GoldenOrb offers open source variant of Google's Pregel
Analytics company Ravel has announced it is releasing GoldenOrb, its massive-scale graph analysis software, as open source. GoldenOrb is based on the ideas behind Google's Pregel architecture which is in turn inspired by the Bulk Synchronous Parallel Model developed in the 1980s.
Massive-scale data problems are commonly associated with projects such as Apache's Hadoop, which itself began as an implementation of Google's Map/Reduce algorithms and has since become a platform for tackling massively distributed computing. GoldenOrb, and Pregel, take on a different class of problem from Map/Reduce, one where the data is represented not by millions or billions of records, but as millions or billions of connections which form a graph.
It uses technology from the Apache Hadoop and Zookeeper projects as a foundation to build a framework where an "OrbRunner" orchestrates the processing of Vertex objects by VertexBuilders and VertexWriters. The developers recommend users read Google's paper on Pregel to acquire an understanding of how the graph alogrithms are applied on a clustered system.
Ravel believes that GoldenOrb could find applications in social graph analysis, data-mining, fraud detection, energy grid optimisation and network analysis. It points out that not only are there no commercial implementations of GoldenOrb, there are no plans to develop any productised versions.
GoldenOrb is available from the project's GitHub repository and is licensed under the Apache Licence 2.0.