Minutes for ICNRG at interim-2013-icnrg-3
minutes-interim-2013-icnrg-3-1
| Meeting Minutes | Information-Centric Networking (icnrg) RG | |
|---|---|---|
| Title | Minutes for ICNRG at interim-2013-icnrg-3 | |
| State | Active | |
| Other versions | plain text | |
| Last updated | 2013-08-19 |
minutes-interim-2013-icnrg-3-1
(Action Item)
: it is necessary to track "implementations" on wiki (Dave Oran?)
Minutes of ICNRG interim meeting Hong Kong, Sunday August 11th, 2013
----------------------------------------------------------------------------------------------------------
First morning session - notes taker: Bengt Ahlgren
09:00 Welcome - Chairs
Agenda bashing - Dirk Kutscher
Welcome by the host - Jianping
09:15 Intro to ICNRG, Status of the baseline documents - Chairs
ICN research challenges - Dirk K
ICN baseline scenarios - Brje (slides by Kostas)
* would like feedback from implementers on the relevance of the document for
evaluation and comparison of different approaches * section 2 is near
completion * section 3 will need more work in the coming months * two draft
update releases planned till Vancouver * looking for volunteers to implement
in a simulator some/all of the proposed topologies/traffic load/etc and
contribute them to the community * in Berlin we agreed to split the scenario
description (sec 2) and the evaluation methodology, including metrics (sec 3)
into two separate drafts
ICN survey document - ???
10:00 SAR: Coupling Service Location with (Inter-domain) Routing, Decoupling
Them From Forwarding - Hongbin LUO Beijing Jiaotong University
(slightly different title of the talk)
Using four namespaces
- Service Identifiers (SIDs) - flat self-certifying
- Node identifiers (NIDs) - flat self-certifying
- Intra-domain routing locators - can be different in different domains
- Path identifiers (PIDs) - negotiated bi-laterally between two domains
Dave: discussing the relation to TRIAD - seems to be similar ideas
Nacho: do the domains sharing path identifiers need to be next to each other?
Nacho (slide 14): Are P5 etc the path identifiers?
Hongbin: yes
Nacho, Dave and Ashok asks about the routing (slide 17)
Hongbin: tier-1 providers need to know all service names
Dirk: requests and responses seem to be forwarded up and down in the hierarchy?
Nacho: does this happen for every request?
Hongbing: yes, but terminates at cached copies
Nacho: when a node caches something it has to be registered with the local RM?
Hongbing: that's right!
Nacho: is there only one RM per domain?
Hongbin: logically, yes
Slide 24: questions about what traffic a certain node can estimate
Dirk: you seem to have the concept of naming all the nodes?
Hongbing: that is correct
Second morning session - notes taker: Byoung-Joon (BJ) Lee
Samsung Electronics, Advanced Institute of Technology
(Action Item)
: it is necessary to track "implementations" on wiki (Dave Oran?)
11:00~12:00: Software-Defined ICN (Wen Qi from City University of Hong Kong)
- proposing generic ICN function module with unified pkt format for
forwarding,
to provide interoperability between different ICN networks via SDN
- future work will include migration to OVS prototyping
- followed by a lab demo in the afternoon, where the NDN and PURSUIT
protocols
were proprietary versions
Additional notes by Dave Oran:
- Two Usage models: (a) is interconnecting ICN islands using different ICN
architectures, (b) migrating content or clients from one ICN protocol to
another. - Approach: common/universal API, make no change at ICN clients. Then
use SDN to decouple control and data plane. Q: (Dirk) can see SDN as
implementation method, but for unification, is it enough to invent protocol
headers - don;t you have to have compatible protocol semantics - NDN and
Pursuit are totally different A: (didnÕt understand the question - response was
about how to do the routing mappings) - There is a protocol analyzer (i.e.
gateway) in the Label mapping module of the design.
Q: where is the unified packet changed into a unified packet
A: in the edge switch using a matching policy
Q: controller does it manage the edgoe or the core also
A: it does both - in the middle ther is no translation
- they move the NDN routing into the content index and topology management
modules Q: (DaveO) how do you handle large pursuit objets that break up into
multiple NDN objects A: Cache whole Puesuit object in edge where NDN
translation is done. Future work to be able to cut through. Q: (Dirk) what
about the SDN controller - need extensions to OpenFlow? A: Yes.
- How do we take this work forward
- Dirk: possibilities are numerous - this approach envisions extending OpenFlow
in various ways.
14:00~: Discussion on the ICN survey draft (Cedric Wesphal)
- also describes "service model of ICN" (Dirk K)
- history/plan
: version -00 July 2013
version -01 January 2014
- Question: how to fit "trust management" using ICN in the survey document
: ICN helps solve this problem or not?
: or, to discuss trust management issues in ICN?
- Question: scope of ICN? for the whole Future Internet? or just for "secure
content" delivery?
- Need clarification on "persistent storage"
(Action Item)
: it is necessary to track "implementations" on wiki (Dave Oran)
15:00 (?) ~ lost track of time here: Split into 4 small groups for separate
discussion
- traffic loads/scenarios, packet format, routing, etc
16:00~ : NDN workload (Ashok from Cisco)
¥ Started from IRC traces
¥ Reformatting URLs to names making them hierarchical
¥ Components of names
- test data set with realistic component count and name length
- from IRCache trace (13.5 million traces in 2007, over 8 site, etc)
: every URI beyond 28 components is junk
...and many interesting results
- presentation slide and Cisco data sets will be available on wiki (?)
- How to deal with flat names like YouTube videos?
- What about multihoming?
- What about mobility and mobile content?
-
17:00 ish~: a lab demo from City University of Hong Kong's work on
Software-Defined ICN ...end of the first ICNRG meeting in Asia
Thanks
Byoung-Joon (BJ) Lee
Samsung Electronics, Advanced Institute of Technology