Hammer lightweight graph partitioner based on graph data volumes

Abstract

The graph partitioning challenge is well known and ongoing classical problem. Many heuristic methods tried to propose solutions focusing mainly on load processing and cost-efficiency. With the emergency of big data technology, the graph partitioning challenge became even more demanding, as an imminent need to handle big volume of data in real time. This reveals a new challenge as most of the existing studies does not consider the volume metric with their streaming graph algorithms causing imbalanced workloads and graph storage. With this article, we propose a specific lightweight algorithm which we called “Hammer” Algorithm. Our proposed Hammer algorithm is a streaming graph based on volume metric to ensure optimal load processing and communication cost efficiency. Our proof of concept was done on real world dataset and the Hammer algorithm showed considerable performance against some existing graph partitioning algorithms.