Report on IEPM PPDG Efforts for the Quarter January - March 2006

Report by Les. Cottrell for the IEPM team, SLAC

IEPM Status Reports

Bandwidth/Throughput Monitoring

As part of the Terapaths project we collaborated with ESnet to make measurements of jitter with OSCARS. The results showed that OSCARS worked well both to limit throughput and to reduce jitter.

Many improvments were made to the IEPM-BW output to add more scatter plots, to provide a zoom capability, extend the use of pathload and work with the author to add timeouts, to improve the detection of events using the Holt-Winters method.

At the requests of the SLAC accelerator operations folks we made a case study of an Oracle application's performance between SLAC and ORNL. The problem was resolved to be in the application.

Passive Monitoring

We put together a paper on the effectiveness of passive monitoring for forecasting for CHEP06. The results are promising but do not remove the need for active monitoring.

PingER and Developing Region Monitoring

Based on a paper by Bamba Gueye and others on locating Internet hosts, we are developing a web based Java host location tool called TULIP. This uses ping Round Trip Times (RTTs) from landmark sites to the host whose location is desired to provide triangulation information to determine its location. A problem with Gueye's method was its lack of landmarks outside Europe and N. America, so we developed a reverse ping server that is being deployed at IEPM PingER and BW sites to provide landmark facilities. Besides N. America and Europe we now have landmarks in Brazil, S. Africa, India, Pakistan, Taiwan, and Japan and are actively working to add others. In particular we are working with the NLANR/AMP measurement team to install the landmark code at critical AMP sites in New Zealand, Australia, China, Malaysia, Phillippines, Chile and N. Africa. Besides providing a general purpose tool to help locate hosts for Internet users, it will also be valuable to validate the locations of over 700 PingER sites around the world. Since we supposedly know the locations ofthe PingER sites. It will enable us to evaluate the accuracy of the method.

We made progress on automating PingER management by providing scripts to identify and tabulate non-responding hosts.

We finished up the yearly ICFA/SCIC review on network performance worldwide. We then used this as a basis for a detailed study of performance for S. Asia and Africa that was presented at CHEP06. To simplify producing the performance report figures we are developing a Java application to provide long term performance trend plots.

Testbeds

We completed a Memorandum of Understanding with Microsoft to evaluate and report on the performance of the new TCP stack that will be released with the next version of Microsoft Windows (Vista/Longhorn). Following this we installed Windows Vista and Longhorn and completed the first phase of the evaluation using production high-speed networks within the US and trans-Atlantic.

Admin, visits, papers, presentations, proposals etc.

The IEPM group hosted a successful two week vist by two staff members from the Pakistan National University of Sciences and Technology (NUST) Institute of Information Technology (NIIT). The visit was to discuss further collaboration and projects. As a result we put together projects to: Les Cottrell visited NUST/NIIT in Rawalpindi/Pakistan for a week to work with students and faculty and to present findings on Pakistani Academic and Research network performance to various senior ministers and leaders in Pakistan. We also got agreement from NTC and COMSATS to install PingER monitoring hosts at their PoPs in Pakistan.

We submitted the following proposals:

Publications: We made the following presentations: