University Times

Most of us generally are familiar with the concept of communication networks and their use. These computer networks, systems, services and data shared across an organization typically are referred to as enterprise information technology infrastructure.

Cyberinfrastructure implies the use of available enterprise infrastructure combined with purpose-built scientific systems for research and education.

Indiana University has developed one of the more widely cited definitions used by IT organizations supporting scientific research:

Cyberinfrastructure consists of computational systems, data and information management, advanced instruments, visualization environments and people — all linked together by software and advanced networks to improve scholarly productivity and enable knowledge breakthroughs and discoveries not otherwise possible.

CSSD has upgraded the Pittsburgh campus cyberinfrastructure using funds obtained through the National Science Foundation (NSF)’s CC*IIEx (Campus Cyberinfrastructure: Infrastructure, Innovation and Engineering) program. This multiyear program invests in improvements and re-engineering to support computational science and network integration activities tied to bridging campus cyberinfrastructures and achieving higher levels of performance and predictability for science applications and for distributed research projects.

What does this project mean for cyberinfrastructure at Pitt?

More capacity and advanced networks

Partnering with Pitt faculty, peers from other NSF-awarded institutions and other experts in research-IT communities, CSSD upgraded campus network connections to provide more capacity to local high-performance computing resources (HPC) and to cyberinfrastructure resources at the national level.

The project’s technical work included upgrades to 100Gb/s of two campus links connecting to the University’s Network Operations Center (NOC) and to Internet2.

These upgraded links (previously at 10Gb/s) provide more bandwidth for access and movement of research data to and from University HPC resources hosted at the NOC and to sites on Internet2. Internet2, an advanced national network backbone, supports the development and deployment of new applications being developed within research and education communities and is comprised of hundreds of universities, national labs, government agencies and other research and education networks.

These upgraded links form a foundation. From this foundation, additional enhancements can be developed — in a community-driven approach — to meet the rapidly expanding needs of collaborative and multidisciplinary data-driven research here at Pitt.

The Center for Simulation and Modeling and the Pittsburgh Genome Resource Repository already are using this new capacity to more effectively manage the movement of large genomics data sets to mass storage and computational resources at the Pittsburgh Supercomputing Center. They also are taking advantage of Pitt’s ScienceDMZ.

ScienceDMZ

The NSF project established a ScienceDMZ network at Pitt, distinct from the general-purpose campus network and engineered specifically to support data-intensive science. Our ScienceDMZ works by reserving a portion of the network for friction-free, high-performance networking in an environment separate from the business or enterprise systems that constitute the great majority of Pitt’s local area network (PittNet). This separate environment provides a relatively small space that has been optimized for the wide area data-movement needs of systems whose effectiveness depends on high-speed flows of big data. It also gives researchers greater networking capacity to work with colleagues at other institutions with a well-known, managed connection point for campus bridging.

For example, researchers in the Department of Physics and Astronomy are using the new capacity and ScienceDMZ configuration for analysis of particle physics as part of the ATLAS experiment using data from the CERN Large Hadron Collider.

New tools for data movement

The new Enterprise Data Transfer Node (DTN) service was developed and implemented as the first service available on the ScienceDMZ. A dedicated system using Globus Connect software provides faculty and researchers a convenient and effective way to transfer data files to and from Pitt HPC resources. Working seamlessly with the InCommon federated identification architecture and the Globus Online service, Pitt users and external collaborators can use their own campus credentials and not have to manage multiple usernames and passwords.

The Laboratory for Computational Transport Phenomena in the Swanson School of Engineering will use the new DTN and Globus software to move large-scale data sets from simulation and modeling conducted at supercomputing facilities to local HPC resources.

Tools for network management

Finally, the project also included the development and deployment of an initial set of perfSONAR endpoints on the ScienceDMZ for network management and troubleshooting. perfSONAR is a widely deployed test and measurement infrastructure used by science networks and facilities worldwide to monitor and ensure performance. With these endpoints in place, CSSD can test and measure network performance, as well as archive data, in order to pinpoint and solve service problems that may span multiple network and international boundaries.

Technology, people, effect

Most simply, cyberinfrastructure refers to an infrastructure for knowledge, one that improves faculty and researchers’ ability to take advantage of capacities to develop new tools, methods and knowledge for broader impact within and beyond their immediate areas of focus. As such, cyberinfrastructure is a key enabler of Pitt’s strategic commitment to addressing the “grand challenges” facing our world.

Brian Stengel pursues development opportunities in research computing for the University and works closely with faculty and researchers on grant proposals and collaborative research projects.