Warren gave a presentation titled
Configuration, Results & Future Plans
at the 11/5/98 XIWT IPERF group meeting at Reston. He reported that storage
requirements can be reduced by more than 85% by applying "gzip" to raw data text
files. Warren suggested that a short term approach for the central archive site
would be to use his perl scripts instead of purchasing a vendor statistical package.
This was agreed to. Warren also presented some preliminary analysis of the XIWT
Les reorganized how we (SLAC) interface with NetPredict. Coordinated with the XIWT and
NetPredict to arrange a netPredict presentation at the next XIWT meeting in February
in Menlo Park.
Les worked with Colorado State, BaBar and ESnet people to understand the abysmal
performance between SLAC and Colorado State. as a result BaBar decided to place a
new machine for code development at the University of Colorado instead of Colorado
Les discussed the desirability and feasability of producing a PingER brochure with SLAC
publications and DOE. We are currently awaiting guidance from DOE on some policy
Les reviewed a new tool called jping developed by somebody at the Unioversity of
Cologne. Exchanged several emails with the author.
Les exchanged some email with Tobagi of Stanford University
concerning response time required for
Les met with 3 people from Roche Pharmeceuticals (2 from Palo Alto, the other from
Switzerland) who were interested in the PingER system with the possibility of using
it within Roche.
Warren has a new setup for the pingtable program, a different setup for the
configuration file and a new grouping file.
Warren implemented a statistics module for perl rather than calculating the values
in the code. He also added a percentile calculation to the module and the author
says he will include it in the next release.
Future of PingER
Warren has been looking at databases as
part of possibly removing the dependence on SAS of the HEPNRC central archive site.
This is also relevant to the XIWT archive site futures.
Warren discovered a subtle bug in the monitoring software. The upshot is on rare
occasions a returned packet gets labelled to the wrong site. Warren
is working on a
new version which should avoid this, and will encourage monitoring sites to upgrade
when it is ready.
User Front End
Warren talked to Inxsight about
Hyperbolic Tree. However, it currently appears insufficient for our needs.
Warren made more analysis of the Surveyor data and processed October's data. He is
looking at tools to automatically generate the graphs, rather than using Excel.
Warren also obtained data from the XIWT Monitoring effort. They are using the
SLAC code at eight Monitoring sites and archiving the data in a SAS database
at NIST. Warren copied the SAS files from NIST to FERMILAB and dumped
the SAS data to text.
Warren extended the program to create images of the traceroute paths between sites
Routes from Several HEP Monitoring Sites to CERN.
Warren installed Mesa (a free version of OpenGL) and the perl OpenGL module to try
and extend the network snapshots to include several end nodes and do some work on
visualizing the results of the monitoring.
Back to Top
Revised 25 November 1998