Status Report for SLAC/DOE IEPM Project
For September 2001
Work completed or in progress since the last report.
GridWarren has been working with some of the members of the WP7 (Network Monitoring) group within the European DataGrid to learn about their efforts. Warren has registered PingER as a data producer in the NASA GMA test LDAP Directory server. Warren has also worked on making PingER data available from ldap://evagore.slac.stanford.edu:3900/o=Grid-Mon. SLAC has been allocated a unique OID to define Attribute Types and Object Classes.
High throughput/QBSS Connie has developed a script to standardize the collection and recording of iperf measurements at SLAC. Warren worked on reformatting the data for inclusion in the PingER reports. We are also collaborating with someone from the UCL London to integrate their iperf analysis that is based on PingER back into PingER. See http://pc35.hep.ucl.ac.uk/.
The DataGRID monitoring effort (known as work package 7) has used the PingER software and adapted it for use in iperf measurements. See WP7 Iperf Monitoring. presented at CHEP01.
Les worked with Andy Hanushevsky to see how bbcp's self pacing works.
Les worked with Paola Grosso to set up a 100Mbps QBSS test bed using two Cisco 6500s and made measurements on competing throughputs and response times. See http://www-iepm.slac.stanford.edu/monitoring/qbss/measure.html.
INCITEWarren has obtained the chirp code and Matlab Analysis routines. Both Rice and LANL are still in the process of setting up PingER hosts. Since the Internet 2 face-to-face meeting in Austin was replaced with a video conference, we cancelled our trips to Rice also.
PerfMapsWarren made some changes to the PerfMap display for use in the SC2001 exhibit. Warren is also developing a method to zoom in on performance for certain groups such as the PPDG.
PingtableDue to problems with java servlets at SLAC, we have abandoned the QueryHistogram plotting program and ported the functionality to a stable and maintainable perl program. To see the results go to http://www-iepm.slac.stanford.edu/cgi-wrap/pingtable.pl?from=SLAC&to=WORLD and click on one of the lines with SLAC in the 1st column.
We are also working with Maxim of FNAL to provide faster ways and improved formatting of the graphs produced at FNAL.
VoIPWarren is working to extend the VoIP pilot to include the sites involved in the Internet2 VoIP WG. Warren has also added sites in the Czech Republic and Brazil.
Web 100Doug installed the Web 100 patch 3 on Linux 7.1. He is debugging the web100 interface to java JNI program - this allows a java diagnostic program similar to what Tom Dunigan showed to be developed for any parallel stream product without modifying the source. This program is more complete than showing an ftp connection with xxx bytes transferred in xxx seconds. He is working on a test program that adds timestamp functionality to web100 api for tcpdump/web100 integration. This allows any combination of web100 and tcpdump functionality to be combined into a program to monitor performance history. There is no substitute for this capability currently. he is also adding graphics capability via Java swing API to plot results of web100/tcpdump data. This is not a tcpdump/tcptrace program, but a real time plot which will allow users of any tcp socket program to monitor performance, it wont matter whether they ust bbftp or bbcp. The data collection works in real time. The graphics program and timestamp input through the Java JNI API is being debugged.
- Imran's port/re-write of the traceping is proceeding. An inital display program can ve viewed at http://www-iepm.slac.stanford.edu/cgi-wrap/newtraceping.pl
Missing dataARM, INFN/CNAF, KAIST/APAN and the University of Melbourne have not been serving data for several months and have been removed from the analysis routines.
New SitesThere is a new PingER monitoring site at the San Diego Supercomputing Center (SDSC). We now have PingER monitoring at all PPDG sites.
Problem assistanceWe assisted the Technical University of Dresden im measuring the performance to SLAC. To do this Les got an account at TU-Dresden and installed iperf, tcpload etc. See http://www-iepm.slac.stanford.edu/monitoring/bulk/dresden.html
CENICLes, Paola and Warren wrote an article about working with QBSS, which will appear in the next issue oif the CENIC publication.
ESnetLes prepared a talk on ESnet End-to-end monitoring activities for the ESnet review. he attended a 3.5 hour phone meeting while in Beijing to go over the presentations for the ESnet review, attended the one day meeting by Video conference and presented the talk.
HENPLes attended the CHEP01 meeting in Beijing. He gave 4 talks. One prepared by Warren on IPv6 in ESnet. One prepared with Connie Logg on Passive performance monitoring & traffic characteristics at the SLAC border. One in collaboration with Caltech on Achieving high performance throughput in research networks. The last one was the Summary of the Local & Wide Area Networking parallel track. He also met with Chinese HEP networkers and talked to the local Sun Microsystems people and prepared a report on the computing and networking scene for HENP in China.
Internet 2The Internet2 members HENP networking WG meeting was Cancelled. Slides on Grid Monitoring that were to have been presented by Warren at the HENP BOF are now online. Les spent some time working on a presentation on high throughput for the Internet 2 Working Group on HENP networking. Unfortunately the meeting was postponed.
Les prepared and presented a talk on QBSS Applications at the Virtual Internet2 Member Meeting Working Group on QoS.
SC2001 Les wrote and submitted a SC2001 Bandwidth Challenge Proposal for the SLAC/FNAL SC2001 collaboration. Les also worked on getting about 24 other collaborators, accounts and access to hosts at their sites and and configuring hosts.
Created 4 October 2001.
Comments to email@example.com