Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.
Comment: Migrated to Confluence 4.0

...

Recent release of the Network Performance Toolkit

- Presented by Jason Zurawski, Internet2

Jason provided an overview of the pS-Performance Toolkit:

  • Historical Perspective
  • Performance Problems
  • What's in the Box?
  • What's in (outside) the Box?
  • Lookup and Discovery
  • Testing Parameters

See:

http://www.internet2.edu/presentations/fall09/20091005-PerfWG-zurawski.pdfFor discussion:
- Guidelines for scheduling measurement tests (previous wg discussion lead by Joe Metzger)

Discussion:

  • Sometimes there is a need to compromise between number of sites and amount of testing
  • Good idea to have identical hardware at all sites
  • Important to differentiate between problems at the site and problems on the network
  • Both scheduled testing and on-demand testing are needed

Geant's experiences with PERTs

- Presentation Presented by Victor Reijs, HEAnet

Victor provided an overview of The Performance Enhancement and Response Team (PERT) at Geant:

  • Why is e2e performance hard?
  • What is a PERT?
  • Brief history of the PERT
  • What is eduPERT?
  • GN2 PERT experience feedback
  • Challenges for eduPERT
  • Why you should have a PERT

See:

http://www.internet2.edu/presentations/fall09/20091005-performance-reijs.pdf

For discussionDiscussion: - US Regional and State Networks PERT
- Joining efforts with the GEANT distributed PERT

  • Is the PERT concept something we could implement here in the US?
  • Argument in 2001 was that "we don't have the tools."  But now we do have them.
  • It would be great to add this human network.
  • Tools can help us identify performance problems, but we need help in knowing how to isolate the source. What's the decision tree?
  • It was agreed there is lots of potential, and this could add to the value of the Internet2 network. Could help to build value for R&E networking.
  • Concerns: How are we going to fund such an effort?  If it is an extra service that's part of the network, maybe that could be sold. But what is the value a user will put on it?
  • Suggestion: identify network engineers at Tier-2 sites, who need to understand this troubleshooting anyhow to handle LHC requirements, and use them as a cadre from which a PERT could grow. Use a ticketing system and integrate it with Geant's ticketing system.