Report on IEPM PPDG efforts for PPDG for the quarter October - December 2004

Report prepared by Les Cottrell, January, 2005

Collaboration with IEPM, Network Performance Monitoring

Bandwidth/Throughput measurement (IEPM-BW)

We submitted and had accepted an extended abstract on Transport Fairness Characterization and Evaluation - comparison of variants of TCP and UDP based transport with respect to transient and steady state traffic on real networks for the Protocols for Fast Long-Distance networks in Lyon.

Lightweight Bandwidth Estimation

We cleaned up the ABwE web site that describes the lightweight avaliable bandwidth estimation toolkit, and put into production an ABwE server at SLAC. The server is now installed at about 40 sites.

Bandwidth performance anomalous events

We completed implementation, tuning, and checking out of the modified NLANR "plateau" algorithm. We built a library of interesting events as well as a canonical data set of time series bandwidth measurements from SLAC to 40 sites for 100 days. This will be used for comparing the effectiveness of the Plateau versus other algorithms such as Kolmogorov-Smirnov, Holt-Winters and Principal Component Analysis. We also set up a collaboration between Loughborough University (K-S), FNAL (H-W), NIIT/Pakistan (PCA) and SLAC (Plateau and leadership) for implemnting and comparing the effectiveness/applicability of the various algorithms against a common data set. We put tother a draft paper describing this work.

Traceroute Analysis and Visualization

PingER

We worked with Florida International University and CalTech to install a PingER monitoring site at FIU. This will be particularly used for monitoring sites on the Americas Path Network (AMPATH).
Working with an Institute in Bangalore, India we set up a PingER monitoring site there. This will be valuable to see the performance of paths from and within the developing world.
Hosted visit and talk by geographer from UNE Armitage Australia, who has been using the PingER data to detect a wave of congestion that moves around the world with daytime. This was used very succesfully as a demo for SC2004.

SC2004 Bandwidth Challenge

We co-led (with CalTech) a collaboration with Caltech, FNAL, University of Manchester, England, several companies (e.g. Chelsio, S2io, Sun), ESnet, National Lambda Rail and others to participate in this year's SC2004 Bandwidth Challenge. We secured the loan of over $400K of routers/switches/servers/10GE NICs plus two 10Gbits/s links from Sunnyvale to Pittsburgh, and space at the QWest and Level(3) colocation facilities in Sunnyvale. We successfully defended our Bandwidth Challenge for sustained throughput achieving over 100Gbits. This was reported in the national news including Yahoo, CCNews, SlashDot as well as Internationally. For more details see the web site at http://www-iepm.slac.stanford.edu/monitoring/bulk/sc2004/hiperf.html.

Proposals and Representation etc.

SLAC is a co-PI on the US Department of State and the Pakistan Ministry of Science and Technology funded project for Measurement and Analysis for the Global Grid and Internet End-to-end performance (MAGGIE). We have regular fortnightly phone meetings We will exchanging visits in the near future to forward the collaboration.

SLAC is a partner in the TeraPaths: A QoS Enabled Collaborative Data Sharing Infrastructure for Peta-scale Computing Research. A two page summary of the new proposal can eb found at http://www.slac.stanford.edu/grp/scs/net/proposals/iepm-bw/dgnmi-2p.doc.

SLAC is a partner in the UltraLight optical testbed proposal (led by Caltech) which was funded by the NSF. We attended the kick-off UltraLight meeting at CalTech in December 2004.

The proposal with Texas A&M, NASA and others to NASA on developing and monitoring IP based protocols for NASA satellites etc. was rejected. Submitted paper with FNAL to CHEP04 on Wide Area Networking System for HEP Experiments at FNAL.

We served on the Program Committees of the Protocols for Fast Long Distance Networks (PFLDnet 2005) workshop to be held in Lyon in February 2005, and the Passive and Active Measurements workshop (PAM2005) in Boston in 2005. We reviewed about 20 papers and made recommendations.