10:00 - 10:05 Agenda Bashing
10:05 - 10:10 Adding Analytics to I2RS protocol
10:10 - 10:20 Conflict Resolution [Pedro Guiterrez]
10:20 - 10:30 Discussion on Conflict Resolution
10:30 - 10:50 I2RS Filter-Based RIB [Sue Hares]
10:50 - 11:00 FB-RIB Discussion
11:00 - 11:20 Service Topology
11:20 - 11:30 Service Topology Discussion
Sue Hares running the meeting
presentations at https://www.ietf.org/proceedings/interim/2016/01/13/i2rs/proceedings.html
purpose - interactive discussions
JH: 3 different aspects - conflict resolution, collecting data, analytics
SH: analytics not part of I2RS, but we need to collect enough data to allow for analytics
JH: may need more sophisticated DMs to allow for analytics - this is within scope
NetIDE Intermediate Protocol - supports currently NETCONF, OpenFlow
Composed applications - triggered by Network Events, result in config operations
Conflict handling - define when it arrises and resolution strategies - e.g. by weights of modules
lessons: match network commands to network event, it matters
global knowledge - when it arises, if they can be resolved,how>
is this line of work interesting?
JH: observations about atomicity consistent with what we observed
where conditions apply? For example, you may have two different events when receiving.
Pedro: I agree that there are multiple events on input. We are looking at the Open Flow Controller.
SH: what other types of events
PA: OpenFlow packet arrival
JH: so what controller does, it does in response to packet in? We may have different events for I2RS that are different from OpenFlow.
PA: We are starting with the Open-Flow triggers.
JH: This makes sense. There is enough conflict in what you started with.
Sue: Lots of events with packet events, and beyond for I2NSF.
Pedro: Thank you for the feedback. It will help this project and other project
Filter-based topology RIBs
I2RS filter-based RIB is not only filter-based forwarding
minimalsitic ECA policy
it's not doing all the customer wants
can we align filters at different levels?
N-Tuple filters - standardized?
FB-RIB - ephemeral still in discussions with NETCONF
Filters in I2RS FB-RIB
how they are stored if BGP
flow specification actions
more proposed filter tyoes (e.g. nvo3)
two ways to combine rules - order, specify in rules
BGP FlowSpec vs. I2RS filter
neet to have opstate
DR: hot discussion about opstate just ongoing in NETCONF - need contributions and cross-discussion
why discuss today?
I2RS SM aligns with L3SM?
JH: L3 service model in not to be used beyond I2RS
service topology model may look differently than the model that initially used - service atop L2VPN / L3VPN - one more layer of topology
bottom-up, top-down, or combination
should we be doing a bottom-up combination of L3VPN and other things? other difficulties?
JH: service chaning is an overlay topology
SH: maybe we should rename service as overlay
Xufeng: not really close to the overlay model that they do
DF: inyour diagram - where is layer 2?
optical missing - atop of optical
DF: peering in overlay capability is not captured
SH: need to read back some of the documents - may cover the logical links that instantiate peering