Beehive is a Java based framework for parallel programming of graph applications using a transactional model of task execution. Many data analytics applications for large scale graph data require parallel processing utilizing a cluster computing environment. mParallelism in many graph problems tends to be fine-grained and irregular, and it is not easy to extract parallelism through static analysis and data partitioning. This is called amorphous parallelism.
Graph problems with amorphous parallelism cannot be easily partitioned for programming using the MapReduce model.The Beehive framework addresses this problem based on transactional model of parallel programming. In Beehive, vertex-centric computation tasks for a problem are executed as serilaizable translations using an optimistic model for concurrency control.
Table of Contents
Click here to download Beehive.