Systems & Resources

High Performance Computing Clusters

Pegasus – Top 500 System (389)

  • Documentation (update pending)
  • IBM iDataplex and Blade Chassis architecture
  • 10,000+ x86 (5,500 2.4 GHz Sandy Bridge, 4,500 Harpertown) 200 TFlops aggregate performance
  • FDR-14 Infiniband Interconnect
  • 10 Gigabit Ethernet LAN connections
  • 10 Gbit Internet 2 connection
  • 40 Intel Phi Co-processors (61 cores each) – 40 TFlops Performance
  • 18 TB Aggregate RAM
  • 500 TB High Performance Parallel Storage (GPFS)
  • Full LSF Scheduler support
    • Multi-tier approach – SSD, SAS, and SATA are utilized
  • Optimized Libraries and Compilers: 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

These systems are configured to provide development and job running environments.
For access to Pegasus, you will need a CCS user account.  Click here to request an account.

See photographs of Pegasus being built here!

Ares

  • Documentation
  • 576 CPU IBM Power5+ SMP cluster, with high-speed interconnect
  • 5 TFlops
  • Over 1 TB RAM
  • 75 TB High Performance Parallel Storage  (GPFS)
  • Load Leveler Scheduler
  • Optimized Compilers and Libraries: IBM and GNU Fortran and C/C++ compilers, numerical libraries including ESSL, PESSL, MASS, LAPACK, BLAS, and IO libraries including NetCDF and HDF

These systems are configured to provide development and job running environments.
For access to Ares,  you will need a CCS user account.  Click here to request an account.

Relational Database Cloud

  • MySQL Cluster
    • Highly available databases
    • Data replication
    • 5 TB Tier 1 storage
    • Multi-core server environment
    • Complete backup
  • Oracle Cluster
    • Highly available
    • Data replication
    • 15 TB Tier 1 storage
    • Multi-core server environment
    • Complete backup

Flexible User Storage Environment [Cloud]

  • 1.8 PB of dynamically allocated storage
  • Full cloud service
  • OS X/Windows/Linux CIFS access
  • Group controlled
  • User authenticated
  • Parallel filesystem (GPFS)
  • Redundant serving and storage
  • Designed for research data
  • Backup daily for DR (one copy)

Data Analysis and Visualization Distributed Cloud (DAViD Cloud)

  • 800 TB of dynamically allocated storage
  • Client based access
    • SSH, SFTP, ASCP, BBCP
  • Completely interactive access (not scheduler based)
  • Multiple 10 Gig connections to processing units
  • Highly available/redundant server grid (10 servers)
  • 15 Independent storage stacks (scalable to 300TB each)
  • Analysis and visualization tools.
  • Supported applications include:
    • Grads, Vapor, Paraview, Matlab, R, Octave, Perl, Python, C, Fortran, C++, Java

Software

  • All common scripting and programming languages supported
  • Currently over 300 software tools supported and installed
  • Optimized compilers including Intel, Portland, and AMD
  • All common optimized libraries

Expertise

  • Code porting
  • Algorithm design, development, and implementation
  • C, C++, Fortran, Perl, Python, Java
  • RDBMS design and implementation (Oracle, MySQL, PostgreSQL)
  • SOA and PaaS (platform as a service) architecture
  • Data management system design and development
  • HPC architecture, design, development, and implementation

HPC  Service Center

The High Performance Computing (HPC) core is an established university wide service center with a mission to provide the University of Miami’s academic community with comprehensive HPC resources ranging from hardware infrastructure to expertise in designing and implementing HPC solutions. Click here for more information on the services available through this core.

Contact Us

For assistance with any of our HPC Systems and Resources, please email CCS HPC Support.