Status Report for SLAC/DOE IEPM Project
For September & October, 1999
- TimepingWarren amended timeping.pl (the code that sends the ping probes at the monitoring sites and records the results) to allow send and size directives to be given from the configuration file. This is to allow us to gain extra precision on the measurements to consistently well performing sites such as those on ESnet. To ensure backwards compatibility, if the directives are not given, the program continues the usual 11x100 byte pings followed by 10x1000byte pings.
Warren looked at changing the timeping.pl to first try a DNS lookup on the node name and only using the IP address (if one is provided in the list of sites file) if DNS fails rather than use the IP address only. This is because one particular node changed IP address but didn't tell anyone. And as many sites had the old IP address listed the node appeared unreachable for quite some time.
- GuthrieYvonne continued to devlelop the Guthrie Program for managing the details of the large number of nodes involved in the monitoring by using a local Oracle database. The program now provides a separate list of the countries and continents that have monitoring and remote sites. This is so if a new site is added then the web cgi program pingtable.pl can automatically add this site to the drop-down lists. Also, we are investigating the java Run-Time-Environment (RTE) to allow users to run the guthrie application locally rather than remotely using X.
- Development at HEPNRCThe PingER archive site at HEPNRC have a new person. Warren listed some current and future development that the new person may contribute to;
- a stable poisson pinger
- retaining all rtt (not just min, avg, max)
- more flexibility ( but backwards-compatable ) in the sent and size parameters
- new duration field to help with ULM format compliance
- installation or patch routines to ease upgrades
- better database connectivity.
New ProjectsSeveral new research assistants (RA's) joined the group this month. There are now five RA's working 20hrs/week and 1 part time person working 5hrs/week. All are Stanford students who come to SLAC between courses. With all this extra manpower we are in a position to move forward on a number of development and analysis projects.
- McentyreMit Shah is investigating the use of rate limiting, and how it might affect performance measurements in order to identify a signature. He is studying a known case of rate limiting, to sites in Israel, and working on the Cisco Committed Access Rate (CAR) technology.
- BrooksKaushik Mittra is investigating the Java Analysis Studio and the ROOT packages to allow easier and faster plotting of PingER reports. In particular Kaushik is applying the Servlet Howto to produce graphs of the HTML reports. Future development may involve providing complete files for the user to manipulate within the packages.
Kaushik Kaushik worked with one of the authors of JAS, Tony Johnson, to understand the requirements. Warren worked with Paul Raines to get access to the servlet directory on the test web server.
- WhitmanFaisal Zakaria Siddiqi is investigating better use of databases for the analysis of data, using perl and java database connectivity.
This project may also involve extracting data directly into tools to improve visualization, some of which may be listed at http://www.cybergeography.com/atlas/atlas.html.
Pathologies/ProblemsLarge day-of-week and hour-of-day delays from SLAC to certain XIWT/IPERF sites were noted by Jeremy Hylton. One hypothesis is that SLAC network traffic is dominated by user traffic, which dies down when users go home at the end of the day, while most of the other networks are dominated by customer usage which is spread more evenly over the day. However, SLAC's connection to ESnet is far from overloaded, and no similar effects are seen to other sites. Another hypothesis is that it is the exchange point at MAE-WEST that suffers the above effect.
Les talked with Vern Paxson on some pathological ping behavior.
Les worked with Chritina Vistoli of INFN/CNAF to fix problems accessing the ping_data_plot.pl code. Warren contacted the webmaster at INFN/Napoli to fix a broken reverse traceroute server.
Warren investigated some anomilies in data from the monitoring site at Daresbury where it seemed more packets were being received than were sent. Robin Tasker, the PingER administrator for Daresbury discovered the cause was a bug in ping on Red Hat 6.0. Warren worked with Robin Tasker to install the NIKHEF ping, which doesn't suffer the problem.
Les worked with our friends in Hungary to understand the pathological behavior of Internet performance in Hungary.
Les corresponded with someone at TIFR Bombay to understand why packet loss between ESnet and TIFR was so much better than other U.S. sites to India when the traceroute from SLAC and from TIFR shows the routes to be the same. It is possible there is some kind of shaping involved.
New Sites and New GroupsWarren worked with Shiqi He at HEPNRC to get the new monitoring sites added to the data archiving and plotting and to link into the participants page.
Les worked with Pat Krietz to provide details of the requirements for a possible new monitoring site at BINP in Novosibirsk. Warren responded to an enquiry from NASA about installing a monitoring site.
Warren worked with Dave Millsom and Teresa to create a new Subnet for the IEPM machines.
VoIPWarren and Kishan continued to try to understand the loading of the VoIP line with TTCP/UDP. Warren used SNMP on the router and used MRTG to plot the utilization. Warren worked with Becca Nitzan of ESnet to better understand what should happen. Becca reported our findings to Cisco, and in response Becca upgraded the IOS and the RSVP code. In addition there were further routing problems on the testbed.
Les met with Bjorn from NetPredict to discuss using/loaning a Smartbits traffic generator. Bjorn agreed to lend SLAC the machine and Warren worked with Rajiv Krishnamurthy to install a Smart bit box. Rajiv installed the control software and tests have begun.
Warren configured a new machine to sit on the VoIP network next to Nereus to spy on the traffic to/from Nereus. Warren created accounts for Kishan and Rajiv and worked with systems to allow them access to the new machine in case of problems.
Warren and Les worked with Becca Nitzan to Co-ordinate with other users of the testbed
Les looked at the time correlation for the SLAC - LBNL ping RTT data.
Les investigated the Matlab package and created a Fourier transform
IPv6Warren worked with Becca Nitzan at ESnet to trouble shoot some issues with the IPv6 set up. Eventually it was found that the solution was that the IOS needed upgrading. Warren worked with SLAC net-admins Tony and Jai-Young to do this.
Warren worked on porting the PingER tools to PingERv6, and porting IPv6 applications to perl using the XS tools.
RIPE-TTThe project co-ordinators for the RIPE Test traffic project installed a program to generate alarms if the delays break away from a long term average. SLAC has been involved in several alarms. Warren is investigating if this is due to route changes.
SurveyorLes and Charley worked on analysing the surveyor data to obtain details on loss patterns and defined metrics such as error-free-seconds.
Warren retrieved the data for September, and now also has an account at Advanced to retrieve the data rather than request it from the project co-ordinators each month.
NGI-TestbedWarren worked with operations and systems to install a new machine called Evagore to serve PingER data from LDAP. Warren created a schema and made the data available. If you have OpenLDAP, you can see a demo by typing
ldapsearch -L -h evagore.slac.stanford.edu -p 3900 -b "o=ngi-mon" "(objectclass=*)"
Missing dataThere were relatively few glitches this month. SLAC suffered several power cuts. The Monitoring site Stanford failed and the system had to be replaced.
AdministrationWarren and Les worked with the new R.A.'s to complete the required paperwork and determine the hours they would be at SLAC.
MeetingsWarren and Les met with Wade Hong during his visit to the HEPiX conference to discuss Canadian Internet connectivity and monitoring tools in use.
Warren and Bjorn met Prof David Cheriton from Stanford and discussed networks and monitoring tools. Prof Cheriton suggested we also talk to Prof Fouad Tobagi and will put us in contact with developers at Cisco.
Les arranged to meet with people at RAL and Daresbury While he will be in the U.K.
Les attended the PPDG meeting at SLAC. His notes are available at http://www.slac.stanford.edu/xorg/ngi/ppdg/ppdg-slac-sep99.html
Warren and Les attended a voice conference with XIWT/IPERF.
Back to Top
PapersWarren, Les and Bill Lidinsky worked on an abstract for CHEP2000 called How No Collisions Make Better Physics. Warren, Les and Becca Nitzan worked on another abstract for CHEP2000 called "1-800-CALL-H.E.P."
TalksWarren presented a talk to the HEPiX/HEPNT conference at SLAC.
Revised 22 October 1999
Comments to email@example.com