Status Report for SLAC/DOE IEPM Project

For April & May 1998

SLAC Home Page


Les was invited to be the "Discussant" for one of the 2 sessions on Statistical Aspects of Measuring the Interface at the  Interface98 the 30th Symposium on the Interface: Computing Science and Statistics  held on May13-16, 1998 in Minneapolis. He reviewed the papers from Vern Paxson, John Quarterman and Walter Willinger that were presented at the session. Les' presentation including a summary of the session and some Internet measurement challenges for statisticians, can be found at At the symposium Les met with Valerii Federov and Steve Batsell of ORNL to discuss closer cooperation with ORNL to look at optimal choice of monitoring and remote sites using statistical experiment design. ORNL havebeen funded to work with us on this. We agreed to ramp up on this in July. They will provide the statistical analysis and programs to do the experiment design or to evaluate the existing design and make suggestions. We will provide the data together with explanations of what it means and work with them. We have sent them information on our monitoring sites and on the beacon sites and a pointer to the May98 ICFA NTF WG on Monitoring report. We are also reviewing their paper (98 pages of pretty heavy math, so it is slow going), and we have sent them comments on what we have plowed through so far.

Discussed possible monograph on Internet Monitoring and Statistics with Valerii Fedorov of ORNL. He has been talking with Springer and they seem  interested. Valerii would act as editor and a major contributor, Vern Paxson and Walter Wllinger will also be approached.

Les attended the ESnet Review at Berkeley and presented a talk on ESnet Internet End-to-end Monitoring see The review committee seemed to be pleased and recommended a continuation of the monitoring activities. At the review, a reviewer from the University of Kentucky was interested in monitoring states that have poor connectivity. He gave us a list of such states (e.g. VT, ND, ME) and we sent him results from 3 of the states for which we have monitoring results.

We helped Walter Toki (a co-speaker at the ESnet Review) to understand the the performance of HENP sites using the Internet.

We wrote a report for the ICFA-NTF Working Group on the Internet end-to-end monitoring. It is available at It was sent out for review.

Reviewed 2 SBIRs on Internet Measurements and sent in responses to DOE.


We have a new monitoring site in Taiwan, unfortunately they are only sending 56byte packets, so their data does not appear in any of our reports.

Grouping the Data

Various groups have been added into the pingtable display including TEN-34, vBNS, and XIWT. A new concept called "Beacon Sites" has been introduced and the requirements for being a Beacon sites has been defined. 50 sites have been identified ( see http://www/xorg/iepm/pinger/beacon.html for definition of beacon sites) with a lot of feedback from the monitoring sites. The list has been distributed to the monitoring sites, and work is in progress on getting the monitoring sites to include them. An option in has been added to show only the beacon sites. Documentation on the Beacon sites as well as a table of all the monitoring and remote sites has been provided. A map showing the geographical location of the beacon sites has been made (see )

NIMI and Surveyor

Had several email and face to face conversations with Vern Paxson of LBNL to: understand the brittleness of TCP connections and on modeling TCP data-flows and correlations between packet loss and throughput and between response time variability and load; understand more about TCP retry timeouts; discuss access to the NIMI data; discuss how we might incorportae PingER into the NIMI.

Les talked with Guy Almes about the Surveyor project to get information on the how to access the data, and to clarify when the Surveyor will arrive at SLAC.

There were some problems with the NIMI probe, Warren worked with the guys from PSC to resolve it. Warren recompiled the FreeBSD kernel and eventually the problem was resolved. We still don't understand why it didn't work the first time.


We've updated the traceroute server code to allow easy addition of a non standard traceroute (e.g. the NIKEF traceroute), to allow traceroutes to other hosts besides the browser client, and to make it reasonably secure against denial of serice attacks.

Back to Top

Revised 22 May 1998
Comments to