Development and Investigation of an Interoperable, Hybrid Technology System for Local/Remote Scientific and Visual Computing

Point Of Contact: Robert A. Ballance, ballance@ahpcc.unm.edu
Principal Investigators:

• Susan Atlas, AHPCC & Physics&Astronomy
• David Bader, AHPCC & EECE
• Robert Ballance, AHPCC & CS
• Thomas Caudell, AHPCC & EECE
• Brian Smith, AHPCC & CS
• John Sobolewski, AHPCC, CIRT, and EECE

This project addresses the issues of interoperability in scientific computing environments. We will proceeed by assembling and investigating a hybrid computing environment that combines differing computing platforms (hardware and software), networking fabrics, storage subsystems, and programming methods. Selected real, scientific problems---including the biophysics of protein folding, the quantum mechanics of materials engineering, the dynamics of atom-ion collisions, the physics of multiphase flows, characterization of data mining, and the optimization of workloads in high performance computing---will serve as test cases for investigating the structure and dynamics of interoperating hybrid computer clusters. Integrated visualization using cluster resources will provide a means for scientists to interact with their computations dynamically in real time. The resulting system will be a highly productive environment for high-performance computing and visualization.

The project has 3 phases:

1. Construct an IBM Netfinity/Gigabit Ethernet Linux Cluster using PCI based adapters. This phase will study processor and network performance, interoperability with the Albuquerque High-Performance Computing Center (AHPCC) Linux Superclusters via an IBM SP Switch Router, and will initiate a comparison study of distributed visualization using software rendering versus hardware accelerated rendering.
2. Commencing with the installation of an IBM SP Winterhawk-II system connected by a Trailblazer switch and the attachment of Scalable Graphics Engines (SGE) to the Trailblazer switches at both the Maui High-Performance Computer Center (MHPCC) and AHPCC, issues such as workload characterization, scheduling, remote access, hybrid programming, and parallel I/O performance will be investigated. Scientific applications will be ported to exploit the mixed-mode parallelism and the capability for doing large-scale computation and concurrent, immersive visualization to steer and track using both the SGE and the clusters.
3. Finally, the entire system will be migrated toward complete MPI interoperability through switch-attached interconnection and an SP switch router. Additional issues such as remote visualization, web-based access and application performance, will round out this technically challenging project.

As part of the project, UNM researchers, in collaboration with researchers at Sandia National Laboratory, Los Alamos National Laboratory and IBM Research, plan to develop and port a number of computational applications to this computing environment. UNM will also connect the hybrid cluster to its existing AHPCC Linux Supercluster -- a shared resource on the National Computational Grid -- to demonstrate complete interoperability in Next Generation Internet applications such as portal-based supercomputing, remote graphics delivery and rendering, and Internet multicasting for scientific collaboration.

The following is a list of the UNM SUR Project milestones to be completed for phase I:

• Demonstrate the basic operability of the Netfinity cluster using the 10/100/1000 Ethernet fabric

  The tasks required to accomplish this milestone include the installation of the Linux operating system, a file system and the open MPI communication system on the 8 Netfinity nodes. Correct operation of MPI will be validated by using standard communication validation programs which accompany the MPI installation and other communication test programs provided by UNM researchers.

• Demonstrate the effective parallel operation of the cluster.

  This milestone requires the completion of several subtasks:

  • running local scientific application programs in parallel using multiple nodes
  • running an application program that relies on the storage system
  • running a parallel job demonstrating the communication between the Netfinity cluster and one of the UNM Linux Superclusters
  • running an application that demonstrations the graphics capabilities of the cluster
• Distribute preliminary baseline benchmarks on the Netfinity cluster to project participants.

  The benchmarks will measure both computational and communication performance of the Netfinity cluster.

• Provision of training courses on the effective use of both the Netfinity and the SP2 cluster at AHPCC and MHPCC.

  The training courses will use modules developed by IBM's Advanced Computing Technology Center.