A Research Infrastructure for Collaborative
High-Performance Grid Applications

Thanks to the  National Science Foundation for supporting this research under a grant entitled numbered EIA-0202048.


Investigators: David S. Wise, Beth Plale, Andrew Lumsdaine, Geoffrey Fox, Randall Bramley, Kay Connelly, Dennis Gannon, and David Leake.


This is an infrasturcture grant supporting the projects linked below.
Equipment purchased so far includes:

January 2005

• 16-node Apple Xserve G5 Cluster. Cluster Name: Woton
  • 16 Compute Nodes
    • Dual 2.3GHZ PowerPC G5
    • 4GB Ram
    • Gigabit Ethernet
    • Infiniband
    • 80GB SATA Local Drive
  • 1 Head Node
    • Dual 2.3GHZ PowerPC G5
    • 8GB Ram
    • Gigabit Ethernet
    • 2x250GB SATA Local Drives

December 2004

• Part 1 - Cluster Name: Tyr
  • 16 Nodes
    • Dual 2.0GHZ Opterons
    • 16GB Ram
    • 40GB IDE Local Drive
    • CDRW-DVD Combo Drive
    • Floppy
• Part 2 - Cluster Name: Odin
  • 128 Compute Nodes
    • Dual 2.0GHZ Opterons
    • 4GB Ram
    • 40GB IDE local Drive
    • Floppy
  • 8 Head Nodes
    • Dual 2.0GHZ Opterons
    • 8GB Ram
    • 40GB IDE local Drive
    • CDRW-DVD Combo Drive
    • Floppy
• These functionally seperate clusters are tied together by Gigabit Ethernet. The machines labeled Compute Nodes in the Odin Cluster are also connected via Infiniband. The Head Node machines, as well as all of the machines in Part 1, are connected via Fibre Channel to a 3.5TB SAN Array, constructed of 16 250GB SATA drives in a RAID configuration

July 2004

• Access Grid node for LH101.

August 2003

• New staff member Jon Burgoyne to join department.

July 2003

• 8-node Aspen Systems Linux cluster - Cluster Name: Thor
  • Each node is a dual-processor system with 2.8GHz Intel Xeon processors.
  • Each has 2GB RAM memory.
  • Each has 6 PCI slots (4 of which are PCIX).
  • RedHat Linux 8.0.
  • Connected to the SAN via a 2Gbps Qlogic SANBlade 2300 Fibre-Channel Host.
  • Ethernet connected now.
  • Myrinet, InfiniBand, and Quadrics connectivity.
• Storage Area Network (SAN)
  • 2.2TB of data storage.
  • 16-port, 2Gbps Qlogic SANBox2 Fibre-Channel switch.
  • Mylex FFx-2 Fibre-Channel RAID controller.

The NCSA Alliance Expedition "Scientific Portal" is the gateway to the Grid on the TeraGrid. It provides tools for Grid services for use by large numbers of collaborating scientists. Through a Scientific Portal, a scientist can accomplish key tasks including authenticate once, configure and launch a set of applications, and manage the input and output data sets all from a desktop environment. This expedition will deliver and deploy a Science Portal toolkit that uses the best-of-breed technology that can be easily adapted to any grid application

XCAT is the Indiana University, Extreme Lab implementation of the Common Component Architecture (CCA). It supersedes our earlier effort called CCAT. The CCA specification describes the construction of portable software components that may be re-used in any CCA compliant runtime frameworks. It is expected that the CCA group will build frameworks that are tuned for a variety of application environments. Some cases are designed for applications that run on massively parallel computers.

XSOAP is the underlying communication layer of our implementation of Common Component Architecture (CCA) specification, called XCAT. Our XCAT implementation allows to write CCA compliant components in both Java and C++ that can work in distributed Internet environment. This allows to create powerful scientific applications that can take advantage of emerging web services.

The Security for Ubiquitous Resources Group (SURG) The Security for Ubiquitous Resources Group (SURG) investigates security issues in ubiquitous computing. The key attributes of our target environment are:
Sensors and processors embedded in the physical environment
Mobile devices and users
Collaborative applications
Context-aware (physical and virtual) applications
Non-technical users

LAM/MPI is a high-quality open-source implementation of the Message Passing Interface specification, including all of MPI-1.2 and much of MPI-2. Intended for production as well as research use, LAM/MPI includes a rich set of features for system administrators, parallel programmers, application users, and parallel computing researchers.

OSCAR version 2.2.1 is a snapshot of the best known methods for building, programming, and using clusters. It consists of a fully integrated and easy to install software bundle designed for high performance cluster computing. Everything needed to install, build, maintain, and use a modest sized Linux cluster is included in the suite, making it unnecessary to download or even install any individual software packages on your cluster.

dQUOB (dynamic QUery OBjects)
Beth Plale is a middleware system providing continuous evaluation of queries over time sequenced data. The system provides access to data in data streams through SQL queries. Queries are portable entities that are embedded into the data streams at runtime, and managed remotely during application execution.

The dQUOB system conceptualization of 'data streams as a database' is an intuitive way for users to think about managing time sequenced data in data streams. dQUOB SQL queries have the power to filter and aggregate data, combine streams, and create new streams.

The Opie project offers language support for Morton-ordered matrices. Its core is a compiler to transform programs in C and other iterative languages into equivalent programs that replace row-major matrix representation and cartesian indexing with Morton-order matrices, similarly indexed. Support for recursive/quadtree indexing is also a goal.

The Arcee project's goal is to develop a paradigm of programming, as well as support for it, to express efficient, block-recursive algorithms for standard matrix problems. Support includes augmenting usual gerneic progamming tools with "recurators," the analog of iterators of conventional looping style. It is supported by libraries regenerated with Opie

MxN Research: Increasingly software components and even complete applications in scientific computing are being composed together to create new large scale multidisciplinary simulations. High performance software is often SPMD parallel, which leads to the MxN problem: connecting components running on differing numbers of processors. In research partly funded by the Department of Energy we are exploring the semantics of parallel remote method invocation, specification methods for data distribution templates, and dynamic data redistribution between parallel components.