The Rensselaer Institute for Data Exploration and Applications (IDEA) provides a set of resources for campus users needing dedicated computation for applications ranging from data analytics to deep learning systems. Key facilities include:
- A high-performance computing cluster with five Intel Xeon and two AMD compute servers in various configurations ranging from 24-40 cores (48-80), 256GB-1TB RAM, and up to four GPUs per machine (Tesla K40m or Titan RTX). The IDEA Cluster includes two dedicated storage servers totaling more than 40TB of usable space. The IDEA Cluster is designed for dedicated data mining, machine learning, and neural computing-intensive jobs using popular toolkits.
- IDEA provides oversight for three high-performance hypervisors to host project-specific virtual machines for IDEA, Tetherless World Constellation and dedicated HEALS research
- Researchers and students within IDEA have ready access to leading data analytics platforms including RStudio Server, Jupyter Notebooks and MATLAB, all enhanced with parallel computing capability; machine learning technologies including TensorFlow and PyTorch; and scalable database technologies including ElasticSearch. Data analytics applications are regularly deployed to campus users and others via the Web using the R Shiny platform.
- Researchers within IDEA can scale up their code developed on the IDEA Cluster and, with code compatibility, run them on Rensselaer's AiMOS supercomputing system which provides over 130,000 cores and 1500 high-end GPUs. The machine is rated 24th in the world, and currently has a rate over 8 petaflops. Details on the AiMOS supercomputer can be found at https://www.top500.org/system/179781.
- A visualization laboratory featuring Rensselaer's Campfire visualization suite – a patented visualization system for multiscale, collaborative data analytics coupled with several large screen systems to provide an immersive capability.
Use of these facilities are available to RPI researchers at no charge for projects that are externally funded.
See also Accessing the IDEA Cluster