Minutes from MMN Meeting of 12/11/98, in Lancaster Attendees Jon Crowcroft UCL J.Crowcroft@cs.ucl.ac.uk David Hutchison LANCS dh@comp.lancs.ac.uk Dan Waddington LANCS dan@comp.lancs.ac.uk Mark Banfield LANCS banfield@comp.lancs.ac.uk Chris Edwards LANCS ce@comp.lancs.ac.uk Jason Knight LANCS j.knight@lancaster.ac.uk Brian Cowe Cam Brian.Cowe@cl.cam.ac.uk Mark Sanford Loughborough j.m.sanford@lboro.ac.uk David Parish Loughborough d.j.parish@lboro.ac.uk Nat Pryce IC np2@doc.ic.ac.uk Frank Ball OBU frank@dsrg.brookes.ac.uk Ian Marshall BT marshall@drake.british-telecom.co.uk Paul McKee BT paul.mckee@bt-sys.bt.co.uk Hamid Gharib BT hamid..gharib@bt.com Apologies from: Morris Sloman mss@doc.imperial.ac.uk (flu) James Cowan J.Cowan@cs.ucl.ac.uk (business) Iain Phillips I.W.Phillips@loughborough.ac.uk (children) Agenda agreed Introduction: UCL - Jon Presented work on latest server stack management that James has embarked on IC - Nat talked about the progress on the MIDAS and other IC contributions Cambridge - Brian talked about the switchlet progress including the novel ability now to download new behaviours as tcl scripts - also about dividers. There was some discussion about the divider access to hardware/switches - GSMP II is preferred but quite a few real switches dont run this, although lots implement SNMP; however, normal standard SNMP doesnt give an efficient way to packetize the type of requests that are needed to do divider setups, and many SNMP agents are slow. Hamid noted though that France Telecom and Sprint are doing large trial networks with switches with GSMP. Jon suggested that cambridge write a draft RFC proposing a mapping of the functionality they need into an enhanced SNMP - this might make more headway than trying to get GSMP accepted. Ian suggested that an informational model should be written up too, as this could be independnet of the protocol wars! ACTION on Cambridge to write this up. OBU - Frank outlined the modeling/theory work on queuing that they have been doing. Lancaster - Dan talked about the PVM (Protected Virtual Machine) work he has been doing and gave an excellent demonstration of the Nt implementaion of the scheduler. Jon asked about the API to the "guaranteed" part of the scheduler, and whether it could be parameterised similarly to the packet schedulers (e.g. SCR+PCR+burst). Dan said that this was future work... Loughboropugh - David talked about adding intelligence to the AIR monitoring system they have built. Mark presented more thoughts on classification. There was a lively debate about what would be possible (concerning how well for example large delays could be classified as anomolous and associated more with rare high delays that are just part of a large deviation in a self-similar traffic ensemble. Mark and David stressed that they are not triggering automatic actions because of such classifications, and rather, just giving hints (possible additional evidence) for the network managers, and that practical work with the BT operators confirmed the validity/usefulness of their data processing. BT - Paul gave a talk about BT's future research directions in distributed systems which had come out of a recent brainstorming session - protocol oriented and "post CORBA" thinking were the order of the day, including idieas for self organisation and auto configuration of mobile agents etc etc..... Papers, Patents, Projects. We talked a bit about Framework V, and about patent progress. David asked if someone had picked up the ball at BT with the Loughbortough patent now that Tim SPence was no longer on the project. ACTION on Ian Marshall to check this. Integration Cambridge and Loghborough are in the throes of integrateing their work now. IC and UCL are too....things are busy. AOB None T&D Next Meeting Cambridge. Computer Lab, 11th March 1999