Minutes IETF97: nmrg
minutes-97-nmrg-00
The information below is for an old version of the document.
| Meeting Minutes | Network Management (nmrg) RG Snapshot | |
|---|---|---|
| Date and time | 2016-11-15 00:30 | |
| Title | Minutes IETF97: nmrg | |
| State | Active | |
| Other versions | plain text | |
| Last updated | 2016-11-17 |
minutes-97-nmrg-00
<html>
<head>
<meta http-equiv="content-type" content="text/html; charset=utf-8">
<title></title>
</head>
<body bgcolor="#FFFFFF" text="#000000">
<pre class="agenda"><a
href="https://datatracker.ietf.org/meeting/97/session/nmrg">41st Network
Management Research Group (NMRG) meeting</a>
IETF-97
November 15th 2016 - Seoul, South Korea
09:30 - 12:00 : Tuesday Morning session I - Studio 2<br><br>Access to the
Meetecho recordings:
http://ietf97.conf.meetecho.com/index.php/Recordings#NMRG<br> <br>----------------------------------------------------
09:30 - 09:35 : <a
href="https://www.ietf.org/proceedings/97/slides/slides-97-nmrg-1-agenda-bashing-00.pdf">Agenda
Bashing</a> (Laurent Ciavaglia, Lisandro Zambenedetti Granville)
Thanks to our Jabber scribe: Wes Hardaker<span class="chat-talker"
style="color:#923541"></span> Minutes taker: Laurent Ciavaglia<br>
Number of participants: 37 on site (blue sheets) and ~10 on
line.<br><br>----------------------------------------------------
09:35 - 09:55 : <a
href="https://www.ietf.org/proceedings/97/slides/slides-97-nmrg-2-anycast-vs-ddos-evaluating-nov-30-00.pdf">Anycast
vs. DDoS: Evaluating the November 2015 Root DNS Event</a> (Moritz Muller)
Provides analysis / insights on the attack on Root DNS occurred on Nov.
30 2015. Explain anycast behavior under normal or stress conditions.
Illustrations (measurements) of effects of the attack on various
servers. Overall DNS reacted well and cope with the attack, however
some letters were more affected than others. Site flips as reaction and
implications (more or less good effects in return). Collateral damage:
an attack can affect non-targets (e.g. shared data centers). Q from ?:
did you see such behavior? A: Atlas probes do not ... Resolvers made
the changes so DNS was reacting fine. Q Lisandro: you mentioned further
research? A: develop testbeds, more insights on the resolvers behaviors
----------------------------------------------------
09:55 - 10:15 : <a
href="https://www.ietf.org/proceedings/97/slides/slides-97-nmrg-3-autonomic-networking-use-case-for-distributed-detection-of-sla-violations-00.pdf">Autonomic
Networking Use Case for Distributed Detection of SLA Violations</a> (Jeferson
Campos Nobre)
<a class="moz-txt-link-freetext"
href="https://tools.ietf.org/html/draft-irtf-nmrg-autonomic-sla-violation-detection-04">https://tools.ietf.org/html/draft-irtf-nmrg-autonomic-sla-violation-detection-04</a>
One of the Use Cases of the UCAN BoF/ANIMA WG.
Deployment of monitoring probes to measure the service levels at
network layers. Based on active probing: better accuracy, privacy. IPPM
active mechanisms. Cost of mechanisms/probes activation and coverage
optimization. In practice:
-deployment based on network engineer expertise and reactive
--> inefficient (time-wise, coverage-wise) -embedded in
management SW: lack enhancements in scalability and
Solution approach (autonomic solution): a complete solution to steer
the process of measurement probe activation. 4 advantages of the
solution. Intended user and administrator experience: lower the
requirements on expertise, still some information necessary: SLOs,
address of compliant devices. Parameters and information involved: at
the device and among devices levels. Comparison with current solutions:
no standard, scripts, passive monitoring, partial solution (faulty
links), ... Related IETF work: LMAP WG (SLA violation screening,
decrease human workload), IPFIX WG (extension to passive mechanism),
ALTO WG (measurement-based topology). Security considerations: ANIMA
bootstrapping of new devices, protection of the measurement data
exchange. Possible attacks (local DoS, forged results). Updates and
outlook: several updates in NMRG, ask RG last call. Q Alex Galis:
1-Empahsis on monitoring probes, collect a lot of information,
transport of this information for evaluation to the right
places. Did you take care of this? 2-SLA violation based on
measuring network traffic. Not the only source of information.
Extend to
A Jeferson:
1-scalability provided by the supporting peer-to-peer
infrastructure. a centralized point will be a bottleneck.
2-IPPM WG meeting, some works to model a way to monitor service
metrics, could be extendable with such new (composite) mettrics.
Q Laurent: how to configure the autonomic mechanism rightly?
A Jeferson: comparison in Cisco how to configure an IP SLA on existing
system and AN system: informally new system more simple, but not a
formal model. could be future work.
Q Alex Clem (jabber): <span class="chat-text">configuration of the
system will involve a few points: What is the likelihood with which
SLA violations will be detected (autonomic system determines how much
probing to do)
</span><span class="chat-text"> The configuration knobs are higher level:
what's the accuracy objective? How much resources can you spend?
<br></span><span
class="chat-text">----------------------------------------------------
</span>10:15 - 10:35 : <a
href="https://www.ietf.org/proceedings/97/slides/slides-97-nmrg-6-a-reference-model-for-managing-large-scale-software-defined-network-sdn-infrastructures-01.pdf">A
Reference Model for Managing Large-Scale Software-Defined Network (SDN)
Infrastructures</a> (Klaus Wehmuth)
SDN networks can become quite complex.
Multiple (sub-)layers.
How to represent / model these multi-layer networks: MAG MultiAspect
Graph and then MultiLayer Graph. Edge types: inter-layer, intra-layer,
mixed edges. Composite vertices representation. Paths are constructed
by antecessors and successors relations. Generic example for SDN
layers. Take Away Messages: Can represent SDNs with arbitrary number of
layers. Is equivalent to a directed graph. Can be represented by
matrices or any other form of direct graph representation. Can use
well-known graph algorithms for the analysis of the SDN structure e.g.
controller location, management of distributed controllers, study of
intra- and inter-layers flows, … Q Laurent: looks like a good
theoretical "tool". Any concrete examples on SDN? A Klaus: work in
progress. two types of use: off-line: What-if scenarios, and on-line:
inside controllers (working on it). Q Wes Hardaker: similar to marked
directed graph where you can have multiple attributes. in your case
attribute is the layer dimension. A: ... Q ?: data layer is also
considered multi-layer? A Klaus: yes.
<span
class="chat-text">----------------------------------------------------</span>
10:35 - 10:55 : <a
href="https://www.ietf.org/proceedings/97/slides/slides-97-nmrg-5-a-reference-model-for-autonomic-networking-01.pdf">The
ANIMA reference model</a> (Michael Behringer)
Overview of the ANMIA reference model.
Devices perform self-management.
Q Sheng Jiang (as ANIMA WG co-chair): for clarification, 6 ongoing work
items in ANIMA. reference model includes also future work items. WG at
a stage of final work on GRASP, ACP, Bootstrap, Use cases. Need to
define the semantics for the ASAs to understand they speak about the
same objective. I encourage people to make proposal on this for future
work. Q: Jeferson: use case BOF (UCAN), several use cases were
proposed. basis for future work in ANIMA phase 2. A Michael: the use
case of Jeferson would be an ASA. Jeferson: not yet modeled as an ASA
but planned for next step. To enable self-management, an important
concept is "intent" (term coined 5 years ago). Target role(s) of
devices, not device individually. Q Alex Galis:
1-how to deal with large number of intents, updated
frequently... (scaling) problem of managing intents? 2-why
orchestration are not part of the reference model?
A Michael: <br> 1- low update rate for intent. no definite
answer now. working bottom-up. <br> 2- there will be
orchestration. in SDN how to tell the switch where is their controller?
we need mix of ANIMA and SDN. ANIMA provides base services (ACP,
bootstrapping).<br><br><span
class="chat-text">----------------------------------------------------</span>
10:55 - 11:15 : <a
href="https://www.ietf.org/proceedings/97/slides/slides-97-nmrg-6-slice-networking-and-management-research-challenges-and-trends-01.pdf">Slice
Networking and Management - Research Challenges and Trends</a> (Alex Galis) <a
class="moz-txt-link-freetext"
href="https://www.ietf.org/archive/id/draft-galis-anima-autonomic-slice-networking-01.txt">https://www.ietf.org/archive/id/draft-galis-anima-autonomic-slice-networking-01.txt</a>
Description of slice networking definitions, requirements, components
and possible mapping aver ANOM, SDN, and NFV reference models. Q Julien
Maisonneuve: in EU research project, slice a first order object. what
are the justification to add this new layer/component? from the network
service viewpoint, the slices can be grouped in the different network
services. What are the real use cases that would require network
slices? A Alex: you might be right or wrong, so am I. Practical way ton
move forward, should be considered seriously. If tens/millions of
slices to deploy different services, it will be the norm rather than an
exception. progress on slice networking is happening in various SDOs
around the world. When you want to handle the management of "new"
networks, you need to include such concept also for IP networks. Based
on use case, not always. Consider the non-functional requirement. If
slice is important for operators, need to be progressed. Q Julien: at
which level to expose the slices...? e.g. in 5G NORMA with 3
orchestrators... A Alex: other project with more orchestrators.
separation of orchestration and management activities (my own list).
<span
class="chat-text">----------------------------------------------------</span>
11:15 - 11:35 : <a
href="https://www.ietf.org/proceedings/97/slides/slides-97-nmrg-7-supa-policy-based-management-framework-01.pdf">SUPA
policy-based management framework</a> (Will Liu)
Description of the SUPA Policiy Management Framework, of the SUPA
assumptions and models. Q Laurent: short on time. It would be
interesting to discuss further the relationships and implications of
the various models presented today.<br><br><span
class="chat-text">----------------------------------------------------</span>
11:35 - 12:00 : <a
href="https://www.ietf.org/proceedings/97/slides/slides-97-nmrg-8-nmrg-present-and-future-00.pdf">NMRG
Present and Future</a> (Laurent Ciavaglia, Lisandro Zambenedetti Granville)<br>
NMRG chairs presented a brief summary of wht NMRG has achieved recently
and possible evolution for the RG.<br> Call for suggestion on topics, approach
for future meetings.
Q Alex: co-locating with other academic events. ask for conference
committees about putting in place a standard-related track. The RG
chairs will follow-up on the mailing list for proposals: official
announcement (and follow-up) on topics (5G, IoT, Intent, Autonomic,
measurement...) and how they relates.
<br>--- end of meeting --- <span
style="font-size:9.0pt;mso-bidi-font-size:10.0pt;font-family: "Courier
New";mso-ansi-language:FR"><o:p></o:p></span></pre><div
class="moz-signature"><div class="WordSection1">
</div>
</div></body></html>