The goal of FUNL is to bring meaning to data at an affordable price. FUNL incorporates commodity hardware and novel software techniques to exploit the hardware capabilities. Efficient I/O mechanisms for GPU-based graph processing, a library of graph analytic and machine learning solutions, and visualization are integrated into a unified system that provides end-to-end solutions to Big Data problems.
- Data production is increasing exponentially
- Complex processing requires expensive investments in hardware/software
- Require hardware with large amounts of RAM, high-speed interconnections and many CPUs
- Combines the I/O efficiency of PSW
- Hybrid CPU/GPU algorithms
- Compression techniques to deliver performance at a lower cost
- Use off-the-shelf desktop hardware