Super fast distributed computing

A low latency distributed computing framework for processing large scale Data. Faster was inspired on Apache Spark, a state of the art Big Data processing framework. Although, Spark was designed for a specific context where, input data is so large that framework overheads are acceptable. Faster implements most Spark’s operators in C++ focusing on low latency and low overhead. Also, Faster implements specialized operators for exploring data locality. With these operators, Faster can achieve a 30x speedup over Spark running PageRank on a graph with one billion edges.

Faster is implemented in C++, uses OpenMP for thread paralelism and MPI for process communication and job execution. Therefor it is perfectly compatible fith High Performance Computing clusters, supercomputers and grid engines.

Realistic random network generator

RNet is a realistic random network generator. It generates a social network from some statistical parameters. It deliberately assign a community to each node to achieve realistic social behavior.

Also, it is highly flexible, it can accept several parameters. Besides the size of the generated network, it also accepts: outdegree, indegree and community size probability distributions, probability of connecting inside a node community and probability of connecting a node to a totally random node. These parameters make possible to generate a large variety of networks with much smaller restrictions than other generators.

RNet is also very fast, it is possible to generate graphs with millions of vertices in matter of seconds! It uses multithreading to better explore modern CPU power.

A Vim plugin for Ctest + Gtest integration

View tests results inside a Vim buffer.