PSAMP meeting on Nov 18, 2002 reported by Juergen Quittek ======================== Mark Allman on IRTF IMRG ======================== Mark Allman announced the new IRTF Internet Measurement Research Group (IMRG) It is just starting and Mark briefly explained the goals. ========================= Andy Bierman on WG status ========================= Andy Bierman reported on the WG status. He asked whether or not the current list of documents to be produced represent the correct grouping of content. There was no objection to the current grouping. Then the schedule of the WG documents was discussed and Andy asked for volunteers for documents on which work had not started. The current charter is not clear about the document schedule: it does not clearly distinguish initial submission and completion of documents. Nick Duffield explained that initial submission was intended. The following schedules were suggested: Schedule for doc 3 (report formats): Feb03 / Jun03 Schedule for doc 4 (export): Feb03 / Jun03 Schedule for doc 5 (MIB): Mar03 / Aug03 Nick Duffield volunteered to submit the initial version of doc 3, Benoit Claise volunteered for doc 4, and Dan Romascanu volunteered for doc 5. Andy will update the agenda accordingly. ======================================= Nick Duffield on the framework document ======================================= Nick discussed the following framework components - configurability - - per device - - per measurement process - - per collector? - collector driven rate control - selection operations - - framework should discuss but not solve issues - do we want to support iTrace (ICMP traceback) type of applications? - issue on specifying sampling algorithms - - should specify functional requirements and not constrain implementations - sampling types: what can the application expect if 2 sample probes with the same algorithm There was very little discussion on this. Unknown speaker: co-operation with ITRACE WG? Randy Presuhn: Support of IP tracing applications The issue was delegated to the mailing list Juergen: Can sampling be specified precisely enough that interoperability is given? Tanja: Confidentce intervals can be given. ======================================== Tanja Zseby on packet selection document ======================================== Tanja gave an overview on her terminology which is not yet consistent with the framework document. It was agreed to move her terminology section to the framework document. Emile: Is time synchronization considered? It might be desirable for example for path discovery. Tanja: yes, but not explicitly Andy: Time synchronization is not in our scope Then Tanja discussed the information model for packet selection. The model was discussed concerning its completeness and openness. Dan: Concern about openness of information model: might be too complex, better not do it Nick: Concern about openness of information model: might be too restricted. There might come up future sampling techniques that cannot be expressed with this model. ================================ Juergen on Relationship to IPFIX ================================ Juergen suggested to use common terminology for both groups. Further suggestions: - use PSAMP packet selection model in IPFIX - use IPFIX export protocol for PSAMP packet export there might be problems with different reliability requirements. The higher IPFIX requirements might be too much for PSAMP - use PSAMP packet selection configuration for IPFIX meter configuration