Contact: Les Cottrell, SLAC
We have created a web site organized to provide easy access to all aspects of this project. Several new hosts have been added to those monitored by the Internet End-to-end Performance Monitoring – Bandwidth Project (IEPM-BW). We have added pages on documenting procedures for setting up new remote hosts. We added access from the IEPM-BW data to the PingER reporting web site.
We have built tools to facilitate and automate the infrastructure management. This includes downloading of code, gathering the remote configurations parameters (OS, cpu speed, code versions, file locations), verifying windows and streams are set correctly. To assist in auditing links that are having problems we put together a reporting tool to identify problems. We review this on a daily basis to identify where we need to focus effort to keep things running. As a result of this, we made several modifications to the measurement tools to improve robustness and recognize more failure modes.
We have defined and added several new tools to the IEPM-BW suite, including: a simple forecasting tool; We have been working to install the Globus infrastructure at SLAC so we can add GridFTP to the IEPM-BW tool suite. We have made some stand-alone measurements, but are still trying to resolve the certificates issues. We now have certificates for the DoE-SG and Globus and are working on enabling them. We developed tools to measure the throughputs passively using the Cisco NetFlow tools. This enables us to make forecasts for sites to which we are not making passive measurements, and also to compare and validate the active versus the passive measurements.
We have installed the latest Web100 instrumented TCP stack on two Linux 2.4 measurement hosts at SLAC. We are starting to use Web100 to understand the throughput performance.
We have made public the IEPM-BW data and provided documentation on how to access it. The data has been requested by CAIDA, ANL, and the Network Weather Service (NWS). BNL are working on providing a more standard way to access the data.
We have provided assistance to 3 major BaBar collaborator sites (RAL, IN2P3 & INFN/Rome) that were experiencing network problems. We documented the cases so multiple people could work on the problems and to share the information as an instructional tool. To assist others, including PPDG folks, in tackling network related problems we put together a web site of case studies related to poor network performance.
We have made presentations on the new measurement infrastructure and results at several meetings including: the DARPA PI meeting in Washington (presentation), the SciDAC meeting in Washington (presentation); the Internet 2 End-to-end Performance Initiative working group at Tempe Arizona (presentation); the Global grid Forum in Toronto (presentation on QBSS); the Internet 2 HENP networking working group in Toronto (presentation on network measurement); the IEPM-PPDG collaboration meeting in Toronto (presentation); visiting DoE program manager at SLAC (presentation on IEPM-BW, presentation on INCITE work); the International Committee on Future Accelerators Standing committee on Inter-regional Connectivity (ICFA/SCIC) at CERN (presentation); at the International Pure and Applied Mathematics (IPAM) workshop on Internet measurements at UCLA (presentation).
We have had several meetings with various collaborators including 2 phone meetings with the PPDG monitoring group; 2 with the Internet 2 End to end Performance Initiative group; a face to face meeting at SLAC with the LBL team; a phone meeting with Rice; face to face meetings with member of the European DataGrid at CERN and Daresbury Lab; face to face meeting with the Rice/INCITE collaborators and the Delaware/Pathload/Pathrate collaborators.
We submitted a paper to the IEEE on the high performance network measurement results and a second on the effects of compression on throughput to the Edinburgh Global grid Forum meeting in July.
We are in the process of porting the measurement tools to Linux from Solaris. We are also understanding Web100, deciding what parameters are useful to add to the IEPM-BW measurements, and whether and how Web100 can be used to assist application steering
We have two sites