Novel computing facility unveiled at Indiana University: Analysis and Visualization of Instrument-Driven Data (AVIDD) supercomputer facility IU's second teraflop system.

BLOOMINGTON, Indiana — At a ceremony held at Indiana University March 26, IU's new Analysis and Visualization of Instrument-Driven Data (AVIDD) supercomputer facility was dedicated. AVIDD is an extremely powerful distributed computing facility with components at IU Bloomington, IUPUI (Indianapolis), and IU Northwest (Gary) that work together as one single large system.

What is different about AVIDD is that it has been designed by IU scientists to support data-intensive science and research. Data-intensive science requires systems like AVIDD that integrate seamlessly the ability to collect, store, analyze and visually represent vast amounts of data. The ability to do this is essential for dealing with data from atmospheric measurement, high-energy physics experiments, and seismic recordings. As research and development in all fields becomes more dependent on instruments that produce massive quantities of data, facilities such as AVIDD will be essential to transform that data rapidly into knowledge.

Indiana University Vice President of Information Technology and CIO, Michael McRobbie, who is also the AVIDD Principal Investigator, said, "The AVIDD system opens completely new doors for research at Indiana University, allowing scientists to collect, analyze, and visually represent data that were previously well beyond what they could work with. This will enable new research to be carried out in many diverse areas of science and research, such as the life sciences, geophysics, atmospheric sciences, physics, and chemistry. There are few similar facilities in existence today."

AVIDD was made possible in large part by a $1.8M grant from the National Science Foundation (NSF). Peter Freeman, Assistant Director of the NSF who heads its Directorate for Computer and Information Science and Engineering, and who was the featured speaker at the dedication commented, "The scientific community is beginning to understand that advanced computing systems in the US must be able to handle massive data sets as importantly or more so than massive computations. The AVIDD facility at Indiana University is an example of the type of cyberinfrastructure that the NSF sees as the future of advanced scientific computing in the US."

AVIDD is also one of the fastest computer systems in the US and is the second teraflop system that IU has recently acquired. (A teraflop is the capability of a computer to carry out a trillion mathematical operations per second.) This underpins Indiana University's position as one of the country's leading university centers for high performance computing and the development of cyberinfrastructure.

AVIDD consists of:

• A distributed quartet of Linux clusters with a combined total of 1.1 teraflops of processing capacity and 10 terabytes of distributed storage (one terabyte equals roughly 1500 compact discs)
• Massive data storage facilities for storage of hundreds of terabytes of data
• A network of 3D visualization systems installed in laboratories throughout the university.

AVIDD is designed to:

• Manage very large data sets
• Efficiently provide facilities for both long-running and real-time analysis of data
• Enable scientists to interact with and analyze data with advanced visualization equipment.

Alex Dzierba, professor of physics at Indiana University Bloomington, said, "The AVIDD facility is creating critical and immediate benefits for research in high-energy physics. Thanks to AVIDD we are able to analyze data collected in 1995, which were stored unanalyzed for lack of a computing facility that could provide the large-scale data management and computational resources required to interpret these data." Dzierba's research focuses on the basic structure of matter, and a new project he leads to search for a new form of matter will be guided by the analysis carried out with the AVIDD facility.

The implementation of AVIDD has required the development of new software. The challenges faced by university researchers in managing large data sets are similar to those facing private sector research - predominantly, the lack of a combination of computing power and disk storage to process the data in a timely fashion, if at all. The development of AVIDD will lead to technology and techniques that will enhance Indiana's private sector research and development, especially in medicine and life sciences.

AVIDD has an important role in education at IU as well. The system is already in use for educational purposes and is expected to play a particularly important role at IU Northwest. The equipment located on the IUN campus was provided via a Shared University Research grant from IBM. Including a component of the system at IUN opens new doors for education in information technology at this campus, which serves the highest percentage of students from traditionally underrepresented groups of any university in the state. The Linux cluster at IUN is a tangible indicator of IU's and IBM's commitment to diversity.

For more information about AVIDD, see:

http://www.indiana.edu/~uits/rac/avidd/

Technical details

Using a 1.1 teraflops Linux cluster, enhancements to IU's existing Massive Data Storage System (MDSS), and IU-designed advanced visualization equipment, AVIDD is integrated into a single system via the high-speed bandwidth of the I-Light network. The system components are detailed below.

Computational component

The computational component of AVIDD consists of an integrated quartet of IBM Linux clusters, tied together as a single computational resource. Ten terabytes of disk space are available, providing a 10:1 ratio of storage to processing capability. This ratio is unusually high compared to other existing high performance computing facilities and is one aspect of AVIDD that makes it ideally suited for data-intensive computing. A total of 470 processors provide an aggregate processing capacity of 1.1 teraflops:

• Two clusters, each with 208 Prestonia 2.4-GHz processors, provide the bulk of the processing power.
• A third cluster provides 36 64-bit McKinley processors for computer science research.
• The fourth cluster, located at Indiana University Northwest, has 18 1.3-GHzPIII processors. This cluster was funded by a Shared University Research grant from IBM and, along with advanced visualization equipment located at IUN, will play a particularly important role in the educational aspects of the AVIDD facility.

AVIDD has as one of its unique features the capability of temporarily halting a long-running computer analysis in order to meet a time-sensitive need for real-time analysis of data.

Massive data storage

The massive data storage equipment for AVIDD is being incorporated into IU's already expansive HPSS installation. Added equipment includes two IBM p640 servers, 1.7 TB of RAID disk, and six StorageTek 9940A drives. It will be possible to move 1 TB of data from disk cache into the Linux clusters in 2.5 hours.

http://storage.iu.edu/mdss.html

Visualization

IU has developed and tested three separate types of display systems to be deployed as part of the AVIDD visualization component:

• Small, portable systems with 3' x 4' passive stereo displays.
• Wall displays, very similar in technology to the portable displays, but with ceiling mounted projectors. (These may be either rear or front projected.)
• Larger, tiled display walls.

I-Light

I-Light is a very high-speed optical fiber network connecting Indiana University Bloomington, Indiana University - Purdue University Indianapolis, and Purdue University West Lafayette to each other. I-Light also connects all three campuses to the national Internet infrastructure, including Internet2.

http://i-light.iupui.edu/

News Releases    CPO Web page    IT@IU Web page

Posted 25 March 2003
Contact the UITS Communications and Planning Office
Copyright © 2003, The Trustees of Indiana University