AVIDD logo

 

Program

What is AVIDD?

System Diagram &
Photos

Technical Information

Demonstrations

Accessing New Physics With AVIDD

Scott Teige
Department of Physics

Current and future experiments in high energy physics will probe the fundamental nature of matter with unprecedented precision. Theoretical expectations predict entirely new forms of matter, yet to be convincingly observed, will be accessible to these experiments. Accessing this new physics requires the analysis of very large data sets (Terabytes to Petabytes). This demonstration will introduce the physics and demonstrate some techniques used to exploit the resources of AVIDD and the HPSS.

Collaborative Crystallography

Dr. John C. Huffman
IU Molecular Structure Center, Department of Chemistry
Informatics Research Institute, School of Informatics

A group of projects (Xport, "Crystallographer In the Box" and SCrAPS) all with a common goal:

The XPort project (X-ray Portal) addresses the issue of remote instrument access, as well as Internet-based collaboration tools for researchers working with crystallographic data. The project allows "same-as-being-there" access to the remote instrument by offering a variety of video and audio capabilities, as well as remote data access and data storage.

The "Crystallographer in the Box" project aims at the challenge of real-time analysis and calculation of massive crystallographic data sets and visualization of the results using IU's distributed resources.

The SCrAPS (Service Crystallography at Advanced Photon Sources) project provides a mechanism whereby service crystallography laboratories in US academic institutions can access the specialized instrumentation and tunable, high-brilliance radiation sources available only at synchrotron sources located at national laboratories.

Top of page

High Performance Storage for AVIDD Clusters (HPS-AC)

Anurag Shankar, Haichuan Yang, Jeff Russ, Chris Garrison
Distributed Storage Services Group, UITS

  • A dedicated, high performance archival storage system has been designed and implemented for the IU AVIDD clusters. Maximum data transfer rates achieved by the HPS-AC are around 100 megabytes per second.
  • HPS-AC is highly cost effective due to its use of IU's pre-existing High Performance Storage System (HPSS) software and hardware infrastructures.
  • HPS-AC uses parallel storage techniques to enhance data transfer in and out of the AVIDD clusters. Data are striped over four data mover nodes, over their attached disks, and over four tape drives to achieve four times the performance of a single mover, disk, or tape.
  • HPS-AC hardware includes three IBM p640 servers with 1.7TB of attached, high-performance RAID disk and six StorageTek 9940 high-end tape drives. HPS-AC hardware is located at IUPUI.
  • The IUB AVIDD cluster uses the HPS-AC over the I-Light network between IUB and IUPUI and thus eliminates the need for local high performance archival storage at IUB.

MxN: Connecting Parallel Programs Scalably

Felipe Bertrand, Yongquan Yuan, Randall Bramley
Computer Science Department

The MxN problem is one of connecting parallel programs which may be running with different run-time systems and incommensurate numbers of processors. Until now this has been done either by requiring the codes to use the same run-time system and to be started simultaneously in a single run-time system, or by serializing the communications. The first requires extensive changes to code; the second prevents scalability. As part of the Common Component Architecture Forum, we have developed a system using the MPI I/O interface, allowing users the convenient fiction of simply reading or writing to files but in fact having parallel connections automatically created and managed.

Top of page

Visualization and data manipulation of multidimensional LC/IMS/TOF dataset

Xinfeng Gao, Sunnie Myung, David E. Clemmer
Department of Chemistry

A combined liquid chromatography / ion mobility / time-of-flight mass spectrometry (LC-IMS-TOF) method recently developed in David Clemmer's group provides the possibility of multidimensional separations of complex proteome mixtures, which offers insight on the qualitative and quantitative analysis of complex biomolecular mixtures. One of the biggest challenges is to visualize such a dataset and obtain quantitative and qualitative information. Typically a large data matrix (8000 x 1000 x 200) is produced in each LC-IMS-TOF experiment. In order to extract useful information, OpenDX software with our custom-made script running on the AVIDD cluster in a client/server environment has been utilized for visualization and manipulation of the three dimensional dataset.

Top of page

Created March 14, 2003
http://www.indiana.edu/~uits/rac/avidd/demo.html
Send comments and suggestions to rac at indiana.edu
Copyright 2003, The Trustees of Indiana University