OPS-AREA Open meeting minutes
(based on notes taken by Linda Dumbar and Bert Wijnen)

Monday March 28, 2011
Morning Session 

Area Directors and Meeting Chairs: Dan Romascanu and Ron Bonica (remote)

agenda: http://www.ietf.org/proceedings/80/agenda/opsarea.txt
meeting-materials: https://datatracker.ietf.org/meeting/80/materials.html#wg-opsarea

The meeting was hold in a common session with the OPSAWG. 

1. Note well, agenda bashing, note takers, jabber scribe, blue sheets - 5 min 
 
2. Expressing identity: Bruce Nordman, 15 min

- Energy management systems (buildings)
- Gather information from ALL (mains powered) devices on the IP network (and some beyond)
- Usually have no other relationship to devices
- Mechanisms need to be universal
- Many buildings will be unmanaged (all partly)
- Devices come and go synchronously
- Identity-what it is? Whatever makes an entity definable and recognizable, in terms of possessing a set of: 
-- Species: e.g. switch server, 
-- Origin: brand X, model Y
-- Who: name, network address/
- Purpose-provide basic characterization of type of device-primary function
- Not intended to be specific or descriptive
- Not intended to replace any existing mechanisms
- Expect identity to be useful in other contexts
- Internet of Things - In future, most IP devices will not be traditional (IT-oriented) ones found on network today
- Questions
     - What are all relevant MIBs
     - Are there other relevant non-IETF references
     - Does/should any pf types reside outside eman MIBs
     - Any barriers doing this

Mehmet Ersue : this is the identity not done by IANA. We should talk with Geopriv who has been working on this for over 10 years, on location, name, and identity. Maybe better to use a name space instead of IANA registry. Maybe should have a barBOF on this topic. 
Wes Hardaker: I have done fair amount of work in enterprise world. We started with naming a device. We found that a single name is not enough. E.g. cell phone & toast plus a name. My cell phone is different from your cell phone, and different version. Management system needs to know the name and version. Need a table to show type, version, and names. Single name is not enough anymore. We started with global identity, but found out that more detailed. Managers need to exactly understand the specifics of the device. Maybe a table: I am a device of class X and I am at version n
Bruce Nordman - Trying to understand what energy is doing

Melinda Shore - University of Alaska: Need to add more attributes to identity. There has been a lot of work done in the SEC area. 
General issue: energy section know very little about this. 
Bruce - Even having imperfect info is enormous step forward from where we are today in energy world.
Chris Liljenstolpe (jabber) agrees with Wes. We need a label space, rather than an identity, a device will need one or more labels.
Dan Romascanu -  as AD hopes to charter this work soon; as contributor: Identity is not a one dimensional thing
      --  basic level: a label that anybody can fill in
      -- for identity: we need more dimensions (device type, version, location etc)
Bruce - we may only need a small set of
Dan - OK, so start defining the subset you need
Juergen Quittek - Identity in this context seems to be more of a classification of devices
Bruce -  So Identity is probably the wrong word to use
Juergen - Maybe this work (categorization) needs to be done OPSAREA wide
Roman Arch -  If you are looking for very broad, maybe it should be based on the amount of energy consumed, like > 100 watts. 
Moli from Cisco: from EMAN?s point of view, there are a lot of attributes already defined in EMAN. Some of the concept (vendor type, etc) are very common. 
Dan - show of hands to believe it is a useful work for OpWg? 
Around 10 people raised their hands, a couple think it is too much of "boiling ocean"
Aaron Falk - it is interesting piece of work. What Bruce wants is the identity for EMAN, instead of broad range of identity. 
Dan - suggest to write a draft (individual), and do not use term "identity" - Then we can all form a better understanding/opinion
 
3. Learning-based control plane: Dimitri Papadimitriou, 5 min

- Learning ?based control: From static config and pre-determined -decision/executions to learning based control adaptive to environment and running conditions over time
- Role/challenges of learning 
- Diagnose; Internal state, own activity, and network environment over time (detect -> identify -> analyze)
- Adapt decisions and Tune actions in a cost ?effective and timely way
- Determine to operate autonomously and when to cooperate
- Local view (no network global view): ensure scalability and robustness
- Modular (no monolithic or unified plane): ensure gradual development
- Distribution (no uniform or ubiquitous plane): ensure organic deployment
- LCCN initiative:
-- Goals - Community building and consensus building on problem statement and identified challengers
-- Assess level of interest to initia research on these problems/challengers
-- Determine if IRTF is the right place to conduct research work to address 
-- Estimate readiness for BOF request for IETF 81
- LCCn events @IETF 80
-- Get together; Monday at 7:45pm (registration desk)
-- Meeting: Wed march 30: 8pm (Karlin 1)
-- Agenda: https://sites.google.com/site/lccngroup/bof-description-and-agenda> 
-- Document: problem statement: draft-tavernier-irtf-lccn-problem-statement-01
-- Contact: http://groups/google.com/group/lccn/

Juergen Quittek - I did not see there is any learning involved yet
Dimitri - that will be discussed in the meeting
Juergen - Pls tell us HOW you learn.
Dimitri -  The learning we have is used in order to predict the value of the decision
Juergen -  is it a learning mechanisms that is not used thus far
Dimitri - yes. And this cannot be done by centralized NM systems. Assume each entity has its own control. The idea is to how to control them. It is to tell them what to do  
Dan - Clarification Question: is this BOF or IRTF?
Dimitri - It is IRTF space right now
Tina Tsou - from this self-learning protocol, you can?t see Network protocol anymore? 
Dimitri - the idea is how to merge it to NMS in the future. When to decide 
It is not to re-do the NMS plane. There is no intension to run this instead of another. Even if you have enabled this component, we are thinking how this can be integrated with existing NM. Maybe delegate some decisions to the learning component. Also there might be situations where the learning component needs help from NM system.
Aaron Falk: (IRTF chair): This is reseach. This is a proposal to create a RG. Is there energy and interest?It is only IRTF if it is not something that is already understood and has known solutions. So possibly at the next IETF there is a BOF to see if we want to start such an IRTF RG.
DR: comments can be posted on OPSAREA mailing list, Dan will forward to IRTF and topic list.

4. Open mike:

nothing brought up