Status Report for SLAC/DOE IEPM Project
For October & November 1997
Very minimal administration support, apart from writing this report, was required this month.
Warren & Les attended the 1 day XIWT/IPWT (Cross Industry Working Team / Internet Performance Working Team) meeting at Intel on November 13th. There were about 20 attendees from both computer and communications companies. Les took rough notes for the meeting which are available at http://www.slac.stanford.edu/grp/scs/trip/xiwt-nov97.htm. About 6 XIWT collection sites are now running the PingER tools, and they hope to have about 15 sites by early next year. CNRI will try and get an archive site running by the New Year. So far the effort required from the PingER support team to support these sites has been minimal. The XIWT people have ported the tools to BSDI and HP-UX. They were already running on AIX, SunOS and Solaris.
At the XIWT meeting we had discussions with Tracy Monk of CAIDA/NLANR about using their MapNet tools. As part of this we arranged for Warren to visit NLANR in December to coordinate his work on extending MapNet to MapPing which will provide visualization of our PingER measurements. Also agreed to coordinate with NLANR on getting Latitude/Longitude measurements for monitoring sites.
Coordinated with Guy Almes & Vern Paxson to set some guidelines for installation of a Surveyor at SLAC. The intent is to eventually integrate Almes' Surveyor with a NIMI (National Internet Measurement Infrastructure) platform being developed by LBNL & PSC. The next step is for Almes and Paxson to talk at the IETF in December.
Helped Carleton University in Ottawa install the PingER tools and to set up a reverse traceroute server.
Worked with University of Cincinnati and BABar folks to understand UCinn's poor connectivity to ESnet. BaBar is very interested in using UCinn as an HP platform reference site for their program builds. Used the PingER measurements to show the extent of the problem and informed ESnet, and the ESSC representatives.
Provided Internet performance information for the SLAC Scientific Policy Committee meeting and also for the BaBar collaboration meeting at SLAC.
Warren amended the code to read in the median value that we now have running as part of time ping, and added the median value to the table. He also amended the formatting of the table and the slope just appears as an integer value. In addition we soft-coded some of the variables, which allows running the same program as a development version under cgi-wrap as the production version under cgi-bin.
Warren wrote a brief introductions to each file, and amended the configuration file for pingtrends-tabular-report (was Ping Alert) in order to display the correct header in full. In addition Warren amended the code to allow sorting negative numbers and to color code the numbers depending on their magnitude as well as absolute value. Warren also installed the latest version of the library cgi-lib.pl.
The new version of timeping that calculates the median value for the 10 pings and adds it to an extra field in the table, has been successfully running at SLAC since the end of October. We never received any feedback from Intel about specifics for running under BSD, although they did send out some general tips about getting it running which have been installed for the Linux version
Grouping the data
It was suggested that allowing special interest groups would benefit from having a pre-defined set of pairs to review performance. Warren wrote code to allow groups to be defined, and groups can contain other groups, which is passed to HEPNRC graphing routine. It has been agreed that this routine will be best if combined into the HEPNRC code so the groups can be displayed on the front page, and the user gets the benefit of the date range functions.
Warren has been working on adapting CAIDA's MapNet Java Applet to display the ping data SLAC and the other collection sites have gathered. We have a working demo and are currently developing original code for our requirements. This project also requires pin-pointing the geographical location of the server being pinged, so we have spent quite some time using either online tools or a map to find the latitude and longitude of the nodes. This information will be shared with CAIDA who are building a database.
Long Term Analysis
Warren is beginning to develop some code to analyze and present long-term summary files for the data that has been collected.
180 Day Report Plots
Following a series of email discussions on the ICFA and XIWT mailing lists on predicting trends, Les & Lois White added spline fits to the 180 day report plots (see for example http://www.slac.stanford.edu/comp/net/mon/offsite-timing-trends/180Days-slac.stanford.edu__cern.ch.ps) to aid the eye in spotting trends. Also Lois worked with Dave Martin of HEPNRC to finish porting the SAS programs that generate the 180 day report plots so they would work with the new data format which allows multiple collection sites. The next step is to install these SAS programs at the HEPNRC archive site.Back to Top
Revised 23 November 1997
Comments to email@example.com