Status Report for SLAC/DOE IEPM Project
For August & September, 1999
PingERWarren added a new metric to the PingER reports. The TCP throughput based on a theoretical equation linking the RTT and packet loss by Mahdavi et al. Agreement appears to be quite good except for hourly ticks where there is insufficient sampling to provde accuarate input for the equation. We will look at how to improve the accuracy for good links.
Warren fixed some bugs in table.pl and re-enabled the links to other other reports.
Warren and Kishan worked on a syn-ack program to mimic ping but with TCP packets. Kishan developed the code and experiments are being conducted to look at any effects based on ports or load. Caution must be used when running this program as it may appear to a remote site that it is an attack.
Pathologies/ProblemsResolved the following pathologies: big differences in performance to two major sites in Hungary. Worked successfully with Colorado State (Walter Toki) and ESnet to improve performance beween ESnet and colostate. A user of PingER spotted that performance between SLAC & LBNL was badly degraded after the ESnet upgrade, we reported it to ESnet who quickly found the source of the problem and fixed it which improved performance by a factor of 2. Working with India to understand a difference in the performance between Bombay and SLAC/Stanford (SLAC is acceptable while Stanford is bad). We worked with ILAN to allow pings from the Stanford University monitoring site. Helped a user understand the performance between DESY/Zeuthen and SLAC. Trying to contact John MacAllister to ascertain status of traceping. Also trying to contact Christina Vistoli to correct the links to ping_data_plot.pl.
New Sites and New GroupsFour new PingER Monitoring sites came on-line this month: At ANSP at San Paolo in Brazil; At KAIST/APAN in South Korea; At the ESNET-NOC at LBNL; and at ITEP in Moscow, Russia. Les and Warren spent considerable time assisting the sites and getting the documentation updated to incorporate the sites.
In addition Warren also received an inquiry from a site in Canada interested in using the PingER tools.
A new group, SF-ISP, was added to the reports. This group contains several ISP's in the Bay area. Les analyzed the data to show that the more expensive of two DSL ISPs provides a higher performing service.
Still working with sites to get them to install the new Beacon sites. The current status is that, a month after announcing, 15 of the expected 22 sites have installed.
Analysis/documentationLes & Charley Granieri are using the Surveyor data to try and get some measurements of error free seconds and outage length distributions.
Les found and added several sites in E. Europe to help in understanding connectivity to that region. This is partly in response to the upcoming presentation at the NATO Advanced Networking workshop in Tbilisi.
Les put together a web page on some of the uses PingER data has been put to, see Examples of the use of PingER
WLes & Warren had a phone meeting with Lidinsky and He of HEPNRC.
VoIPWarren and Kishan continued to work on monitoring the test VoIP link. Warren enabled SNMP on one of the routers to monitor the loading of the line for use in TOS tests. Kishan found an unexpected pattern in the packet jitter which will get further investigation.
RIPE-TTAfter a long wait and several technical hitches, the SLAC box for the RIPE test Traffic monitoring project became fully operational. The project co-ordinators report the installation is one of the best in the project and the GPS signal is very strong. Analysis will be started and worked on by one of the new RAs.
GuthrieWarren and Yvonne continued to work on the Guthrie node and group management program. Yvonne developed a GUI interface using the Java Swing classes.
NGI-TestbedWarren attended 2 conference calls for the new ngi-testbed monitoring working group. The group would like to experiment with providing the monitoring data via a LDAP server. In response Warren worked with purchasing, operations and systems to get a new machine installed. Warren is now working on securing the system and designing the ldap configuration and schema.
Missing dataWe had the usual glitches, this month there was missing data from KEK, Stanford and TRIUMF. Also we lost data from BNL due to hurricane Floyd. And Sinica due to the earthquake in Taiwan.
AdministrationWarren and Les interviewed 5 students and hired 2 of them as new RAs.
Warren ordered a new PC and arranged for it to be configured ready for use by one of the new RAs.
Warren worked with a vendor and purchasing to obtain an OC3MON device (specialised hardware for use with real time network monitoring).
Back to Top
PapersWarren and Les worked on a paper to be submitted to the IEEE Communications Magazine on Network Traffic Measurements and Experiments.
Warren prepared a first draft of an abstract for a paper to be submitted to the Passive and Active Measurements Conference. Final Abstracts must be submitted by November 15.
Warren attended the SIGCOMM conference. There were many interesting network research papers presented and Warren will follow up with several authors to apply their results to PingER.
Warren, Les and Bill Lidinsky prepared an abstract for the Computing in High Energy Physics conference to be held in February 2000.
Warren worked with Ira Richer from the XIWT/IPERF group to produce a network map for the forthcoming XIWT white paper.
Les attended the ESSC meeting in Santa Fe. He gave talks on ESnet/HENP monitoring and performance between ESnet and US universities, ICFA-SCIC and international performance and NGI monitoing. Made the talks available on the Web (see Talks and Papers on Network Monitoring.)
TalksWarren is preparing a talk for the HEPIX conference at SLAC in October. Les is preparing a talk on QoS measurement for the NATO Advanced Network workshop in Tbilisi in October, and also arranging to visit Daresbury and RAL in the UK on the way back.
Revised 23 September 1999
Comments to firstname.lastname@example.org