Research

Biostatistics harnessing the power of Pegasus!

The Center for Computational Science (CCS) presently provides the Division of Biostatistics with High Performance Computing (HPC) expertise and the Pegasus cluster. The division is presently using 124 cores for intensive data analysis; mainly simulations, large data set analysis and statistical computing that can be done in parallel.

The HPC analysis team focuses on designing algorithms for scientific use and porting algorithms and programs to parallel computing models. Pegasus (pegasus.ccs.miami.edu) is a IBM E-server 1350 cluster is available for M/EPP (Massive/Embarrassingly Parallel) simulations and data analysis tasks (Figure 1). The cluster consists of over 1,000 compute nodes. These nodes are interconnected within their enclosures by Gigabit Ethernet and outside their enclosures with 10-Gigabit Ethernet.

The cluster also features a 35 TB High Performance File System, for maximum performance of both parallel and serial access. This system provides an aggregate of 5,230 Intel Xeon Cores and 7.4 TB of aggregated memory. These systems are configured to provide a development and job running environment similar to those that are available in other HPC centers. Portland Group, Intel and GNU Fortran, C/C++ compilers, numerical libraries including Portland groups ACML, Intel’s Math Kernel Library (MKL), GSL and IO libraries including NetCDF and HDF are installed on these systems. The MOAB job scheduler and resource management system is used to manage production jobs.

Visit the HPC website for more information.