Contact: Les Cottrell, SLAC
We have added Web100 to the measurement toolkit, and have selected an initial set of Web100 parameters to record. We are working on comparing results from Web100 throughput reports with those from the applications and from passive (NetFlow) measurements. We have also looked at the impact of high throughput on smoothed round-trip reported by Web100 and reported on this in a couple of presentations.
We have extended the prediction tools to incorporate exponentially weighted moving averages. It appears to make little, if any improvement compared to an un-weighted moving average.
We have been measuring and analyzing disk system performance to try and understand the effects of file system, caching, and committing writes, see Disk Throughputs.
Working with Richard Hughes-Jones, we have selected reasonable parameters for, and incorporated UDPmon into a test version of the toolkit. We are also separately making measurements with GridFTP, though incorporating into the toolkit requires more work on certificates. We are working on comparing the results from these two sensors with iperf, bbcp, and bbftp
We are working on extending the InCITE topology/tomography tools to make them more portable (remove MatLab). To assist in this we have a PhD student from Rice visiting SLAC for the summer. Jiri is working to provide improved topology/tomography visualization tools and incorporating them with the traceping measurements (see Internet Tomography). We hope this will provide improved ways to visualize the network links on which the Grid is built. Warren also modified traceping to increase its robustness.
We have incorporated modifications to PingER from NASA and EDG into the standard distribution.
We successfully ported the measurement toolkit from Solaris to Linux. The main challenges were related to the different ways the two operating systems treat multiple tasks in Perl. Following this we worked with Richard Hughes-Jones, of the European Data Grid (EDG) who visited SLAC for a week, to port the toolkit and start the measurements on a host at Manchester University. As part of this we parameterized the features that were measurement-host dependent and created a configuration file for each monitoring host to store the parameters. We also carefully documented the steps needed and built tools to enable automating the port procedures. Following this we successfully used the nascent tools to port and bring up the toolkit to FNAL, working with Frank Nagy of FNAL. An extra challenge at FNAL was successfully integrating the toolkit to work with FNAL's strong authentication. FNAL is now making measurements to a single host, extracting the data, analyzing and reporting the results etc. via the web. The next steps are to further improve the porting tools and the toolkit, automatically upgrade the toolkits at FNAL & Manchester, and work with Manchester and FNAL to set up sets of remotely monitored hosts for each site.
We have added Manchester University, NASA to the remote sites monitored by IEPM-BW from SLAC. For Manchester we measured over 460 Mbits/s, see Bulk Throughput - Manchester University.
We have installed and brought up Globus MDS and are working to install the PingER data. We are working on a web interface to provide access to the XIWT/CNRI/Telcordia/IPEX passive monitoring data.
Andy Hanushevsky and Les Cottrell submitted a paper on data compression, and Les and Connie submitted a paper on the new IEPM-BW measurement infrastructure.
Les visited Romania at the invitation of the Romanian Ministry of Telecommunications and Information Technology and the Ministry og Education of Research. He was part of a team from the west (including Harvey Newman, Ian Foster and David Williams of CERN) to provide assistance for "The Romanian Potential regarding Grid Activities and Distributed Computing". Presented talk on Worldwide Network Performance and Monitoring. On the way to & from Romania, Les met with Peter Clarke of UCL and Richard Hughes-Jones of Manchester to go over EDG activities and see how best to coordinate our efforts (see trip report).
We are putting together a draft of the Maggie ( Measurement and Analysis for the Global Grid and Internet End-to-end Performance ) proposal. The initial deadline has been postponed and the proposal will be submitted in August for consideration for FY03.
The following talks were presented:
q Overview of IEPM-BW - Bandwidth Testing of Bulk Data Transfer Tools, presented by Connie Logg at the May 2002 Internet2 Conference in Arlington, VA.
q Comparing Throughputs as Measured by Active and Passive Methods, presented by Connie Logg at the May 2002 Internet2 Conference in Arlington, VA.
q International Connectivity and Performance for the HENP Community, presented by Connie Logg at the May 2002 Internet2 Conference in Arlington, VA.
q INCITE Review, presented by Les Cottrell at INCITE Review at Santa Fe, May 2002.
q Experiences and results from implementing the QBone Scavenger, presented by Les Cottrell at the CENIC meeting May '02 in San Diego.