33rd Network Management Research Group (NMRG) meeting
IETF 89
March 6th 2014 - London, UK

13:00 - 15:00 : Thursday Afternoon Session I - Richmond/Chelsea/Tower
- Around 40 people attended the meeting
- Q: question, A: answer, C: comment

13:00 - 13:05 : Introduction by Lisandro Granville
- Note well
- Objective of the meeting
  . Continue the discussion on autonomics for network management, started in the IETF 88, Vancouver
  . Presentations will show the first outcomes on autonomics for network management

13:05 - 13:30 : Proactive self-healing mechanisms for IP networks by Laurent Ciavaglia
- Q: Can this solution incorporate physical parameters
- A: Yes, since the set of parameters taken into account can be extended

- C: Take a look at microwave networks. Depending on the time window, reactions can be lost

- Q: Can the network evolve to an unstable state if all nodes take unsynchronized decisions?
- A: That has not been investigated yet.

- Q: How long does it take to train the system?
- A: No learning support! Knowledge is explicitly informed by the operator.

- Q: How do you build a more general model?
- A: To be exploited.

- Q: How the operator/user interacts with this solution?
- A: No specific user interface has been defined. Refer to slide 17.
- C: We should provide high-level user interfaces. Policies could be considered.

- Q: Should this effort move on? (from Michael to the audience)
- A: Definitely yes (from the audience)

13:30 - 14:00 : Autonomic Network Definitions by Michael Behringer
http://tools.ietf.org/html/draft-irtf-nmrg-autonomic-network-definitions-00
- Q: Is it receiver-driven or sender-driven?
- A: It’s event-driven and the target network understands the intends

- C: Configuration should exist to say WHAT to do and NOT HOW to do

- Q: Can the autonomic loops be configured or tuned?
- A: Yes. We should have meta-management, i.e., management of the management loops

14:00 - 14:30 : Gap Analysis for Autonomous Networking by Brian Carpenter and Sheng Jiang
http://tools.ietf.org/html/draft-jiang-nmrg-an-gap-analysis-00
- Q: Can a network learn from the external world?
- A: Yes, the network can learn from the intends. But it also depends on the language used to express intends.

- C: In fact that already exists, for example, in OSPF.
- C: An intend defines WHAT the network is expected to accomplish, while the knowledge/intelligence of the network provides the HOW’s to accomplish the intends

- C: Humans can be present just to agree with a network decisions. One the operator trusts the network decisions, he/she gradually leaves the management loop.
- C: In the set of loops, as well on the loops’ life-cycle (e.g., bootstrap, regular operation, adjustments/tuning), define WHERE (in which loops?) and WHEN (in the loops’ life-cycle) should humans be present

- Q: Is there a real-life implementation to be shared?
- A: That will be present in the NMRG draft.

14:30 - 15:00 : Probe Placement Problem by Dimitri Papadimitriou
- Q: Do you assume that there are probes already available in the network ready to be activated?
- A: Yes.

- Q: Why is this specific to probe placement? Isn’t it a more general problem?
- A: To be investigated.