TobyRodwell points out that the GEANT IPPM network monitoring system repeatedly reports large OWD values for sequences of test packets between particular sets of IPPM systems compared to the median value, for example on March 3 2005

http://www.win-labor.dfn.de/cgi-bin/geant/auswahl.pl...

ChrisWelti suspects that the effect might be caused by the IPPM systems themselves 03 Mar2005. TobyRodwell suggests to categorize the issues according to

• IPPM software. Considered unlikely because the relevant part is identical to that of the DFN IPPM system and such an effect is not observed there.
• OS/hardware. Because packets are timestamped in user space, there could be considerable delays due to buffering on the NIC/in the kernel or due to kernel/user-space interaction.
• Network.

Issue Tracker Reference

Progress of this issue is tracked as Issue #8 on the Pilot PERT Roundup Tracker (PRI):

http://issues.pert.geant2.net:8080/pert/issue8

-- AlexGall - 04 Mar 2005