Learning-Capable Communication Networks (LCCN) IETF 81 Meeting Quebec City, Canada July 28, 2011 - PowerPoint PPT Presentation

Loading...

PPT – Learning-Capable Communication Networks (LCCN) IETF 81 Meeting Quebec City, Canada July 28, 2011 PowerPoint presentation | free to download - id: 734113-NDJhZ



Loading


The Adobe Flash plugin is needed to view this content

Get the plugin now

View by Category
About This Presentation
Title:

Learning-Capable Communication Networks (LCCN) IETF 81 Meeting Quebec City, Canada July 28, 2011

Description:

Learning-Capable Communication Networks (LCCN) IETF 81 Meeting Quebec City, Canada July 28, 2011 Dimitri Papadimitriou (dimitri.papadimitriou_at_alcatel-lucent.com) – PowerPoint PPT presentation

Number of Views:18
Avg rating:3.0/5.0
Slides: 10
Provided by: Dimitr48
Learn more at: http://www.ietf.org
Category:

less

Write a Comment
User Comments (0)
Transcript and Presenter's Notes

Title: Learning-Capable Communication Networks (LCCN) IETF 81 Meeting Quebec City, Canada July 28, 2011


1
Learning-Capable Communication Networks
(LCCN)IETF 81 Meeting Quebec City,
CanadaJuly 28, 2011
  • Dimitri Papadimitriou (dimitri.papadimitriou_at_alcat
    el-lucent.com)
  • Wouter Tavernier (wouter.tavernier_at_intec.ugent.be)
  • Didier Colle (didier.colle_at_intec.ugent.be)

2
(Self-adaptive) Learning-based Control
  • Current situation once configured, network
    systems follow explicitly pre-defined behavior,
    persistently decide and uniformly execute
  • gt Self-adaptive learning-based control
  • Aware of their state, activity/behavior, and
    environment over time
  • Capable to react and adapt to changing
    environment and running conditions over time
  • Operate autonomously and when needed
    cooperatively (no global coordination or
    synchronization)

Learning to avoid need to iteratively re-design
systems as they should adapt to their environment
and running conditions
  • Role/challenges of learning
  • WHAT/WHY diagnose internal state, own activity,
    and network environment over time (detect -gt
    identify -gt analyze)
  • HOW adapt decisions and tune actions
    automatically and timely
  • WHEN determine when to operate autonomously and
    when to cooperate

3
(Self-adaptive) Learning-based Control
Augment network systems (e.g. routers) with
learning component
Guiding principles
  • Local view and decision (no network global view
    or coordination) to ensure scalability and
    robustness
  • Modular (no monolithic or unified framework) to
    ensure gradual development
  • Distribution (no uniform or ubiquitous plane) to
    ensure organic deployment

4
Self-X paradigms vs functionality
Functionality
Self-configuration
Self-maintenance
Self-stabilization
Self-diagnosis
Self-protection
Self-healing
Self-repair
Self-optimization
Paradigm
Self-management
Self-adaptation
Self-organization
5
LCCN Research Activity Positioning
Self management capability to operate and
maintain systems and devices depending on the
current system parameters Properties Unified
framework and global coordination (and
re-evaluation) Technique e.g. policy-based
management
Self-management
Self-adaptive ability of the system's components
to adapt to changing environmental and running
conditions over time Properties Distributed,
modular, and local coordination (and
re-evaluation) Technique e.g. learning
LCCN
Self-adaptive
Self-organize structure and functionality at the
global level emerge solely from numerous
interactions among lower-level components without
any external or centralized control Technique
e.g. biology, neuro-science
Self-organizing
6
Activities Timeline
Draft charter update
First Draft charter
11-10
03-11
01-11
07-11
05-11
IETF 79
IETF 80
IETF 81
Doodle Poll
PS doc.
First Get together
LCCN Meeting
https//sites.google.com/site/lccngroup/meetings/p
re-bof-meeting
LCCN Meeting
https//sites.google.com/site/lccngroup/meetings/m
eeting-ietf-81
7
Research Space/Domains
  • Doodle poll (25, about 50 of participants)
  • Outcome several people will present
    state-of-the-art/challenges in these selected
    experimental domains during LCCN meeting _at_ IETF
    81
  • Identify research challenges and potentials wrt
    application of various learning models and
    techniques inside each of these domains and
    between domains (specialization vs
    generalization)
  • Architectural baseline (common principles,
    control model, etc.)

  Network self-defense (incl. distributed intrusion/attack detection, distributed anomaly detection, etc.) Routing (incl. traffic engineering, resiliency, stability, etc.) Control of optical networks Mobile/Ad-Hoc/Sensor networks Other Suggestion
Count 11 21 10 10 5
8
LCCN Initiative
  • One-line abstract learning-based control for
    network systems to automatically and timely adapt
    to network environment and running conditions
    over time
  • Description lthttps//sites.google.com/site/lccngr
    oup/gt
  • LCCN Meeting Wednesday July 27th from 100 to
    300PM (303AB)
  • Goals (of this meeting)
  • Identify research challenges potentials when
    applying learning to i) Routing (incl. traffic
    engineering, resiliency, stability, etc.), ii)
    Network self-defense (incl. distributed
    intrusion/attack detection, distributed anomaly
    detection, etc.), iii) Mobile/ad-hoc/sensor
    networks.
  • Assess level of interest to initiate research on
    them
  • Determine if IRTF is the right place to conduct
    such research work
  • Estimate readiness for opening an IRTF Research
    Group by IETF 82
  • Agenda lthttps//sites.google.com/site/lccngroup/b
    of-description-and-agendagt
  • Problem statement I-D draft-tavernier-irtf-lccn-p
    roblem-statement-01
  • Contact lthttp//groups.google.com/group/lccn/gt

9
LCCN Initiative
LCCN Mailing list countries (60 members)
About PowerShow.com