perfSONAR-PS v0.9 Release

Internet2, together with its collaborators, announced the newest release of perfSONAR-PS, a complementary set of network performance services developed under the umbrella of the global perfSONAR network performance measurement framework. Implemented in the Perl programming language, perfSONAR-PS enables network operators and engineers that already leverage a Perl-based environment to seamlessly integrate comprehensive performance measurement technology into their existing network management and measurement systems while still maintaining interoperability with other standards-based solutions.

 

As a set of software services that implement the perfSONAR network monitoring protocol, perfSONAR-PS provides users a window into the network to generate near real -time performance traffic monitoring and visualization.  Its ability to be utilized on both IP and optical networks as well as hybrid networks like the Internet2 Network or the ESnet Science Data Network, means that monitoring and measurement information can span and aggregate information from different network architectures and debug potential issues across those separate administrative domains even if those domains are autonomous measurement systems. Its real-time, global analysis of network performance problems also makes it possible for users to make immediate adjustments to their applications during run time.

"The continued proliferation of new hybrid packet and optical networks, which are today critical for important global scientific applications, greatly depends upon their ability to maintain the highest levels of end to end performance," said Jeff Boote, Internet2 senior network software engineer. "By extending perfSONAR in this way, a much greater number of our regional network partners and universities will be able to take advantage of perfSONAR's unique capabilities to automatically identify and help resolve performance issues. And in doing so, we anticipate more rapid adoption of new dynamic and hybrid network technologies and architectures."

Designed to be dynamic by nature, perfSONAR includes a discovery service to allow new performance data to be found rapidly and is designed to automatically consider and represent the network topology on which it is operating. In the future, this means that network applications may be designed to interact with perfSONAR to automatically determine network performance and make optimization decisions based upon that expected performance. This capability is critical as Internet2 and other global research networks move toward dynamically-switched hybrid networks making it necessary to use a monitoring platform like perfSONAR that can operate as flexibly as the network can.

"By providing a unified suite of tools for monitoring intermediate and end points of the network, these services will enable monitoring and trouble-shooting of the complete end-to-end path including devices at end sites and along the path. This is a major step forward in our ability to understand the performance of networks," said Les Cottrell, assistant director at Stanford Linear Accelerator Center's Scientific Computing and Computing Services.

In the near term, perfSONAR-PS will provide immediate benefits to tier-2 universities and tier-1 laboratory sites participating in High Energy Physics experiments on the Large Hadron Collider (LHC), a new particle accelerator at CERN in Switzerland which will begin operation next year. Expected to produce roughly 15 Petabytes of data annually, LHC researchers at over 70 U.S. universities will need to download or transmit, about two terabytes of data for analysis within four hour windows, every few weeks. perfSONAR will enable these organizations to exchange near real time network monitoring data, allow for 24x7 troubleshooting of high performance optical   links, and provide advanced application network feedback.

"perfSONAR will prove to be extremely valuable to monitor and troubleshoot network paths for large data Grid environments such as the LHC," said Brian Tierney, staff scientist, Lawrence Berkeley National Laboratory. "perfSONAR can also be used to help with resource selection problems such as deciding which data repository should be used for a given application."

perfSONAR-PS is being released in a beta version to encourage continued testing and refinement of the code. A full production version is anticipated by the end of Q1 2008. The perfSONAR-PS software development is the result of a collaboration between ESnet, Fermi National Accelerator Laboratory, The  Georgia Institute of Technology, Indiana University, Stanford Linear Accelerator Center, University of Delaware, and Internet2.
For more information, visit: http://e2epi.internet2.edu/pS-PS/

spectrum