts. An enormous amount of data comes in the form of interconnected graph data (e.g. the WWW of linked websites or social networks of linked users). To analyze these billionscale graphs in parallel, distributed graph processing systems such as Pregel [1], GraphLab [2], and PowerGraph [3] have emerged in the last years. These frameworks parallelize graph computation by dividing the graph onto different machines using a graph partitioning algorithm. To minimize information exchange across machines during graph processing, an effective partitioning algorithm of the graph is vital. Goal and TasksIn this thesis, a distributed partitioning algorithm should be developed addressing realworld graphs with billions of edges. In particular, the goals and tasks of this work are the following:
The results have to be presented and documented in a written report (English or German language). Bibliography[1] Joseph E Gonzalez, Yucheng Low, Haijie Gu, Danny Bickson, and Carlos Guestrin. Powergraph: Distributed 