perfSONAR Workshop Helps Network Engineers Improve Network Performance for Big Data

“We are here to enable science. Big science, with big data, is affected by network failures.”

And with that, Brian Tierney, group leader of the ESnet Advanced Network Technologies Group at Lawrence Berkeley National Laboratory, kicked off a two-day workshop on perfSONAR, a tool for monitoring and measuring network performance from its starting point to its end point.

Tierney and Jason Zurawski, a science engagement engineer at ESnet, were the featured speakers at “perfSONAR Deployment Best Practices, Architecture, and Moving the Needle,” organized by ESnet and Internet2 and hosted Jan. 21-22 by the Ohio Academic Resources Network (OARnet) in Columbus, Ohio. The network engineers and developers who attended the event hailed from Brazil and across the U.S.; they represented higher education, regional optical networks and businesses that work extensively with researchers.

Many campuses and regional networks deploy sophisticated monitoring systems to ensure good network performance within their respective areas. However, scientists and researchers often collaborate with others across the country, or even across the world.

“Our networks are part of a very large global ecosystem of research and education systems worldwide,” Tierney said. “Because of so many components, it can be quite difficult to identify the source of a problem when one occurs. If you look at each network individually, it may look fine. Only when you look at the complete path can you see a performance issue.”

perfSONAR, which stands for “Performance focused Service Oriented Network monitoring Architecture,” enables these network engineers to build this needed, complete picture of a network from start to finish, or end-to-end, in network parlance, and monitor performance issues across these multiple domains. It is a joint effort between ESnet, Internet2, Indiana University and GEANT, the pan-European research network. The latest release of perfSONAR, version 3.4 released in October 2014, gives network engineers access to more data about network performance as well as increased security protections. Read the new documentation at: http://docs.perfsonar.net. In January 2014, perfSONAR reached a milestone with 1,000 instances of the diagnostic software installed on networking hosts around the U.S. and in 13 other countries.

The workshop, funded in part through a grant from the National Science Foundation’s Campus Cyberinfrastructure - Infrastructure, Innovation and Engineering Program (CC-IIE), focused on a number of operational aspects of deploying and using perfSONAR. The first day included a tutorial on how to install, configure, and use perfSONAR and the perfSONAR dashboard. The second day included presentations from those currently running large perfSONAR test meshes, with a focus on best practices and future growth. At the workshop, Tierney gave an introduction to perfSONAR and presented a session on debugging using the software. Zurawski discussed maintaining a perfSONAR node, described some user case studies and success stories, and provided an overview of perfSONAR as a Regional Asset.

“By providing community education and outreach, collectively we are all better prepared to serve the cyberinfrastructure needs of big data research,” said Paul Schopis, chief technology officer, OARnet, and principal investigator on the project. “Our goal is to provide opportunities for knowledge exchange and best practices. Successful cyberinfrastructure networks depend on a well-trained, collaborative ‘human network,’ combined with widely deployed and easily accessible monitoring and measurement infrastructure, such as perfSONAR.”