Ed note: '--' refers to the slide heading presented ICNRG@IETF85 Agenda Wednesday, November 7th 2012 13:00 -- 14:30, 14:40 -- 15.40 Salon D Intro -- 5 min Note well Blue sheets *** Agenda bashing no comments *** ICNRG Work Item Status *** ICNRG Work Items -- 25 min/each -- objectives for this meeting -- interim meeting tentatively planned on 14/15 Feb 2013 *** ICN Research Challenges Fairness discussion - see wiki (Damien Saucez) http://trac.tools.ietf.org/group/irtf/trac/wiki/IcnProblemStatement Discussion on document content and outline (All) -- Damien Saucez presents Fairness between what: -- flows -- content -- customer -- link/interface fairness -- network fairness Enforcement: end-to-end vs. hop-by-hop Kostas: What is end-to-end? Is the discussion of ends "suitable"? After all, we have a different paradigm here. Not properly reflected in the figures made. Damien: there is also no notion of flow defined. Kostas: doesn't understand why we need to define flow. We need a completely new understanding of fairness. Damien does want to put forward questions, not solutions. The way the slides are ordered -- Summary: -- what is a flow? -- what is a resource? -- what fairness do we want to reach and how to measure it Mirja: Karen Sollins: why is flow on this slide? Flow implies a certain way of thinking about things. Avoid the term flow altogether. Kostas: refers to fairness in peer-to-peer networks. Boerje: a number of new types of resources comes into the picture This presentation is about going beyond flows when discussing fairness. When comparing to IP networks, we need to relate to flows. ...: You need to compare this traffic to other types of traffic. What this fairness is for is a reasonable question to ask. Lars: What is the impact of your traffic on other traffic? Is ICN aware of the capacity sharing and Re-ECN work. Dave: Key: capability of trading off caching for bandwidth. You may have to look at groups of users accessing content. Some drag down content from an upstream link, others use the stuff from the cache. These Matt: This is about mapping content networking to a different paradigm. Think OS schedulers, earlier completion time, etc. Yields a completely different sense of fairness. Lars: Contention your access to an information object causes to others who want to use the same network for other stuff. Fairness is about how you balance this out. ...: ? Dirk: comment on Damien's presentation: different ICN approaches have different concepts for transport (and flows). We should document this in the survey draft. It is difficult to discuss this on a general level. *** ICN Challenges -- document purpose -- why? -- how? -- what? -- possible results? -- input so far -- note -- proposed structure for the document 1. intro 2. problems 3. concepts 4. research challenges 5. impact on IETF work Kostas: monitoring, measurement, and instrumentation did not have traceroute and ping from the beginning -- current status *** ICN Survey document - Presentation of On the Content Retrieval In Information-Centric Network, (Will Liu) http://datatracker.ietf.org/doc/draft-ren-icn-content-retrieval Discussion on document content and outline (All) This is an excerpt of a summary report on ICN. -- motivation -- terminology Kevin: provider vs. author is a relevant distinction here. -- requirements and functionality -- summary (1/3) -- summary (2/3) -- summary (3/3) Kevin: some issues to consider -- scalability in different dimensions -- brittleness (e.g., concerning naming schemes) -- type of persistence wrt the cache Kostas: what's persistent? I may have content on a box, but I may not be able to get it off Kevin: one type of persistence: if I turn of the power, will the content survive? what would happen if I shut down an entire network Dave (from the floor): two dimensions at least to this problem -- vulnerability -- longevity Dave (chair): this may be input to some of the work items how shall we take this work forward? is this input? a starting point? Kostas: should this cover work in the IETF that happens? Lixia: what is the scope of the survey? The presentation we just heard was about file retrieval in a fairly narrow view. Information Centric Networking covers a much broader scope. What's the chairs' assessment regarding the scope of the survey? Dave: wants feedback from the group. Lixia: This should be a survey of ICN research activities, not a survey of traditional CDNs. Dave: should accommodate current research efforts But: we are not writing a 300 page monograph. Striking the balance is important. Lixia: the length of the survey is a function of the level of details in description. It should not change the scope of ICN definition. Dave: good observation. Eve: solicit input on diverse work items the RG wants to get a baseline but also be able to compare (for a particular use case) Maybe a closer look at the scenario and use cases will help. Dave: does the union of the document address Eve's concers? Karen: is concerned that there is a big elephant in the room what we mean by ICN. Agrees with Lixia on the scope to be broad. But some projects are less information-centric in the strict sense. This discussion may open up a huge can of worms. Eve: use case driven approach helps us in scoping Karen: use cases must cover a broad enough set Kostas: we should err on the side of inclusiveness. Must be contribution driven. People can best write about what they do (less about others' stuff). Lixia: yes, an RG is contribution driven. But if we do a survey, it must be a real survey of all existing work, not limited to what the authors know about. Karen commented: are we doign architecture? Lixia seconds this. Kostas: Architecture as a 110th proposal. Lixia: what's the goal for the research group? Boerje: we are free to define our goals as we go along. Lars confirms by nodding. Lars: agrees with Boerje. A venue where people can bring there topic-related research and discuss it. Maybe somebody wants to take ideas and write up an architecture. RGs don't operate by consensus. We may write up and discuss multiple ideas at the same time and may even have multiple competing publications. There is plenty of past evidence on this. Emphasizes that the RG is contribution-driven. RGs are probably worse than WGs from this perspective. Dave: # people (mostly in academia), we are getting a fairly large community. No point in having another place to do the same community stuff as elsewhere. What the special place that ICN RG (that is not redundant) can offer. One example: what's the measurement and comparison metrics? Maybe a meta architecture. How to compare different architectures? How to cut through eternal arguments between groups with different approaches. A survey document may provide some ground truth here. Karen: agrees: differences, strengths and weaknesses missing: there is no group that starts to think about what do we mean by this space. We don't have a common enough understanding of what ICN means. survey is a good starting point. Probably a living document since things evolve all the time. If you do a survey, you need to do the work. Lixia: agrees with Dave; doing architecture does not mean this RG defines its own; various designs exist but comparative evaluation is missing, which can be a task of this RG. The RG needs to have its unique value. Agrees with Karen: providing a definition of ICN makes a good contribution. Kevin: Doing an arch framework document first was a very useful starting point. Survey isn't the same and it may not be easy to extract from other approaches. Lars: on the metrics: provides example for TCP; if you want to propose a TCP extension, you need to run these simulations and present your results for certain cases. Maybe something similar can make sense here. Scenario, numbers and parameters. One way to compare different architectures Dave: good point, but TCP comparison isn't perfect: TCP was around for 15 years when this was done. We are at the starting point. So, this won't be easy. Boerje: was it a good thing to not have done this earlier for TCP? Dave: does not know the answer. Boerje: agrees with measurements and scenarios. An advantage over conferences is that we can talk about things that are not new. ...: ICN vs. CDN. Dave: Hard question: if there seems to be some emerging sense on working on comparison. Who here is willing to do some work on this? For use cases, scenarios, applications on the way towards evaluation frameworks: Bunch of hands up. Boerje: People shall sign up on the Wiki for topics and spec what they wanna do. Will organize the work around subgroups. ...: Defining terminology is really important, also to be even able to talk about architecture framework and specific designs. Boerje agrees. Contributions on this topic welcome. *** ICN Baseline Scenarios Presentation of draft-pentikousis-icn-scenarios-00.txt (Kostas Pentikousis) http://trac.tools.ietf.org/group/irtf/trac/attachment/wiki/IcnBaselineScenarios/draft-pentikousis-icn-scenarios-00.txt Discussion on document content and outline (All) Specific Proposals -- 20 min < Kostas escaped from the room just when his presentation came up :-) > -- draft goals -- draft overview -- social networking Joen (Delft U): have a ppsp reference implementation based upon a single hash request -- example -- real-time -- mobility (multi-access) -- infrastructure sharing -- content distribution -- example (CURLING project) -- ee and dtn Eve: work not just in ndn also at intel labs on home management systems (e.g., about energy usage); considering privacy. Kostas: energy efficient ICT vs. ICT for energy efficient this could be another scenario Eve promises a scenario. -- scenario topology -- for the record ... (Huawai): more discussion on what is unique on ICN (compared to IP) Dirk: nice collection of scenarios (thanks) -- we should be clear about the objective. Draft abstract says it's about comparing approaches -- current text and presentation looks more like ICN use cases. Kostas: wants to collect feedback first Volunteers: please go to the Wiki and add your name *** Coordinated Forwarding and Caching in Information-Centric Networks (Xie) -- 15 min http://sites.google.com/site/haiyongxie/draft-xie-icn-coordinated-caching-forwarding.txt Deployment Aspects -- 30 min -- contributors -- caching -- a motivating example -- a motivating example (continued) -- coordinating caching & forwarding -- leverage content popularity info -- handle consistency -- handle consistency (continued) -- summary Damien: did you look at the problem of scalability and churn? good point. Number of announcements by routers is kept small. Dave: Did you look at caches at certain sizes? content popularity? How do you know where the threshold should fall? Discussing how to deal with ICN deployment issues Discussion Presenters are not available today. No need to spend a lot of time on this today. Basically, people should think about including deployment aspects in one of the three documents we are currently working on. Might also become a separate work item to be worked upon later. But this should not prevent us from discussing it. Input is welcome! Boerje: - postpone to another meeting - separate work item, maybe later - input and discussion welcome Spirous slide on ICN vs. CDNI points (not presented) Migration Incentives Kostas: there is an article in upcoming IEEE commag Proposed reading list: ICN vs. CDN http://tools.ietf.org/id/draft-fmn-cdni-advanced-use-cases-00.txt Information-Centric Network in an ISP, http://tools.ietf.org/html/draft-li-icnrg-icn-isp-01 http://www.psirp.org/files/Deliverables/FP7-INFSO-ICT-216173-PSIRP-D4.6_FinalReportOnDeplIncBusinessModels.pdf http://hal.inria.fr/docs/00/68/44/58/PDF/paper.pdf Other Business -- 20 min ---------------------------------------------------------------------- ICNRG interim meeting Boerje: - want to have longer meetings than just 2 hours at IETF - proposing meeting in February (14-15) in Stockholm, i.e. slide 1 day compared to previous announcement thus having it Thursday-Friday. The co-located SAIL Efraim workshop would then be on the Wednesday. Final decision will be announced on mailing list and Wiki very soon. Dave: - trying to avoid conflict with NSF meeting (presumably Mon, Tue) - will assess interest to attend on Wiki comments: Lixia: - The proposed interim meeting date can be difficult for US academics to attend Dave: - other proposals or hosting offers? Boerje: decide later about Orlando meeting ---------------------------------------------------------------------- NetInf protocol draft announcement ---------------------------------------------------------------------- AOB - nothing