At the end of October I spent a few days in Paris, attending the CNSM 2011 conference in the beautiful original building of Université Paris Descartes. I spent the Monday attending the TPC meeting of the NOMS 2012 conference, one of very few conferences that persist with face-to-face TPC meetings. It was my first time attending a NOMS/IM TPC meeting and I found it to be a very interesting experience; in particular I was very impressed with the detail with which the TPC examined the papers and their reviews. Whilst the ranking based on review scores was a reasonably good indicator of paper quality it was heartening to see that some papers low down on the ranking ended up being accepted (for example because they had 1 unreasonably negative review), whilst others high in the rankings were not accepted (for example because of questions of novelty in relation to authors’ previously published papers). I absented myself from the discussion when my own submission was being discussed, but I was happy to later learn that it was accepted as a full technical paper. Since I am one of the Poster co-chairs for NOMS 2012 my work didn’t finish after the TPC meeting. I spent much of the rest of the week selecting papers to be presented as posters. Some of these were recommended by the TPC meeting, some were selected from the remaining full paper submissions, and 20 of the 30 submitted short papers were also accepted.
On Tuesday CNSM itself started. It’s a single track conference which once again this year attracted a large number of submissions, leading to a low acceptance rate of 14.6%. This meant that the papers presented were uniformly of high quality and authors put a lot of effort into their presentations. Presentations that most stood out for me were Imperial College’s paper on bandwidth management in Home Area Networks (my student Annie Rana is doing similar work), the joint Imperial College / IBM Research paper on refinement of management policies (which reminded me of our own previous work on the authoring and analysis in the policy Continuum) and the paper from KTH on using gossiping algorithms for resource allocation in cloud environments. My colleague Steven Davy did a very good presentation of our paper entitled “Policy-assisted Design and Deployment of Virtual Networks”, which is work that Steven plans to extend over the coming months.
Besides the paper presentations I really liked the keynotes from both Joe Hellerstein from Google and Mark Burgess from CfEngine. Joe talked about the challenges of assigning tasks to machines in Google’s data centres, giving intriguing glimpses of the complexity of the optimisation problems they are solving. Mark’s talk was a lot more high level, almost philosophical in scope, addressing the balance between automation by computers and human intervention. Mark’s slides and notes are available here.